The effect of different types of
communication medium on the
performance of brokers
Master’s ThesisMSc. In Business Administration – Strategy Track
Student: Alex Rood, 10894411
University of Amsterdam, Faculty of Economics and Business Supervisor: Dr. Nathan Betancourt
University of Amsterdam, Faculty of Economics and Business Date of submission: 24-06-2016
Statement of Originality
This document is written by Student Alex Rood who declares to take full responsibility for the contents of this document.
I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it. The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.
Abstract
Choosing the right communication medium might affect the performance of a network broker. This study explains the effects of different types of communication medium on the
performance of brokers. The sample of this study consists of authors and their papers published in the Reviews of Modern Physics (RMP) journal between 1929 and 2013.
Analyses were conducted on a database containing information about authors and a database containing information about papers written by brokers. Contrary to current literature, this research showed that brokers were not consequently outperforming non-brokers. Brokers outperformed non-brokers on some aspects, but not all. The usage of more face-to-face communication by brokers has a positive effect on their performance. A logical explanation for this is that brokers who used face-to-face communication gained more trust from potential partners.
Table of contents
Introduction ... 1 Literature review ... 2 Social networks ... 3 Brokers ... 3 Communication media ... 6 Literature gap ... 8 Theoretical framework ... 9 Method ... 12 Authors database ... 13Brokers’ papers database ... 17
Results ... 18
Authors ... 19
Brokers’ papers ... 22
Discussion ... 26
Conclusion ... 30
Appendix 1 – Network visualization ... 32
1
Introduction
Many scholars explain how actors that obtained broker positions in a network exhibit superior performance (Burt, 1997; 2004; Podolny & Page, 1998). Brokers are those actors in the network that make a connection between two previously unrelated groups (or clusters) of people. A broker makes such a connection by interacting with people from different clusters. By doing accordingly he is building so called bridges. When brokers build bridges between clusters of people, they have access to information and control information flows (Burt, 1997; 2004; Podolny & Page, 1998; Burt & Talmud, 1993).
Because of the advances in communication technology, there is an increasing number of communication medium brokers can choose from when they want to interact with other actors in a network. Before the invention of the telephone and the telegraph, major forms of communication were speaking to each other face-to-face, or sending a letter. While sending a letter might reach people distant to you, it still costs a lot of time and energy compared to speaking with someone face-to-face (Dimmick, et al., 2000; Treviñio, et al., 2000). But, since the invention of the telephone in 1876, a lot of new communication media have been
developed. The internet has provided us with many new communication media, such as email and text messages, that sends our messages across the earth within seconds. While these are good ways to reach people distant to you, they are limiting you in the amount of information you send in comparison to face-to-face communication (Daft & Lengel, 1986). Each different communication medium has different characteristics that make it appropriate in some
situations, but maybe not in others (Lengel & Daft, 1989). Some media might be better at distributing empathy and facilitating mutual knowledge and understanding (Dimmick, et al., 2000), while other communication media enable actors to communicate while living far apart and maybe even in different time zones (Kiesler & Cummings, 2002).
2
The communicator should keep in mind that the message is affected by the chosen
communication medium (Lengel & Daft, 1989). When a message is ambiguous, a consensus on the interpretation has to be reached. In this case, a medium high on richness is preferred (Dennis & Valacich, 1999). When a message is high on uncertainty, additional information is needed and a leaner medium will be more preferred (Dennis & Valacich, 1999). For this reason, choosing the right communication medium might affect the performance of a broker. This study will examine the effects of different types of communication medium on the performance of brokers with the following research question:
What is the effect of different types of communication medium on a broker’s performance?
The results of this study will explain more about the differences between high
performing brokers and low performing brokers. It will also give more insight in the effect of a chosen communication medium. This will enable brokers to choose more consciously the form of communication that will best facilitate their needs and will help reach their goals.
The paper is structured as follows. First, existing literature will be reviewed. Second, the way the analyses were conducted are elaborated in the method section. Third, the results of the different analyses will be presented, followed by the discussion. Lastly, the main conclusions of this paper will be presented.
Literature review
This part will discuss relevant literature for this study. First, the information about social networks available in current literature will be discussed. Next the main findings about brokers in current literature is presented. Finally, the concept of communication medium will be discussed.
3
Social networks
A social network is a set of socially relevant network members (nodes) connected (or tied) by one or more types of relations (Wasserman & Faust, 1994; Marin & Wellman, 2011). In principle, any unit connected to another unit can be studied as nodes in a network (Marin & Wellman, 2011). For example, White, et al. (2004) used journal articles as nodes and citations as their connections. When two nodes have an information-carrying connection there is a tie between them (Granovetter, 1973). According to Burt (1997) a node’s connections is social capital that provides opportunities such as, among others, influence, participation in important projects, and promotions. Bourdieu and Wacquant (1992, p. 119) define social capital as “the
sum of the resources, actual or virtual, that accrue to an individual or a group by virtue of possessing a durable network of more or less institutionalized relationships of mutual acquaintance and recognition”.
In a network, nodes will never be perfectly connected, as there will often be missing ties that can be considered ‘holes’ in the flow of information (Burt, 1992). As people make connections and build their network, it is likely that they do so with people like themselves, socially similar people spend time in the same places, have more shared interests, and are more attracted to one another (Burt, 2009). As people build their network in such a way, groups will form and structural holes will emerge. There is a structural hole between two groups when they are not connected to each other and, thus, do not share information (Pollock, et al., 2004).
Brokers
The absence of connections between groups creates opportunities for brokers (Burt, 2001). Brokers are those actors in the network that make a connection between two previously unrelated groups or clusters of people (Burt, 1992). A broker makes such a connection by interacting with people from different clusters. By doing accordingly he is builds so called
4
bridges that gives access to non-redundant information flows. Non-redundant information is information that is more additive than overlapping to the already obtained information (Burt, 2001). When people in a network are strongly connected to each other or when people link to the same third parties, they have similar information. As a result, they only provide redundant information (Burt, 2001). When brokers build bridges between groups, they have access to non-redundant information and are able to control the flow of information. This causes brokers to exhibit superior performance and personal advantages such as higher
compensation, better performance evaluations, and faster promotions (Burt, 1997; 2001; 2004; Podolny & Page, 1998; Burt, et al., 2013).
