View
11
Download
0
Category
Preview:
Citation preview
Working paper. Please do no not redistribute or cite without author’s permission.
1
Entrepreneurship, Legitimacy and Online Social Communities:
An Empirical Analysis
Sofia Bapna and Mary Benner
Carlson School of Management, University of MN
We seek to address the question of how entrepreneurial firms can leverage online
social networking communities to establish legitimacy. We propose that one
measure of a firm’s legitimacy is the size of its online community, and we
hypothesize about the specific ways that the use of online symbolic actions can
facilitate growth in a firm’s online social community, and therefore, its legitimacy.
We test the impact of four categories of online symbolic actions on online social
community growth. We find that online symbolic actions that i) show product and
industry knowledge, a component of actions reflecting credibility, are positively
associated with online social community growth, ii) show firm accomplishments, a
component of actions reflecting organizational achievements, are also positively
associated with online social community growth, and iii) seek opinions have a
negative impact on online social community growth.
The failure rate for new businesses is staggering. A full quarter of new ventures don’t make it
past the first year, 44% of ventures fail by the end of the third year, and 55% of ventures fail by
the end of the fifth year from founding (Shane 2008). Institutional theorists associate legitimacy,
that is, the acceptance of an organization’s actions as proper or appropriate (Suchman 1995),
with new venture success and survival (Meyer and Rowan, 1977, DiMaggio and Powell 1983).
Stinchcombe (1965) suggests that legitimacy helps new ventures, particularly new forms of
Working paper. Please do no not redistribute or cite without author’s permission.
2
organization, overcome the liability of newness. Aldrich and Fiol (1990) refer to new venture
legitimacy as “especially critical”. Legitimacy is viewed as a resource that can be used to acquire
other resources, to gain unrestricted access to markets (Brown, 1994), to thwart questions about
an organization’s competence, and to raise the status of an organization in its community (Oliver
1991). Meyer and Rowan (1977) note that organizations that lack legitimacy “are more
vulnerable to claims that they are negligent, irrational or unnecessary."
Managers can purposely seek legitimacy through specific actions (Suchman 1995) and previous
literature has provided insights into some of these actions. Suchman (1995) suggests the use of
symbolic actions and public discussion as approaches to enhance legitimacy. Zimmerman and
Zietz (2002) theorize that new ventures can gain legitimacy by adherence to socio-political
regulations, norms and widely held beliefs within a venture’s domain. Lounsbury and Glynn
(2001) propose that new ventures can increase their legitimacy through the use of stories that
reflect the firm’s stock of entrepreneurial capital, which includes patents, credentials,
endorsements, prior performance, and social connections.
Apart from legitimatization in the context of new ventures, past literature has also examined
legitimacy creation in various unique settings including legitimacy in the context of
multinational enterprises (Kostova and Zaheer 1999), legitimacy of networks of firms (Human
and Provan 2000), legitimacy in contexts such as hospitals where technical and managerial
legitimacy are created in ways distinct from each other (Reuf and Scott 1998), and legitimacy in
the context of deep institutional change within a firm (Suddaby and Greenwood 2005).
Working paper. Please do no not redistribute or cite without author’s permission.
3
However, none of these studies have examined the use of online social communities to establish
legitimacy. Several unique features of social networking sites make it conducive to the
legitimation process. Before we delve into these reasons it is important to define social
networking sites and their relevance to businesses. Ellison and Boyd (2013) define a social
networking site as:
“A networked communication platform in which participants i) have uniquely
identifiable profiles that consist of user-supplied content, content provided by
other users, and/or system-provided data; ii) can publicly articulate
connections that can be viewed and traversed by others; and iii) can consume,
produce, and/or interact with streams of user generated content provided by
their connections on the site.”
Henceforth, we refer to a firm’s connections on a social networking site as the firm’s online
social community, community, fans, or fan base. The widespread use of such sites by businesses
is evident in the fact that 11 million businesses used the social networking site, Facebook, in
2012. A survey of business professionals indicates that 55% of them devote employee resources
to manage their social media initiatives (Mzinga Inc. and Babson Executive Education 2009).
Further, the features of social networking sites lead us to four arguments about why such sites
are potentially effective channels for establishing new venture legitimacy.
Firstly, social networking sites allow a firm to communicate with its community members
several times a day via unique content that can be spontaneous, that is, the communications don’t
Working paper. Please do no not redistribute or cite without author’s permission.
4
necessarily require pre-planning as is the case with traditional advertising and marketing. These
characteristics allow for a large variety of content, and give the firm the flexibility to tailor
content to current issues. This makes social networking sites an optimal channel for the use of
symbolic actions (Suchman 1995), which make people draw interpretations about the character
of the organization (Brown 1994), and for story telling (Lounsbury and Glynn 2001). For
example, a firm sharing a photo on a social networking site of products being quality checked is
a symbolic action that can cause viewers to perceive the firm as having professional processes.
Secondly, community members are able to respond to content published by a firm on the social
networking site, and firms are able to respond back either with a generic response that is geared
to all community members, or with a response that is geared to a specific member. The cycle of
communication is quick with the possibility of an exchange of several messages over a single
day. Further, social networking sites allows for communication between community members
who don’t necessarily know each other, and the communication between community members is
visible to the firm. This three-way communication, between firm and community member,
between community member and firm, and between community members, is akin to a
conversation. This is consonant with the idea of participating in public discussion to create
legitimacy (Suchman, 1995).
Thirdly, social networking sites allow for social diffusion via peer influence (Aral and Walker
2010). People who are connected to members of the firm’s community, but not necessarily to the
firm can see interactions between the firm’s community member and the firm and are thus
exposed to new information and stories about the firm. A community member can choose to
Working paper. Please do no not redistribute or cite without author’s permission.
5
indicate her interest in a story shared by the firm on the social networking site. The community
member can also comment on the story or share it. Everyone that the community member is
connected to can see such actions. Thus social networking sites allow firms to leverage their
community members to diffuse information and stories. This is aligned with the idea of
broadcasting stories about the firm’s entrepreneurial capital (Lounsbury and Glynn 2001) and
also about the firm’s adherence to commonly held beliefs and norms in the domain that the firm
operates (Zimmerman and Zietz 2002).
Lastly, social networking sites publicly articulate information about how many people a firm is
connected to on the site. A firm’s connections on the site, which is also referred to as fan count,
is the most popular metric for measuring social media success (Chief Marketer’s 2011 Social
Marketing Survey, via Comscore 2012). These connections or fans are people who have
affirmatively stated their affinity for a particular organization, for reasons that may include “self-
expression, communicating positive associations of that brand to others, staying in
communication with that brand, or receiving deals and promotions” (Lipsman, Mudd, Rich and
Bruich 2011). Facebook fans as well as their friends on Facebook who have been exposed to
content by a retailer on the social networking site are found to be significantly more likely to
make a purchase from that retailer than an average Internet user (Lipsman, Mudd, Aquino and
Kemp 2012). With the intent of leveraging their fans, some iconic brands have amassed more
than twenty million fans. Based on arguments of mimetic isomorphism which suggest that under
conditions of uncertainty, firms seek to gain legitimacy by following the practices of other firms
that they perceive to be successful (DiMaggio and Powell 1983), we expect that new firms will
attempt to establish a presence on social networking sites and grow their fan base if this is a
Working paper. Please do no not redistribute or cite without author’s permission.
6
popular practice in their industry. This is because new firms, by virtue of their nascency are
faced with uncertainty.
These arguments point to an important gap in research in our understanding of legitimacy. Thus,
in this empirical study, we seek to address the question of how entrepreneurial firms can leverage
online social networking communities to establish legitimacy. We propose that one measure of a
firm’s legitimacy is the size of its online community, and we hypothesize about the specific ways
that the use of online symbolic actions can facilitate growth in a firm’s online social community,
and therefore, its legitimacy.
Online symbolic actions are actions that, like traditionally defined symbolic actions, carry
subjective meaning that is interpreted by the observer (Brown, 1994). However, online symbolic
actions are limited to those that can be carried out on a social networking site via posts. A post is
publicly visible content that is published by the firm on a social networking site. Posts can
include text, photos and icons. Posts can also include links to articles, blog posts and videos. The
following examples differentiate a symbolic action from an online symbolic action. An
entrepreneur can take a venture capitalist to a supplier meeting to display the venture capitalist’s
endorsement, or indicate reliability through a fashionably decorated office (Zott and Huy, 2007).
These are symbolic actions. However, on a social networking site the entrepreneur can share a
publicly available story about being funded by a venture capitalist or share photos of their office.
Because online symbolic actions are limited to posts on a social networking site, such actions
often take the form of representation of actions through media such as text and photos. However,
social networking sites also permit the firm to perform some unique actions like conducting a
Working paper. Please do no not redistribute or cite without author’s permission.
7
public discussion in which the firm seeks an opinion. Fans can view each other’s opinions,
interact with one and other, and the firm can participate in the discussion. All posts have the
potential of being online symbolic actions whether they are intended as such or not. A manager
may publish some content on a social networking site with the intention of it being viewed as an
online symbolic action, or just go about sharing something without intending it as an online
symbolic action. Either of these actions is an online symbolic action if the audience attaches a
symbolic meaning to the action.