While brokers are associated with better performance, there are performance differences among brokers. Burt (2013) explained that even when brokers have the same number of connections, one might be advantaged compared to others because of his position in the network. When a broker in a network has the same number of connections as another, but has relatively more connections that are network bridges connecting to other groups, it will give him access to less redundant information. This can lead to better timing, and has more influence in whose interests are served when contacts come together (Burt, et al., 2013). In addition, Levin & Cross (2004) found that trust is an important mediator for transferring knowledge, especially for strong ties. This may imply that even though there are multiple ties, not all of them are equally efficient in transferring knowledge. Some scholars argue that personality combined with someone’s position in the network also affects performance (Mehra, et al., 2001). This means that the advantages of brokers do not result from access to diverse information alone, it is a by-product of processing diverse information (Burt, et al., 2013). Intellectual and emotional skills are the real advantages that separate high performing brokers from low performing brokers (Burt, et al., 2013).
5
A number of studies show why some people become brokers. People that are high self-monitors are more likely than low self-monitors to fulfill a broker role (Sasovova, et al., 2010). There is also evidence for the relation between self-monitoring orientation and organizational performance (Mehra, et al., 2001). The personality of a broker can affect the way he or she builds his network over time (Mehra, et al., 2001). For example, Brass (1985) found that informal interactions are important for the amount of influence and promotions.
It could be that the position of an actor in the social structure matters for obtaining rewards (Sørensen, 1996), but some scholars argue that better-performing actors came to occupy broker positions because of their ability to secure the opportunities that lead to these positions (Lee, 2010). Actors with superior quality should be able to identify and exploit the opportunities coming from non-redundant ties. Therefore they are more likely to fulfill a brokerage role, implying that it is not only their position that enables them to outperform non-brokers (Lee, 2010).
While the absence of connections between sparse networks create an opportunity for brokers, it also causes an implementation problem. Dense networks are better at coordinating action, which is needed when implementing innovations (Obstfeld, 2005). But, Obstfeld (2005) notes that dense networks also make idea generation more difficult, which leads us back to the need for brokers.
Granovetter (1973) explained that there are differences in the strength of ties. The strength of a tie is defined as: “a (probably linear) combination of the amount of time, the
emotional intensity, the intimacy (mutual confiding), and the reciprocal services which characterize the tie (Granovetter, 1973, p. 1361)”. Strong ties, according to Granovetter,
cannot be a source of redundant information. Weak ties however, can provide non-redundant information (Granovetter, 1973; Levin & Cross, 2004). Strong ties are not a source of non-redundant information, because when person A has a strong tie with person B and
6
person C, chances are there is a tie between B and C as well. In this case, A has redundant information from B, because C is also able to interact and obtain information from B.
Communication media
Kiesler & Cummings (2002) found that many distributed work groups are adapting their interactions to new communication technologies. These technologies allow for the exchange of work information without face-to-face communication and for spontaneous communication (Kiesler & Cummings, 2002). It is however important to note that some communication media might be better at distributing empathy and facilitating mutual knowledge and
understanding, as Dimmick, et al. (2000) explain in their research. For example, when people need to have the same understanding about a certain topic, expressing emotions might help to perceive the intended meaning of a message. In this case, Dimmick et al., (2000) found that the telephone will be better than e-mail at conveying emotions and reaching the goal of understanding. Some media better enable the process of communication and make it more efficient. For example, the phone is better for some purposes because of the efficacy of a familiar human voice in conveying affect or emotion in real time (Dimmick, et al., 2000). Even though with e-mail there are certain symbols, for example emoticons, this type of communication medium does not carry the load of information that the sound of a voice does (Dimmick, et al., 2000).
Lengel & Daft (1989) created the Media Richness Theory, which argues that the performance of a task will be improved when the task’s needs are matched to a medium’s richness.
Richness is defined as “the information-carrying capacity of data (Daft & Lengel, 1986, p. 11)”. If data gives a lot of understanding it is high in richness. When data provides little understanding it is considered low in richness. Understanding information of a message is influenced by equivocality and uncertainty (Daft & Lengel, 1986). According to Daft, et al. (1987) managers that are able to understand and choose a communication media on the basis
7
of the uncertainty and equivocality of a message are performing better. Some messages that an actor wants to send can be interpretated in multiple and conflicting ways, also called equivocality. Equivocality is high when there are multiple interpretations of a message (i.e. when a message is confusing and ambiguous). When a message is high on equivocality, a consensus on interpretation has to be reached. In this case, a communication medium high on richness is preferred (Daft & Lengel, 1986). Uncertainty exists when the message has little information (Dennis & Valacich, 1999). When more information becomes available, uncertainty decreases. So, when a message is high on uncertainty, additional information is needed and a leaner medium will be more preferred (Dennis & Valacich, 1999). With this insight Daft and Lengel (1986) created a Media Richness Hierarchy which explains how different communication media diverge in their ability to transfer information. Physical Presence, or face-to-face communication is the richest form of communication. Next in the hierarchy are Interactive Media (e.g. telephone). On the third place in the hierarchy are Personal Static Media, for example letters or texts. At the bottom of the hierarchy, and providing only limited information are the Impersonal Static Media, such as flyers. The differences in classification are mainly chosen by the differences in a medium’s ability for immediate feedback (e.g. to check interpretation), the number of cues and channels utilized (i.e. ways in which information could be communicated e.g. body language, tone of voice), personalization (i.e. the ability to personalize a message), and language variety (i.e. ability to convey natural language rather than just numeric information).
Even though face-to-face communication is richest in information transfer, there are also benefits when using non-face-to-face forms of communication. E-mail is better at fitting in with people’s work schedules and communicating with people in different time zones (Dimmick, et al., 2000). Actors in a network that need to engage in long-distance communication are more likely to choose email and other written media as opposed to
8
meeting face-to-face (Treviñio, et al., 2000). Written communication media help to
communicate with people that are geographically far away, or for which you don’t have time to meet in person (Dimmick, et al., 2000). Email and other written communication media are also useful for sending a message to a large number of recipients (Treviñio, et al., 2000). According to Dimmick, et al., (2000) interpersonal-relationship activities can be successfully accomplished with e-mail just as with other forms of interactive medium. However, Kraut, et al. (1998) found that greater use of the internet leads to a decrease in the size of one’s local social network.
It may be that particular relationship strategies are better suited to particular media forms due to their reliance on social context cues and synchronicity (Dimmick, et al., 2000). It should however be noted that for some media to be used, some form of knowledge and skills is required. For example, the effort and experience it takes to use e-mail might make it less preferable for some people (Dimmick, et al., 2000; Treviñio, et al., 2000). Also, typing and reading are more time consuming than speaking and listening, dragging the efficiency of communication by text (Siegel, et al., 1986).
Literature gap
A lot of research has been done on the advantages of having a broker position, the individual characteristics and personality of brokers, and the differences between communication media. However, still little is known about how a communication medium might impact the
performance of brokers. It is unknown how a communication medium affects a broker’s performance and its consequences such as higher compensation, better performance evaluations, and faster promotions. Different communication media might lead to different levels of a broker’s performance.
9
Theoretical framework
A social network is a set of socially relevant network members (nodes) connected (or tied) by one or more types of relations (Wasserman & Faust, 1994; Marin & Wellman, 2011). This study will regard authors of scientific papers as the nodes, and collaboration of authors on a paper as their connection. This means that when two authors worked together on a paper, this study will regard them as having a tie.
In order for this study to examine brokers, they first have to be identified. Burt (1992; 2001; 2004) used network constraint to measure brokerage. Network constraint measures the extent to which an author’s network contains redundant contacts (Burt, 1992). It can be seen as an inverse measure of social capital. Network constraint decreases when there is an increase in strength of connections and more sharing of information among actors of the network. Network constraint has been associated with different kinds of performance such as better performance evaluations, faster promotions, and higher compensation (Burt, 2004). Burt (2004) used, among others, staff officers within several divisions of a large financial organization, senior managers in a large electronics manufacturer, and senior managers across functions in a division of a chemical and pharmaceuticals company as samples for his studies. This study will use authors of scientific papers published in a scientific journal as the sample. While staff officers and managers are likely to pursue better performance evaluations, faster promotions, and higher compensation, authors of scientific papers might have other goals. The number of publications in peer-reviewed journals and the times an author has been cited has been commonly used to assess an author (Bedeian, et al., 2009). More publications in a peer-reviewed journal and more citations are seen as indicators of performance (Bedeian, et al., 2009). More publications and citations means that authors have produced work of higher quality, according to (Bedeian, et al. (2009). For this study, publications and citations will be
10
used as proxies for an author’s performance. Together with the notion that brokers exhibit superior performance, the following hypotheses will be tested:
Hypothesis 1: Authors of scientific papers in a brokerage position publish more papers than authors that are not in a brokerage position.
Hypothesis 2: Authors of scientific papers in a brokerage position get cited more than authors that are not in a brokerage position.
Since the invention of the telephone in 1876 and the World Wide Web in 1955, new ways of communication are constantly being invented. New communication media create new forms of interaction and new kinds of social relationships between people (Crowley & Mitchell, 1994). Crowley and Mitchell (1994) distinguish different types of communication. For this study we will use face-to-face and mediated (or non-face-to-face) communication. When communicating face-to-face, the participants are in the presence of one another. When engaging in face-to-face communication the participants use a multiplicity of symbolic cues such as gestures, words, and facial expressions to convey messages and interpret the messages sent by others (Crowley & Mitchell, 1994). Mediated communication is communication with the use of a technical medium (e.g. paper, electrical wires, electromagnetic waves, etc.) that enables the spreading of information to others who are remote in space, time, or both
(Crowley & Mitchell, 1994). Sending a letter, an e-mail, a text message, or having a telephone conversation are examples of non-face-to-face communication.
As technology developed, more forms of non-face-to-face communication with different benefits became available, partially substituting face-to-face communication. This might have made working with people from different cities increasingly easier. When it
11
becomes easier for people from different cities to collaborate, the amount of potential people someone can collaborate with increases. With this increase in potential collaborations, the chance to find the right person to collaborate with outside the author’s city also increases. In time, as more non-face-to-face communication media became widely known and used, more people from different cities were able to work together. This leads to the following
hypothesis:
Hypothesis 3: The year of publication has a positive effect on the percentage of people from different cities working together.
New non-face-to-face communication media, such as email and text messages, send our messages across the earth within seconds. While these are good ways to reach distant people, they are limiting the amount of information sent in comparison to face-to-face
communication. Different types of communication medium have different characteristics that make it appropriate to use in some situations but maybe not in others (Lengel & Daft, 1989). Lengel and Daft (1986) argue that a message is affected by the communication medium that is used. As such, performance can be affected by the chosen type of communication medium. Because we use the number of publications and citations as an indicator for performance, the following hypotheses will be tested:
Hypothesis 4: The percentage of people from different cities working together is related to the amount of publications of a broker.
Hypothesis 5: The percentage of people from different cities working together is related to the number of times a broker has been cited.
12
Besides looking at the overall performance per broker, information about the effect of a chosen type of communication medium on the performance of individual papers, written by brokers, also helps to answer the research question. Therefore, the following hypothesis will be investigated.
Hypothesis 6: The percentage of people from different cities working together is related to the number of times a broker’s paper has been cited.
Method
The sample of this study consists of authors and their papers published in the Reviews of Modern Physics (RMP) journal between 1929 and 2013. In total, 3140 papers were included in the dataset. These papers were written by a total of 4780 authors. The RMP journal was chosen because it contained the most information needed for this research. Other journals lacked some information, for example which university authors were affiliated to. Some information was already available in an existing database. This database contained information about papers published in RMP, their authors, their date of publication, and which previously published papers they cited. With this data a new database was created. Additional information, such as the affiliated university of each author, was retrieved by consulting the Reviews of Modern Physics website (Reviews of Modern Physics, 2016) and added to the database.
This database only contains the number of papers published by an author, and the total amount of times his papers have been cited. It does not contain information about which exact paper was written by the author, and how many times each individual paper has been cited. Moreover, it only gives the total of co-authors whose university was affiliated to the same
13
city. For this reason an additional dataset was constructed, based on papers written by brokers. With this database, individual papers can be compared. The first database compares authors, the second database compares papers. Because the goal of this study is to investigate the effect of a chosen type of communication medium on brokers’ performance, only papers published by brokers were included in the second database. The first database is called ‘Authors database’, and the second is called ‘Brokers’ papers database’.
Below the method used for the creation of the two databases and the methods used to test the hypotheses are described. First, the construction of the authors database will be explained, followed by a description of the chosen methods used for testing the hypotheses. Next, the creation of the brokers’ papers database and the used methods for testing the hypotheses are explained.
Authors database
The authors’ database consisted of 4780 authors. The database contained information about authors’ constraints, network roles, total number of publications in RMP, the total amount of times their articles have been cited, and the percentage of authors they collaborated with who were affiliated to the same city. This database was used to see if there was a difference in performance between brokers and non-brokers. This database was also used to examine the effect of a chosen type of communication medium on the performance of brokers.