To compile a comprehensive list of possible online symbolic actions and measure their impact on
fan growth, we started with symbolic actions suggested in past literature that are believed to be
associated with legitimacy, such as those noted by Zott and Huy (2006). From this list we
selected those symbolic actions that could potentially translate to the social media domain. We
augmented the list with additional symbolic actions that can be carried out over a social
networking site. Based on past literature and our theory we identified thirteen online symbolic
actions that can potentially lead to legitimacy. We grouped each of these online symbolic actions
into one of four broad categories of online symbolic actions. They are those that: reflect
credibility, reflect professional organizing, reflect organizational achievements, and seek
opinions. We propose that the frequent use of these four categories online symbolic actions by a
new venture leads to the growth of its online social community. We empirically test our
hypotheses by manually analyzing the content of close to 10,000 posts by new ventures.
Our empirical analysis points to three main findings. The first finding is that online symbolic
actions that show product and industry knowledge, which is a component of online symbolic
Working paper. Please do no not redistribute or cite without author’s permission.
8
actions reflecting credibility, are associated with online social community growth. Examples of
posts that fall under this category are posts that show how to use a product, posts that contain
broad educational content or posts that relay industry related news. The second finding is that
online symbolic actions that show firm accomplishments, which is a component of online
symbolic actions reflecting organizational achievements, are also associated with online social
community growth. Examples of posts that fall under this category are posts that show awards
won by the firm, as well as milestones achieved by the firm. These two findings indicate a form
of preferential attachment at work in the dynamics of social networking sites. People want to
connect and interact online with firms that they perceive to be knowledgeable and noteworthy in
terms of achievements. The final empirical finding is that contrary to our prediction online
symbolic actions that seek opinions have a negative impact on online social community growth.
We believe this may be due to an over use of questions posed by some firms to their community
members.
While previous studies have examined the impact of symbolic actions on legitimacy, this study
extends our understanding of legitimacy and symbolic actions to the domain of social media and
has three distinguishing factors. First, the study defines online symbolic actions and describes
how they are performed. The study also identifies which symbolic actions from previous
literature translate effectively to the social domain and also suggests some additional online
symbolic actions that are relevant to social media sites. Next, the study empirically examines the
impact of various types of online symbolic actions on online social community growth. Last, the
study suggests a theoretical model whereby legitimacy is created via social networking sites.
Working paper. Please do no not redistribute or cite without author’s permission.
9
This study has implications for managers of new ventures because it suggests a path to
legitimacy and growing online social communities. While managers have access to up to the
minute data analytics information such as engagement level and fan count, there are few large-
scale empirical studies that suggest a mechanism and path for growing online social
communities.
The remainder of this paper is organized into five sections. The first section describes the
boundaries of the study and the choice of social networking site for this study along with a
description of the site’s features and functionality. In the second section we review past literature
on symbolic actions in the context of legitimacy, develop four hypotheses that relate online
symbolic action to online social community growth and describe our theoretical model for
legitimacy creation. In the third section, we elaborate on our research context, methodology
employed, and the process we used for data collection and validation. In the fourth section this,
we present our empirical findings based on the analysis of about 10,000 Facebook posts for 15
entrepreneurial firms from a single industry, with a median of 67 weeks of data per firm. In the
final section, we discuss our contribution to research, practice and to the social networking sites
phenomenon in the context of e-commerce businesses.
Boundaries of the Study and Institutional Context
In the following two paragraphs we describe the boundaries of this study, and the choice of
social networking site for this study along with a description of the site’s features and
functionality.
Working paper. Please do no not redistribute or cite without author’s permission.
10
We limit our attention to entrepreneurial firms that have a significant online component, and are
business-to consumer (B2C) firms that sell physical products. An entrepreneurial context is
important for this study because as a firm becomes well established it is difficult to determine
whether the firm’s online social community is growing because of brand effects, existing status,
existing legitimacy or due to other factors. This is addressed, at least in part, by the fact that our
sample consists of new ventures, that is, those within the first three years of founding, that don’t
have the history and the legitimacy that is associated with well-established firms. Further, online
businesses typically engage in online social networking activities since such business target
customers who are also likely to be users of social networking sites. Thus, online businesses are
a good context for this study. Additionally, this study is limited to B2C and product firms
because the dynamics of usage of social networking sites by business-to-business and service
firms are considerably different from that of B2C product focused firms. This study focuses on
firms that sell physical products rather than digital products (such as a digital newspaper or
software) because physical and digital products lend themselves to different dynamics on social
networking sites in terms of community member contribution to discussions and content. Finally,
this study analyzes posts by the firm rather than user generated content, because we seek to
understand what online symbolic actions by the firm lead to legitimacy.
We focus our study on the social networking site, Facebook, because of the site’s extensive
adoption. This is illustrated by the fact that in the United States Facebook accounts for 90
percent of the time spent on social networking sites (comScore 2011). Furthermore, 11 million
businesses used the site in 2012. Firms have their own uniquely identifiable profile page on
Working paper. Please do no not redistribute or cite without author’s permission.