This database was used to create a network based on co-authorship. For example, when three authors published a paper together, each author is considered to have a connection with the other two authors. All relationships between co-authors are assumed to be
symmetrical. The advantage of looking at a co-author network is that it makes it possible to identify authors’ network roles and compare them based on these network roles. However, a disadvantage of looking at a co-author network is that the contribution of a single author is hard to distill from a paper that has been written by multiple authors.
14
In order to calculate an author’s constraint, a node list had to be created. As mentioned above, network constraint measures the extent to which an author’s network contains
redundant contacts (Burt, 1992). A node list shows which author (ego) has a tie with other authors (alters). A program was written in Visual Basic for Applications (VBA) that automatically creates such a node list in Excel. The program went through each author, registering what papers he or she has written. Next, the program registered all the co-authors for each paper. The program linked all the co-authors’ names to the name of the initial author. At this point each author in the database was linked to a set of names with whom he or she wrote an article. The final stage of the program was to report this information in a node list.
The node list was used to calculate constraint scores by using UCINET. The network constraint of an author is high if his or her authors also worked together or when the co-authors published with a central contact. Network constraint decreases when the number of contacts increases. A network constraint value of 0 means that the author in question did not have any co-authors, and was thus not connected to the network. In this study, when an author had a network constraint value between 0.001 and 0.400 his contacts were regarded non-redundant. These authors were called brokers. When network constraint values were higher than 0.400 contacts were regarded redundant. Authors with a constraint value of .401 or higher were called non-brokers. Authors that had a constraint score of .4 or lower were coded as 1. Authors that had a constraint score of .401 or higher were coded as 0.
Each paper written by a broker was retrieved via internet through the website of RMP (Reviews of Modern Physics, 2016). The DOI of the papers was used to retrieve additional information about the city of the university to which authors were affiliated, authors’ affiliation to a top university, and the total number of co-authors. This method was chosen because other options, for example a survey asking authors about their used types of communication medium, was unfavorable because of the large amount of authors in the
15
dataset and the reliance on their response. In addition, the dataset contains information going back to 1929. Some authors may have deceased.
In this study, a variable called ‘same city percentage’ was used as a proxy for the usage of face-to-face or non-face-to-face communication between authors. The same city percentage of brokers (x) was calculated by counting the number of co-authors affiliated to the same city as the affiliation of the broker in question (y), divided by the number of co-authors affiliated to the same city (y) plus the number of co-co-authors not affiliated to the same city (z).
𝑥 = 𝑦
𝑦 + 𝑧
where
x = same city percentage.
y = co-authors that are affiliated to a university that lies in the same city as the affiliated university of the broker in question.
z = co-authors that are affiliated to a university that does not lie in the same city as the affiliated university of the broker in question.
This method was chosen because it does not just count the amount of co-authors affiliated to the same city as the broker, it gives a percentage. This method is therefore less affected by the number of co-authors publishing the paper. For example, a paper published by a lot of authors might give a high number while not all authors are affiliated to the same city, compared to a paper published by only a few authors.
The affiliation to a top university was registered binary, 1 for affiliation to a top university, 0 for affiliation to other universities. The following universities are considered as top universities: Yale, Harvard, Columbia, New York, Stanford, Chicago, and the University of California at Berkeley (Phillips & Zuckerman, 2001). Because of their well-known
16
reputation in the physics field the following universities were added: Massachusetts Institute of Technology (MIT), University of Cambridge, University of Oxford, Princeton University, California Institute of Technology (Caltech), and ETH Zurich.
The number of authors of each paper was counted. Another variable was created by subtracting one from the total number of authors of each paper, resulting in the number of co-authors per paper. For all papers a broker has worked on, the value of this variable was summed. This resulted in the total number of co-authors of the broker.
Because the number of citations depends on the number of papers an author has
published, a Standardized Citation variable was created by dividing the number of citations by the number of publications.
To adjust for skewness and kurtosis logarithmic, square root and one-divided-by transformations were conducted. To be able to transform the variables, values of 0 were increased by 0,0000001 in order to prevent the exclusion of observations with a value of 0, when conducting a logarithmic or a one-divided-by transformation. The transformations with the best results for skewness and kurtosis were used for further analyses.
A correlation matrix was constructed for the following variables: brokerage, percentage of co-authors in the same city as the broker, number of papers published, number of citations, Standardized Citations, average affiliation to a top university, and number of co-authors. To compare the difference between brokers and non-brokers, an independent t-test was used for the following transformed variables: number of papers published, number of citations, and Standardized Citations.
Linear regressions were used to calculate the effect between the number of co-authors that were in the same city as the broker, the independent variable, and number of papers
published, number of citations and Standardized Citations as dependent variables. Also, the affiliation to a top university and the number of co-authors were controlled for.
17
Brokers’ papers database
This database contained information about the papers published by at least one broker consisted of 709 papers. The database contained information about the number of citations, same city percentage of co-authors, years of publication, affiliations to a top university, and the number of authors. This database was used to see if the chosen type of communication medium among the authors of a paper had an effect on the performance of the paper. Also, the database was used to see if the year of publication of a paper had an effect on the chosen type of communication medium.
The same city percentage of a paper was constructed using the following formula: 𝑥 = ∑ 𝑎𝑛
𝑦 𝑛=1
𝑦(𝑦 − 1) where
x = same city percentage y = the number of authors
a = the amount of co-authors affiliated to the same city
This formula divides the total possible amount of authors affiliated to the same city by the actual amount of authors affiliated to the same city. Just like in the authors database, this method was chosen because it does not just count the amount of authors affiliated to the same city, it gives a percentage. This method is therefore less affected by the number of authors publishing the paper. Two examples: the paper ‘Energy Levels of Light Nuclei’ by Hornyak and Lauritsen (1948) had a same city percentage of 100%, because both Hornyak and
Lauritsen were, at the time of publication, affiliated to the California Institute of Technology in Pasadena California. The paper ‘Solar models, neutrino experiments, and helioseismology’ by Bahcall and Ulrich (1988) had a same city percentage of 0%. Bahcall was, at the time of publication, affiliated to the Institute for Advanced Study, Princeton in New Jersey while Ulrich was affiliated to the University of California at Los Angeles.
18
The affiliation to a top university was coded by looking at the affiliation of the first author. If the first author was affiliated to a top university the paper was coded as 1. A paper was coded as 0 when the first author was affiliated to other universities. The following universities were considered as top universities: Yale, Harvard, Columbia, New York, Stanford, Chicago, and the University of California at Berkeley, Massachusetts Institute of Technology (MIT), University of Cambridge, University of Oxford, Princeton University, California Institute of Technology (Caltech), and ETH Zurich.
The number of authors for a paper was calculated by using a VBA program. This program went through the existing database and counted the amount of times a paper’s DOI was connected to another author. This amount is the same as the amount of authors that published the specific paper.
A correlation matrix was constructed with the following variables: the number of citations, same city percentage of co-authors, years of publication, affiliation to a top university, and the number of authors. Then a multiple regression analysis was conducted with citations as the dependent variable and same city percentage as the independent variable, while controlling for the amount of authors and the affiliation to a top university. Finally, another multiple regression analysis was conducted to show the effect of year of publication on same city percentage, while controlling for the amount of authors and the affiliation to a top university.
Results
In this section the outcomes of the different analyses are discussed. First the results for the authors database will be presented. After that, the results of the analyses on the papers of brokers will be presented.
19
Authors
High skewness and kurtosis are measured for number of papers published by authors (Skewness = 5.73, Kurtosis = 51.61), times an author has been cited (Skewness = 5.40, Kurtosis = 38.74), and Standardized Citations (Skewness = 5.78, Kurtosis = 43.21). A more normal distribution is measured for the number of co-authors (Skewness = 1.47, Kurtosis = 2.44), the same city percentage of co-authors (Skewness = 2.35, Kurtosis = 4.80), the average affiliation to top universities (Skewness = 1.04, Kurtosis = -.86), and the average year of publishing (Skewness = -.21, Kurtosis = -1.26). Using a transformation of dividing one by the number of papers published resulted in a skewness of -2.05 and a kurtosis of 2.64. In addition, to increase comprehension of further analyses by undoing the value inversion of the one-divided-by transformation, the logarithmic number of papers published was subtracted from one. Using the logarithmic transformation for citations gave the best results and resulted in a skewness of -2.35 and a kurtosis of 4.24. The logarithmic transformation also improved the Standardized Citations best (skewness = -2.36, kurtosis = 4.29).
Correlations
Table 1. Correlation Matrix Authors RMP
Note. * p < .05, ** p < .01
The results show that constraint is significantly and negatively correlated with the number of papers published (r = -.09, p < .01), but significant and positively correlated to the number of citations (r = .17, p < .01) and Standardized Citations (r = .18, p < .01). Furthermore, network
1. 2. 3. 4. 5. 6. 7. 1. Number of co-authors
2. Average top university .13**
3. Average year of publishing .14** -.36**
4. Constraint -.78** -.09* .14**
5. Co-author same city percentage -.12** .35** -.53** .22**
6. Papers published .14** .25** -.19** -.09** .34**
7. Citations .11** .09* .21** .17** .19** .16**
20
constraint is also correlated to the percentage of co-authors affiliated to the same city (r = .22,
p < .01), the number of co-authors (r = -.78, p < .01), average affiliation to a top university (r
= -.09, p < .05), and the average year of all of an author’s publications (r = .14, p < .01). The same city percentage of co-authors of a broker is significantly and positively correlated to the amount of papers published (r = .34, p < .01), the number of citations (r = .19, p < .01), and Standardized Citations (r = .17, p < .01). The percentage of co-authors affiliated to the same city is also correlated with the number of broker’s co-authors (r = -.12, p < .01), the broker’s average affiliation to a top university (r = .35, p < .01), and the broker’s average year of publication (r = -.53, p < .01). The amount of papers published by a broker is correlated to the total number of co-authors for that broker (r = .14, p < .01). The amount of papers published is also correlated to the broker’s affiliation to a top university (r = .25, p < .01), the average year of publication (r = -.19, p < .01), the number of citations (r = .16, p < .01) and
Standardized Citations (r = .11, p < .01). The total number of times a broker has been cited is correlated to the total number of co-authors (r = .11, p < .01), affiliation to a top university (r = .09, p < .05), and average year of publishing (r = .21, p < .01). The amount of citations and Standardized Citations are highly correlated (r = 1, p < .01) because of the way Standardized Citations have been measured (Standardized Citations = Citations / Publications).
T-test
When conducting a t-test, Levene’s test for equality of variance turned out to be significant for all the variables: number of papers published (p < .01), number of citations (p < .01), and Standardized Citations (p < .01). Because of this, equal variance is not assumed. On average, authors of scientific papers published more papers when they are in a brokerage position (M = .13, SE = .26) than when they are not in a brokerage position (M = .08, SE = .20). This difference is significant t(4780) = -3.91, p < .001 and shows support for hypothesis 1. However, on average brokers had less citations (M = .41, SE = 2.83) than non-brokers (M =
21
.59, SE = 2.55). This difference was non-significant t(4780) = 1.47, p = .11. Finally, the Standardized Citations were also lower for brokers (M = .31, SE = 7.78) than for non-brokers (M = .54, SE = 2.53). This difference is also non-significant t(4780) = 1.86, p = .06. This does not support hypothesis 2.
Regressions
A multiple regression was conducted to see if same city percentage predicted the number of papers published by a broker while controlling for the average year of publication, the amount of co-authors, and the broker’s average affiliation to a top university. This model was
significant (F(4, 555) = 39.33 p < .01), with an R2 of .22. However, the increase in R2 was only .02. The analysis shows that same city percentage of co-authors significantly predicted the number of papers published by a broker (Beta = .18, t(660) = 5.16, p <.01). This shows support for hypothesis 4.
Another multiple regression was conducted to see if same city percentage predicted the times a broker has been cited while controlling for the average year of publication, the amount of co-authors, and the broker’s average affiliation to a top university. This model was significant (F(4, 555) = 14.49 p < .001), with an R2 of .06. However, the increase in R2 was only .02. The analysis shows that same city percentage of co-authors significantly predicts the times a broker has been cited (Beta = .18, t(660) = 3.60. p <.01). A multiple regression was also conducted to see if same city percentage predicts the Standardized Citations while controlling for the average year of publication, the amount of co-authors, and the broker’s average affiliation to a top university. This model is significant (F(4, 555) = 6.77 p < .001), with an R2 of .05. However, the increase in R2 was only .02. The analysis shows that same city
percentage of co-authors significantly predicts the number of papers published by a broker (Beta = .17, t(660) = 3.37, p <.01). This shows support for hypothesis 5.