11
Facebook, where they periodically share content via posts. Visible to each Facebook user is a
live updated feed of the activity on the site by users she is connected to. Users can also see which
of their connections are fans of a particular firm. To become a fan or a community member of a
firm, a user needs to ‘like’ the firm’s Facebook page by clicking the ‘like’ button after visiting
the firm’s Facebook page. When a user becomes a fan of a page, posts by the firm are displayed
on that user’s live feed. The number of Facebook fans that a firm has at any given point in time
is publicly visible. Site users can interact with a firm by commenting on a post by a firm, sharing
it, or by indicating their interest in a post by ‘liking’ it. A user can like a post by clicking the
‘like’ button associated with the post. When a user likes, shares or comments on a post, the
action is publicly visible on the firm’s profile page and is also displayed on the feed page of the
user’s connections on the social networking site. If the firm permits them to do so, users of the
social networking site can add content on the firm’s profile page. This is called user generated
content. Firms can consume and interact with user-generated content on the social networking
site.
Symbolic Actions in the Context of Social networking sites
All objects and actions can potentially convey both an instrumental dimension and a symbolic
dimension. The instrumental dimension is based on objective criteria and relates to the attributes
of the products or service (Lievens and Highhouse 2003). It is measured by economic or
performance criteria such as speed and profitability (Rafaeli and Vilnai-Yavetz 2004). In
contrast, the symbolic dimension is subjective (Lievens and Highhouse 2003) and evaluated
according to criteria like values, feelings, and predilections of the observer (Rafaeli and Vilnai-
Working paper. Please do no not redistribute or cite without author’s permission.
12
Yavetz 2004). Symbolic actions are those actions to which subjective meaning is associated
beyond the substantive impact of the action (Brown 1994), that is, beyond the instrumental
dimension.
Several researchers have suggested specific symbolic actions as a means to establish legitimacy,
which in turn allows entrepreneurs to gain resources. Aldrich and Fiol (1994) propose that
symbolic communication can enable the creation of cognitive legitimacy. Santos and Eisenhardt
(2009) propose that firms that use identity claiming mechanisms such as templates, stories and
leadership signals become cognitive referents, that is the organization is accepted as
characterizing the new market. Lounsbury and Glynn (2001) posit that entrepreneurial stories
help legitimate a new venture by crafting its identity. Higgins and Gulati (2003) theorize that in
the biotechnology industry, upper echelon experience serves as symbols of a new venture’s
legitimacy to critical outsiders such as prominent underwriters.
Drawing a parallel form these studies, we suggest that a firm can use online symbolic actions to
establish legitimacy. We propose that there are four categories of online symbolic actions that
can lead to legitimacy. These are ctions that reflect credibility, reflect professional organizing,
reflect organizational achievements, and seek opinions. We explain each of these actions in the
following sub-sections.
Symbolic Actions Conveying Firm Credibility
Credibility refers to “the quality of being believable or worthy of trust” (Collins English
Dictionary, 2013). Past research has shown that entrepreneurs have depicted themselves as being
Working paper. Please do no not redistribute or cite without author’s permission.
13
credible by displaying personal capability and commitment to their ventures (Zott and Huy
2007), using association with prominent others (Higgins and Gulati 2003) and via obtaining
endorsements (Starr and MacMillan 1990).
We define online symbolic actions that convey firm credibility as those that fall into one of three
categories
i) Actions conveying the capability of the founder or key employees: Specifically these are
posts that portray the entrepreneur(s) or key team members as capable company builders.
Such capabilities can be displayed through posts that talk about the founder(s) or key
employees winning awards, being speakers at an event or conference, being interviewed
by the media, or being mentioned in the media. From an instrumental perspective, such
actions show past accomplishments, while from a symbolic perspective such actions are
an indicator of future competence and capability of team members.
i) Actions conveying association with industry experts: These are posts showing content
such as an interview with an external expert; product picks, tips or suggestions by an
external expert; or a live Q&A session with an external expert. The instrumental
dimension of such actions is that they communicate the expert’s opinions. Symbolically
these actions convey that established people recognize and want to associate with the
firm.
ii) Actions conveying product or industry knowledge: These are posts that provide
educational or value added informative content related to the firm’s products or the
industry to which the firm belongs. Instrumentally, educational or informative content
disseminates knowledge and views while symbolically such posts portray the firm as
being knowledge and having expertise (or having access to knowledge and expertise)
Working paper. Please do no not redistribute or cite without author’s permission.
14
about the products the firm sells and the industry in which it operates.
These ideas suggest the following hypothesis:
Hypothesis 1: The more frequently that an entrepreneurial venture performs online symbolic
actions that reflect credibility the more its online social community grows.