22
Table 2. Regression matrix Authors1
* p < .05, ** p < .01 *** p < .001. B, unstandardized regression coefficient; SE, unstandardized standard error; β, standardized beta.
Brokers’ papers
The papers database consists of all the papers of the RMP journal between 1929 and 2013 of which at least one of the co-authors is a broker. In total 257 papers met these qualifications. Papers are cited between 0 and 768 times. On average these papers are cited 73.04 times. From the 257 papers 50 are written by one person, 207 are written by two or more people. On average 42% co-authors are affiliated to the same city.
A high kurtosis (Skewness = 3.21, Kurtosis = 12.65) has been measured for the number of citations of papers. Also, the control variable number of authors has a non-normal
1The following ‘authors’ got excluded by SPSS: APS council review committee, APS study group
participants, C. Guillaud, M., M.T. Weiss. These authors were excluded because no same city percentage could be calculated because they only attended discussions, only wrote papers individually, or both.
Variable B SE B β B SE B β B SE B β
Constant 12.02 1.29 51.15 15.30 40.73 15.09
Average top university .05 0.03 .09* .12 .30 .02 .08 .29 .01
Average year of publishing -.01 .00 -.39*** -.03 .01 -.15** -.02 .01 -.12**
Number of co-authors .00 .00 .18*** .03 .01 .13** .02 .01 .12**
R2 .20 .04 .03
F 45.68*** 7.30*** 5.15**
Papers Citations Standardized Citations Model 1
Variable B SE B β B SE B β B SE B β
Constant 9.45 1.42 24.11 16.90 15.66 16.70
Average top university .03 0.03 .05 -.13 .30 -.02 -.15 .30 -.02
Average year of publishing -.01 .00 -.30*** -.01 .01 -.07 -.01 .01 -.05
Number of co-authors .00 .00 .20*** .03 .01 .15** .03 .01 .13**
Same city percentage co-authors .20 .05 .18*** 2.10 .58 .18*** 1.94 .58 .17**
R2 .22 .06 .05
F 39.33*** 8.81*** 6.77***
Model 2
23
distribution (Skewness = 3.60, Kurtosis = 16.97). To normalize the data, different
transformations were conducted. The square root transformation provides the best results in terms of normality for number of citations of papers (skewness = 1.23, kurtosis = 1.71). The logarithmic transformation provides the best results for the number of authors. The number of authors was further transformed by subtracting the one divided by number of authors from 1. This results in a skewness of .74 and a kurtosis of .07.
Table 3. Correlation Matrix Brokers’ Papers
Note. * p < .05, ** p < .01
The number of times a broker’s paper has been cited and the percentage of co-authors in the same city are negatively but non-significantly correlated (r = -.10, p = .14). The percentage of authors affiliated to the same city is significantly correlated to the number of authors (r = -.27,
p < .01) and the paper’s affiliation to a top university (r = .32, p < .01).
1. 2. 3. 4.
1. Number of authors
2. Top university -.08
3. Year of publication .24 -.04
4. Same city percentage -.27** .32** -.06
24
Table 4. Regression matrix Paper’s citations
* p < .05, ** p < .01 *** p < .001. B, unstandardized regression coefficient; SE, unstandardized standard error; β, standardized beta
A multiple regression was conducted to see if same city percentage predicted the times a paper has been cited while controlling for the amount of authors and the affiliation to a top university. The model as a whole is non-significant (p = .29), (F (4, 202) = 2.95, p = .02) with an R2 of .06. The analysis shows that same city percentage of authors does not significantly predict the number of citations of a paper (Beta = -.08, t(206) = -1.06 p = .29). This does not provide support for hypothesis 6.
Variable B SE B β Constant -106.94 38.28 Year of publication .06 .02 .23** Top university .64 .86 .05 Number of authors -.14 .06 -.15* R2 .05 F 3.56* Citations
Model 1
Variable B SE B β Constant -94.09 40.14 Year of publication .05 .02 .20* Top university .85 .88 .07 Number of authors -.15 .05 -.17*Same city percentage authors -.01 .01 -.08
R2 .06
F 2.95*
Model 2
25
Table 5. Regression matrix same city percentage
* p < .05, ** p < .01 *** p < .001. B, unstandardized regression coefficient; SE, unstandardized
standard error; β, standardized beta.
Another multiple regression was conducted to see to what extent year of publication predicts the same city percentage of authors of the scientific papers while controlling for top
universities and the number of authors. The model is significant (F (3, 203) = 20.95, p < .001), with an R2 of .24. The analysis shows that year of publication significantly predicts the percentage of authors affiliated to the same city (Beta = -.30, t(206) = -4.36 p < .001). This shows support for hypothesis 3.
Variable B SE B β Constant 40.48 4.25 Top university 19.38 5.61 .32*** Number of authors -1.64 .44 -.25*** R2 .17 F 20.16***
Same city percentage
Model 1
Variable B SE B β Constant 11154.44 255.80 Top university 19.38 5.72 .22** Number of authors -1.09 .42 -.16* Year of publication -.56 .13 -.30*** R2 .24 F 20.95***Model 2
26
Discussion
In an effort to explain the effect of a chosen communication medium on brokers’
performance, a database of authors of scientific papers published in the journal Reviews of Modern Physics (RMP) was analyzed. This part will discuss the outcomes of the results.
The number of publications is correlated with Standardized Citations, indicating that authors who get published more also get cited more. This could mean that both the number of papers published and citations are measuring performance. However, the correlation shows only a small effect.
On average brokers publish more than non-brokers. However, brokers are not cited more per paper than non-brokers. These contradicting results are only partially in line with current literature that associate brokers with higher performance (Burt, 1997; 2004; Podolny & Page, 1998). This indicates that brokers outperform non-brokers on some aspects, but not all.
For this study physical distance was used as a proxy for the used communication medium. As the results indicate, as papers got published later in time more people from different cities worked together. This is more than likely due to the fact that, as we moved through time, more forms of non-face-to-face communication medium were invented. While less strong in transferring information, these forms have multiple benefits such as being able to communicate over long distances. With the ongoing technological innovations, different non-face-to-face communication media will most likely become better at transferring
information, narrowing the gap with the information richness of face-to-face communication. For example, the developments in virtual reality and holography will enable us to transfer more information such as body language and immediate feedback while still having the benefits of current non-face-to-face communication media such as communicating over long distances.