Symbolic Actions Conveying Professional Organizing
Entrepreneurs convey the quality of their venture’s organizing efforts by showing that the firm
has adopted professional structures and processes, for instance firms us fashionable front offices,
dress codes and professional looking websites in order to convey professional organizing (Zott
and Huy 2007). We define online symbolic actions that convey professional organizing as those
that fall into one of the following two categories. The first category is posts that communicate
the use of professional processes or procedures. For example, a post that includes a photograph
of products being quality checked before they are shipped to a customer indicates that the firm
employs professional processes. The second category constitutes posts that convey the existence
of professional structures. An example of a post that conveys professional structures is an
interview with an employee in which the employee describes her role in the firm. Instrumentally,
both these types of posts provide behind the scenes information about the firm. Symbolically,
such posts depict the professional nature of the firm’s structures and procedures, thus portraying
the firm as being both professionally run and experienced. Thus we hypothesize:
Hypothesis 2: The more frequently that an entrepreneurial venture performs online symbolic
Working paper. Please do no not redistribute or cite without author’s permission.
15
actions that reflect professional organizing the more its online social community grows.
Symbolic Actions Conveying Organizational Achievement
In previous studies, entrepreneurs have been shown to demonstrate organizational achievement
such as prototypes, awards and demonstrations, venture age and number of employees to
establish legitimacy (Zott and Huy 2007). Santos and Eisenhardt (2009) propose that firms that
use stories and leadership signals such as setting standards become cognitive referents in a new
market. Rao (1994) in his study of the auto industry discusses how winning certification contests
legitimate organizations and create favorable reputations.
We define online symbolic actions that convey organizational achievement as posts that either i)
convey milestones, partnerships and awards won by the firm ii) convey that the firm sells award
winning products or those that are featured in the media. An example of a post that describes a
milestone is a photograph of the firm’s employees celebrating the firm’s first anniversary.
Instrumentally, such posts document past accomplishments or past certification of the products
sold by the firm. Symbolically, posts that convey information about milestones indicate that the
firm has persisted over time or grown. Posts that convey information about partnerships or the
fact that the firm sells award-winning products symbolically indicate that the firm is recognized
by established entities its environment. We therefore hypothesize:
Hypothesis 3: The more frequently that an entrepreneurial venture performs online symbolic
actions that reflect organizational achievements the more its online social community grows.
Working paper. Please do no not redistribute or cite without author’s permission.
16
Symbolic Actions Seeking Thoughts or Opinions
We define online symbolic actions that seek others’ thoughts or opinions as posts that show that
the firm wants to engage in a conversation with their audience by asking questions. An example
of such a post is sharing pictures of two different prints for a product and asking which one the
reader prefers. On a social networking site, such questions are not rhetorical because readers are
able to respond to the questions and are also able to view and react to the responses by others.
The instrumental dimension of such posts is that the firm is seeking an opinion or input, while
the symbolic dimension is that the firm wants to connect with their audience and that the firm
cares about what its audience thinks. Thus we hypothesize:
Hypothesis 4: The more frequently that an entrepreneurial venture performs online symbolic
actions that seek opinions the more its online social community grows.
We summarize these four hypotheses in Figure 2.
Theoretical model
Our proposed theoretical model (see figure 1) that describes how new ventures use online
symbolic actions to create legitimacy is partly based on arguments of mimetic isomorphism,
which suggest that under conditions of uncertainty firms mimic the actions of other firms that
they perceive to be successful, in order to be seen as legitimate (DiMaggio and Powell 1983).
New ventures by virtue of their nascency are faced with uncertainty. Therefore, we expect that in
industries where established firms have large online social communities, new ventures will seek
to establish a presence on social networking sites and grow their fan base. When a new venture
Working paper. Please do no not redistribute or cite without author’s permission.
17
uses an online symbolic action, the post acts as a stimulus around which existing community
members engage because they perceive the firm as being knowledgeable, and noteworthy in
terms of achievements. Community members engage by posting comments to, sharing, or
registering an interest in the post (Ellison and Boyd 2013). People who are connected to
members of the firm’s community, but not necessarily to the firm can see interactions between
the firm’s community member and the firm, and are thus exposed to new information and stories
about the firm. This serves as a trigger for activities on the site (Ellison and Boyd 2013). Thus,
the process of social diffusion of information, that is, getting information about the firm through
their connections may cause users of the site to join the focal firm’s online social community.
The firm continues to infuse the online social community with online symbolic actions to keep
its existing community engaged and simultaneously grow the firm’s online social community. In
this model, a high frequency of online symbolic actions is important in order to remain visible
through the torrent of information that is received by typical users of social networking sites.