27
The analyses showed that when brokers used more face-to-face communication, they are cited more. It should however be noted that the model used for citations only explained 6% of the variation and the model used for Standardized Citations only 5%. The results show that brokers that use more face-to-face communication also publish more papers. However, the chosen type of medium shows no significant effect on the performance of individual papers. These findings indicate that different types of communication medium lead to different levels of brokers’ performance, but not necessarily that of a paper. It could be that this difference is caused due to a need of trust among the authors when collaborating. Trust is “when you commit to a relationship before you know how the other person will behave” (Burt, 2005, p. 101). The more unspecified the terms of working together, the more trust is needed for collaboration to happen (Burt, 2005). When tasks are ambiguous, knowing people that are trustworthy becomes more important. Being a broker could lead to distrust by the two groups the broker connects, because both do not see the broker as part of their group (Xiao & Tsui, 2007). This means that trust becomes a more important factor for brokers than for non-brokers, which might be better achieved through the use of face-to-face communication. It may be that brokers that used face-to-face communication, because of its ability to transfer a lot of information, reduced uncertainty and therefore gained more trust from potential co-authors. Because of this trust instilled by face-to-face communication, potential co-authors could be more willing to work with brokers, increasing the likelihood of brokers publishing papers.
Non-face-to-face communication media have a lot of benefits for brokers. However, face-to-face interaction leads to higher levels of brokers’ performance. While using non-face-to-face types of communication medium increases the amount of potential people a broker can collaborate with, it is not associated with higher levels of performance. Collaborating with
28
someone using non-face-to-face communication might be more challenging because of the lack of trust.
Limitations
During this study, the city of the university to which an author is affiliated was used to determine the used communication medium. When an author was affiliated to a university that was located in the same city as a university of another author, it was assumed that they mainly used face-to-face communication. When the universities of two authors were not located in the same city, it was assumed that they mainly used non-face-to-face interaction for their communication. However, authors affiliated to different cities could have visited each other in person and authors from the same city could have used only non-face-to-face communication media such as e-mail. Also, distance between cities was not used in the equation. When two authors were not affiliated to the same city, but the physical distance between the two cities was rather small, it could also be that they mainly used face-to-face communication. Also, when two authors were affiliated to the same city, but this particular city was very large, it could be that they mainly used non-face-to-face communication. In addition, relative distance was not used in the equation. When someone affiliated to a well-connected city (e.g. good roads and a lot of public transport) relative distance decreases, face-to-face communication becomes less obstructed and thus more likely. And, also, when someone is affiliated to a badly connected city, the relative distance increases, causing face-to-face communication between two authors to be more obstructed.
Another limitation of this study that, for the calculation of same city percentage, only the city of the university an author is affiliated to, was used. However, it might be that an author has a different hometown than the city of his or her affiliated university. It could be the same city as a author’s home town or his or her university’s city. This could mean that
co-29
authors used face-to-face communication while the universities they are affiliated to are not located in the same city.
For this research, only the Review of Modern Physics journal was used. More journals were not used because of time constraints and unavailability of additional information
provided, such as to which university an author was affiliated.
One of the control variables used registered if an author or paper was affiliated to a top university. First, the classification of a top university is subjective to some extent. Also, the classification of top universities was based on the reputation in 2016, while the data ranged from 1929 to 2013. During this time, universities might have gained or lost.
It might be that a broker that knows exactly when to use face-to-face communication and when to use non-face-to-face communication, combined the two for maximum efficiency. Therefore, it might be that the broker is coded as same city, while in reality he or she mainly used non-face-to-face communication.
In this paper, citations have been used as an indicator of performance. However, there are some problems with using citations to rate performance (Garfield, 1979). Even though it is rare, sometimes papers are cited in a negative way. This means that a paper is cited because the citing paper disagrees with the cited paper, which is not an indicator of performance.
Finally, this research only made the distinction between face-to-face communication and non-face-to-face communication. However, there are a lot of differences between the different non-face-to-face communication media. For example there are a lot of differences between a telephone call and a letter. The effect of different non-face-to-face communication media are not incorporated in this research.
30 Future research
With the ongoing technological innovation new communication media will be invented. These new communication media might give new benefits that increase the use of non-to-face communication. Also, inventions such as virtual reality enable a lot of aspects of face-to-face communication, such as body language, to be transferred across the world in seconds. These new communication media and their benefits can have an effect on the performance of brokers. Future research could investigate these new communication media and their effects on performance.
The results of this research show that face-to-face communication leads to higher levels of performance. It could be that this type of communication is better at activating the network because of trust. Future research could investigate the effects of a chosen
communication medium on trust and the activation of ties.
Conclusion
This study explains the effects of different types of communication medium on brokers’ performance with the following research question:
What is the effect of different types of communication medium on a broker’s performance?
In answering this question more will become known about what distinguishes high from low performing brokers. It will also help brokers to choose a type of communication medium that will best facilitate the pursuit of their goals.
Contrary to current literature, this research showed that brokers were not consequently outperforming non-brokers. Brokers outperformed non-brokers on some aspects, but not all. This indicates that previous literature might have overestimated the performance of brokers.
31
When looking at the chosen type of communication medium, it is clear that as we move further in time more forms of non-face-to-face communication medium were invented and used. However, when looking at the effects of different types of communication medium on a broker’s performance, face-to-face communication remains important. The usage of more face-to-face communication by brokers has a positive effect on their performance. A logical explanation for this is that brokers who used face-to-face communication gained more trust from potential partners.
In the future more communication media will probably be invented that will shrink the gap of information richness between face-to-face communication and non-face-to-face
communication. These new communication media might have a positive impact on the performance of brokers. If this is true, then face-to-face communication will gradually
become less important. However, meeting face to face will probably never be fully substituted by non-face-to-face communication media.
32
Appendix 1 – Network visualization
33
References
Bahcall, J. & Ulrich, R., 1988. Solar models, neutrino experiments, and helioseismology.
Reviews of Modern Physics, 60(2), p. 297.
Bedeian, A., Van Fleet, D. & Hyman, H., 2009. Scientific achievement and editorial board membership. Organizational Research Methods, 12(2), pp. 211-238.
Bourdieu, P. & Wacquant, L. J., 1992. An Invitation to Reflexive Sociology. Chicago: University of Chicago Press.
Brass, D. J., 1985. Men's and Women's Networks: A Study of Interaction Patterns and Influence in an Organization. The Academy of Management Journal, 28(2), pp. 327-343. Burt, R., 2005. Brokerage and closure: An introduction to social capital.. Oxford: Oxford University Press.
Burt, R. S., 1992. Structural holes. Cambridge, MA: Hardvard University Press.