Legitimacy that is established through a large online social community can be either
consequential or exchange legitimacy depending on the lens of the stakeholder. Exchange
legitimacy is awarded based on “expected value” or “practical consequences” to a particular set
of constituents (Suchman 1995). Thus, it is conceivable that when a prospective financial backer
is evaluating the legitimacy of a venture, she considers the potential economic benefits of a large
online social community such as leveraging the community for future revenues and for diffusing
information about the firm. Consequential legitimacy, in contrast, is based on what an
organization has accomplished (Meyer and Rowan 1991). Scott and Meyer (1991) note that
consequential legitimacy may be awarded based on criteria such as hospital mortality rates and
Working paper. Please do no not redistribute or cite without author’s permission.
18
automobile emission standards. In a similar vein, it is likely that a stakeholder such as a potential
customer may make inferences about the quality or value of a firm’s products based on the size
of the online social community that has voluntarily associated itself with the firm.
Alternate explanations
An alternate explanation to using symbolic action for online social community growth is the use
of more traditional marketing related posts. Such posts are characterized by a focus on firm or
product related content as well as product promotions or discounts. These are posts that do not
simultaneously convey any of the four online symbolic actions hypothesized in this study.
Additionally, it is conceivable that inducements such as sweepstakes, contests, giveaways, and
rewards may motivate people to join the online social community of a firm. To understand the
impact of these alternate explanations on online social community growth, our study measures
the impact of product and promotion related posts as well as posts that involve inducements.
Furthermore, previous literature indicates that user generated content (UGC) which refers to
posts by community members, is positively correlated to economic outcomes (Duan, Gu, and
Whicston 2008). In a similar vein it can be argued that UGC creates greater engagement and
hence larger online social communities. Our sample has only two firms that do not allow user
generated content that is allow their community members to create new posts on the firm’s
Facebook page. Since our sample does not have adequate variability with respect to firms that do
not allow UGS vs. firms that allow UGC, we do not control for UGC.
Working paper. Please do no not redistribute or cite without author’s permission.
19
Empirical Setting, Data, and Model
Empirical Setting and Sample
We test our hypothesis with a sample of firms in the online flash sales segment of the retail
industry. Such firms meet our criteria for analysis because they sell physical products and their
business model includes a significant online component. Further, the online flash sales segment
is one of the fastest growing e-commerce segments. The segment’s growth has accelerated since
2007 when the declining economy led to an unprecedented supply of excess inventory. Flash
sales were seen as a solution to move excess merchandise off the market quickly while
maintaining an aura of exclusivity (Cocotas 2012). These firms operate by purchasing excess or
out-of-season inventory at steep discounts from various brands. The products are then sold online
at a deep discount for a limited time (typically between a day and a week) to customers who
usually have to be members; members are invited by either the firm or by other members. The
need to market a new set of products on a weekly, or even daily, basis has led these firms to
depend heavily on email and marketing via social networking sites. In addition, the growth of the
segment since 2007 has given us a sample of several start-up firms founded around the same
time and operating under the same business model.
Our sample is based on listings from two directories of flash sales firms, Lokango.com and
FashionInvites.com as well as the business press. The latter source was included to compensate
for the time lag for some newly created firms to appear in the directories. The selected firms
covered all the major categories of flash sales businesses, namely apparel, household goods,
Working paper. Please do no not redistribute or cite without author’s permission.
20
travel and life style items. The criteria for inclusion in our sample are that the firm was in
business as of the time of data collection, has operations or an office location in the US, sells
products in the US, was founded in 20101, and has a Facebook fan page
2. We identified 23 firms
that met our criteria, and our goal was to obtain all Facebook content that is post text and weekly
fan count from the date each firm joined Facebook until July 1, 2012.
Dependent Variable
Our dependent variable is online social community size. We obtained the total Facebook fan
count per week for each firm in our sample from monitor.wildfireapp.com, This weekly data
spans the period of our study.
Independent Variables
Based on a literature review and our theory we identified thirteen dimensions associated with our
four online symbolic action constructs that reflect credibility, reflect professional organizing,
reflect organizational achievements, and seek opinions. Following Zaheer and Soda (2009), we
invited a panel of three experts to validate our constructs using the Q-Sorting technique (Segars
and Grover 1998). The list of 13 dimensions was randomly ordered and provided to a panel of
experts, who were asked to sort the dimensions into the four constructs. The instructions
included options for panel members to say that a dimension did not belong to any of the
constructs, or to say that a dimension belonged to multiple constructs. Based on the experts’
comments we reworded the dimensions related to two constructs. We then conducted another
round of surveys to validate our constructs; the percentage of correct classification and
1 In the case of firms with many subsidiaries, we include in the sample firms whose flash sales subsidiary was founded between 2010. 2 For firms that operate both a regular retail business and flash sales business we included only those firms that had a separate Facebook fan page for the flash sales part of their business.
Working paper. Please do no not redistribute or cite without author’s permission.
21
agreement was 100% for all three experts. Our independent variables were constructed by
categorizing each Facebook post in our sample along these dimensions.
The Facebook application programming interface (API) was used to collect the text of all past
posts for the firms in our sample, up to July 1, 2012. For some firms, we were not able to gather
posts all the way back to the date the firm joined Facebook, possibly because of restricted access
setting by some firms, or limits set by Facebook on historic data. This limitation combined with
the fact that different firms joined Facebook on different dates resulted in an unbalanced panel of
posts by each firm, that is each firm in the sample does not have posts for the same number of
weeks. A total of 12689 posts were collected.
Raters were asked to categorize each post in our sample of 12689 posts. This involved
determining whether the post belonged to any of the 13 dimensions associated with the
constructs. A web based questionnaire was given to the raters to perform the categorization
process. Measurement error due to context effects like grouping of items (Kline, Sulsky, and
Rever-Moriyama 2000) and item priming effects (Salancik 1984) was avoided by randomizing
the order of items that form the various constructs (Tourangeau and Rasinski 1988). The
questionnaire was peer reviewed to ensure principles of good item writing such as avoiding
double barreled questions, jargon, leading items, and negatively worded items. Further, the
questionnaire was pre-tested via a “think aloud” (Sudman, Bradburn, and Schwarz 1996) and via
a pilot test for 100 posts. We modified the questionnaire based on feedback from the think aloud
and the pilot tests. While we use a questionnaire for our independent variables, our dependent
variable is obtained from a different source, thus avoiding common source bias (King, Liu,
Working paper. Please do no not redistribute or cite without author’s permission.
22
Haney and He 2007).
Studies that analyze the content of textual data from social media typically use text mining tools
because the volume of data involved makes manual analysis difficult. For instance, Goh, Heng
and Lin 2013 use commercial text mining tools to capture the informative and persuasive nature
of content shared on Facebook. Text mining however, has many limitations. For example, it does
not analyze text embedded in images, or consider what the image itself conveys, nor does it
allow the analysis of the content of videos linked to posts. Also, text mining does not analyze
icons and symbol incorporated in posts. Depending on the way the text mining data is collected
and configured, the analysis may not include content linked to posts such as blogs and articles, as
wells as captions associated with photos. Further, fine-grained analysis of text can be challenging
when using text-mining tools. For example, in our study, we needed to differentiate between
three types of awards that may be mentioned in post content, that is awards to founders or key
employees, awards to the firm, and whether the firm sells award winning products. To allow for
a granular and nuanced analysis of posts we opted for manual rating of posts and used Amazon’s
Mechanical Turk (AMT) to find our raters. These raters not only analyzed the text of the post but
went to Facebook to find the actual post (based on identifying information) and rated the post
after reading articles linked to the post, watching videos embedded in the post, and scrolling
through photos associated with the post.
Since workers on AMT are significantly more diverse than workers from typical American
college samples (Buhrmester, Kwang and Gosling 2011) the probability of measurement error
due to response biases such as social desirability, and acquiescence (Bagozzi and Yi 1991) is
Working paper. Please do no not redistribute or cite without author’s permission.
23
reduced by using AMT workers. In addition, a study by Buhrmester et all (2011) indicates that
the data obtained from workers on Mechanical Turk are “at least as reliable as those obtained via
traditional methods” and that AMT can be used to obtain, “high-quality data inexpensively and
rapidly.” This is supported by the fact that workers AMT receive negative ratings that are
publicly displayed as part of the workers profile if their work does not meet the expectations or
quality standards or of the job requestor.
Further, to ensure consistency of understanding of the questions, terminology and task, raters
were trained via videos that we posted online. Additionally, to filter out raters who had a poor
understanding of the questions or the task we “pre-qualified” each rater by asking them to rate a
set of 15 test posts. The degree of agreement between raters was measured using Cohen’s kappa,
where the K value is interpreted as the degree of agreement between raters after taking into
account probability (Cohen 1960). Literature on using kappa, suggests that a coefficient of .61
indicates reasonably good overall agreement (Kvalseth 1989), so a rater was qualified if his or
her categorization of the test posts resulted in an overall kappa coefficient of > .61.
A total of 15 raters were qualified using this metric. Qualified raters were given online access to
categorize posts for this study and we were available by email to answer any questions. To
ensure that inter rater reliability was maintained, we periodically selected random posts for each
firm and checked the reliability of the rating by asking another rater to rate the randomly selected
posts. This scalable, manual process applicable to high volume content analysis that we
developed for qualifying and managing workers, and for evaluating their output, is illustrated in
Figure 3.