Burt, R. S., 1997. The Contingent Value of Social Capital. Administrative Science Quarterly, 42(2), pp. 339-365.
Burt, R. S., 2001. Structural Holes versus Network Closure as Social Capital. Social capital:
Theory and research, pp. 31-55.
Burt, R. S., 2004. Structural Holes and Good Ideas. American Journal of Sociology, 110(2), pp. 349-399.
Burt, R. S., 2005. Brokerage and Closure: An Introduction to Social Capital. Oxford: Oxford University Press.
34
Burt, R. S., 2009. Structural holes: the social structure of competition. s.l.:Harvard university press.
Burt, R. S., 2010. Neighbor Networks: Competitive Advantage Local and Personal. Oxford: Oxford Unversity Press.
Burt, R. S., Kilduff, M. & Tasselli, S., 2013. Social Network Analysis: Foundations and Frontiers on Advantage. Annual Review of Psychology, Volume 64, pp. 527-547. Burt, R. S., Kilduff, M. & Tasselli, S., 2013. Social Network Analysis: Foundations and Frontiers on Advantage. Annual Review of Psychology, Volume 64, pp. 527-547. Burt, R. S. & Talmud, I., 1993. Market niche. Social networks, 15(2), pp. 133-149.
Crowley, D. D. J. & Mitchell, D., 1994. Communication Theory Today. Stanford: University Press.
Daft, R. & Lengel, R., 1986. Organizational information requirements, media richness and structural design. Management science, pp. 554-571.
Daft, R. L. & Lengel, R. H., 1986. A proposed integration among organizational information requirements, media richness, and structural design. Management science, Volume 32, pp. 554-571.
Daft, R. L., Lengel, R. H. & Trevino, L. K., 1987. Message Equivocality, Media Selection, and Manager Performance: Implications for Information Systems. MIS Quarterly, 11(3), pp. 355-366.
Dennis, A. & Valacich, J., 1999. Rethinking Media Richness: Towards a Theory of Media Synchronicity. Systems Sciences.
35
Dimmick, J., Kline, S. & Stafford, L., 2000. The Gratification Niches of Personal E-mail and the Telephone. Communication Research, 27(2), pp. 227-248.
Garfield, E., 1979. Is citation analysis a legitimate evaluation tool?. Scientometrics, 1(4), pp. 359-375.
Goldhaber, A. S. & Nieto, M. M., 2010. Photon and graviton mass limits. Reviews of Modern
Physics, 82(1), p. 939.
Granovetter, M. S., 1973. The Strength of Weak Ties. American Journal of Sociology, 78(6), pp. 1360-1380.
Hornyak, W. F. & Lauritsen, T., 1948. Energy Levels of Light Nuclei. Review of Modern
Physics, 20(1), p. 191.
Inkpen, A. & Tsang, E., 2004. Social capital, networks, and knowledge transfer. Academy of
Management Review, 30(1), pp. 146-165.
Kiesler, S. & Cummings, J. N., 2002. What do we know about proximity and distance in work groups? A legacy of research. Distributed work, Volume 1, pp. 57-80.
Kraut, R. et al., 1998. Internet Paradox: A Social Technology That Reduces Social Involvment and Psychological Well-Being?. American psychologist , 53(9), p. 1017.
Lee, J., 2010. Heterogeneity, Brokerage, and Innovative Performance: Endogenous Formation of Collaborative Inventor Networks. OrganizationScience, 21(4), pp. 804-822.
Lengel, R. H. & Daft, R. L., 1989. Selection of Communication as an Executive Skill. The
Academy of Management Executive, 2(3), pp. 225-232.
Levin, D. Z. & Cross, R., 2004. The Strength of Weak Ties You Can Trust: The Mediating Role of Trust in Effective Knowledge Transfer. Management Science, 50(11), pp. 1477-1490.
36
Marin, A. & Wellman, B., 2011. Social Network Analysis: An Introduction. In: The SAGE
Handbook of Social Network Analysis. London: SAGE Publications Ltd, pp. 11-25.
Maula, M. V., Autio, E. & Murray, G. C., 2009. Corporate venture capital and the balance of risks and rewards for portfolio companies. Journal of Business Venturing.
Mehra, A., Kilduff, M. & Brass, D. J., 2001. The social networks of high and low self-monitors: Implications for workplace performance. Administrative Science Quarterly, Volume 46, pp. 121-146.
Nelson, R. E., 1989. The Strength of Strong Ties: Social Networks and Intergroup Conflict in Organizations. The Academy of Management Journal, 32(2), pp. 377-401.
Obstfeld, D., 2005. Social Networks, the Tertius Iungens Orientation, and Involvement in Innovation. Administrative Science Quarterly, 50(1), pp. 100-130.
Podolny, J. M. & Baron, J. N., 1997. Resources and Relationships: Social Networks and Mobility in the Workplace. American Sociological Review, 62(5), pp. 673-693.
Podolny, J. M. & Page, K. L., 1998. Network Forms of Organization. Annual Review of
Sociology, Volume 24, pp. 57-76.
Pollock, T. G., Porac, J. F. & Wade, J. B., 2004. Constructing deal networks: brokers as network "architects" in the u.s. IPO market and other examples. Academy of Management
Review, 29(1), pp. 50-72.
Reviews of Modern Physics, 2016. Reviews of Modern Physics. [Online] Available at: http://journals.aps.org/rmp/
Sasovova, Z., Mehra, A., Borgatti, S. P. & Schippers, M. C., 2010. Network churn: The effects of self-monitoring personality on brokerage dynamics. Administrative Science
37
Sørensen, A. B., 1996. The Structural Basis of Social Inequality. American Journal of
Sociology, 101(5), pp. 1333-1365.
Treviñio, L. K., Webster, J. & Stein, E. W., 2000. Making Connections: Complementary Influences on Communication Media Choices, Attitudes, and Use. Organizational Science, 11(2), pp. 163-182.
Wasserman, S. & Faust, K., 1994. Social network analysis: Methods and applications. Cambridge: University press.
White, H. D., Wellman, B. & Nazer, N., 2004. Does Citation Reflect Social Structure? Longitudinal Evidence From the “Globenet” InterdisciplinaryResearch Group. Journal of the
American Society for Information Science and Technology, 55(2), pp. 111-126.
Xiao, Z. & Tsui, A. S., 2007. When brokers may not work: The cultural contingency of social capital in Chinese high-tech firms.. Administrative Science Quarterly, 52(1), pp. 1-31.