Working paper. Please do no not redistribute or cite without author’s permission.
24
For reliability testing, 370 randomly selected posts, approximately 3% of our original sample,
were examined by a second rater. The overall kappa coefficient for the 740 posts was 0.79. We
also examined the kappa coefficient for inter rater reliability for each firm. Three firms had a
kappa coefficient of <0.6 indicating that the raters did not have reasonably good overall
agreement. These three firms were dropped from our sample. Thus, this study develops and
employs a scalable and high quality manual content analysis process using the online work force
available through Amazon’s Mechanical Turk. This process enabled us to overcome the
previously stated limitations of text mining tools and resulted in nuanced content analysis.
An additional five firms were also dropped from our sample. Of these, two firms were dropped
because post content was not available possibly due to restricted access by the firms, and two
others were dropped because significant post content was unavailable (because the firm deleted
the content) after the firm had initially posted something. One firm was dropped because its
product line was focused on Indian fashion and home décor, and thus the post content targeted a
very narrow audience.
The three online symbolic actions that convey credibility, professional organizing, and
organizational achievements are multi-dimensional aggregate meta constructs (Law, Wong, and
Mobley 1998) that are formed as a composite of two or three sub-constructs (as indicated Figure
2). The sub-constructs under a particular construct may or may not co-vary. Further, each of
these sub constructs is composed of one or more dimensions. The construct corresponding to
seeking opinions has no sub-constructs and has only one dimension. Each dimension
Working paper. Please do no not redistribute or cite without author’s permission.
25
corresponds to a question that the rater rates about the post content. Figure 2 depicts the
relationship between the constructs and the sub-constructs. Table 1 lists the constructs, sub-
constructs, and dimensions underlying the sub-constructs along with their corresponding survey
question(s).
Control Variables
We include two control variables that relate to the alternate explanations discussed earlier. The
first is a control for posts that invite people to participate in contests, giveaways and
sweepstakes. The second control variable is for posts that focus on firm or product related
content as well as product promotions or discounts. These are posts that do not simultaneously
convey any of the four online symbolic actions hypothesized in this study.
Model Specification
The final sample included 15 firms and a total of 9,826 posts. The median number of posts
analyzed per firm is 674 (mean 655). The median number of weeks of data per firm that was
analyzed is 67 (mean 57). Only two firms that were analyzed had less than 50 weeks of data.
We constructed a weekly panel data set that included our independent and dependent variables
and the controls. The independent variables and controls were computed by summing weekly
scores corresponding to sub-constructs and constructs.
We estimate the following model:
Working paper. Please do no not redistribute or cite without author’s permission.
26
Fan Countit = β0ij + β1Fan Counti(t-1) + β2Credibilityit + β3Professional Organizingit +
β4Organizational Achievementit + β5Opinionit + β5Contestit + β7 ProductOrSaleit + uit (1)
In the equation above, Fan Countit is the size of the Facebook community for firm i during week
t and Fan Counti(t-1) is its lagged value by one week. The equation also includes the four
constructs and the two control variables discussed earlier.
There are a few potential concerns with a simplistic regression of the model presented in
equation 1. First, it is conceivable that an unmeasured variable like existing status or legitimacy
of a firm is driving growth in the fan base. This is addressed, at least in part, by the fact that our
sample consists of new ventures that don’t have the history and the legitimacy that is associated
with well-established firms. Second, omitted variables such as the sub-segment that the firm
belongs to for example apparel or household goods, may impact one or more of the explanatory
variables and also the fan count. This form of endogeneity causes biased estimates (Wooldridge
2002). Third, the presence of the lagged dependent variable Fan Counti(t-1) gives rise to
autocorrelation. To address the latter two issues, we use the to use the Arellano-Bond (1991)
difference GMM estimator. The second concern, omitted variables, is addressed so long as the
omitted variables are time invariant, that is, they are fixed effects (Mileva 2007). All of the firm
fixed effects are contained in the error term, uit, in equation 1, which is composed of both the
unobserved firm specific effects, vit,, and the observation-specific errors, eit,. The difference
GMM uses first-differences to transform equation (1) into:
ΔFan Countit = β0ij + β1ΔFan Counti(t-1) + β2ΔCredibilityit
Working paper. Please do no not redistribute or cite without author’s permission.
27
+ β3ΔProfessional Organizingit + β4ΔOrganizational Achievementit
+ β5ΔOpinionit + β5ΔContestit + β7 ΔProductOrSaleit + uit (2)
Since the unobserved firm specific effects, vit, are time invariant, the fixed firm-specific effect is
removed, and from equation 2 we get:
uit-ui,t-1= (vi− vi)+( eit - ei,t-1) = eit - ei,t-1 (3)
The third concern, autocorrelation, is addressed because the dynamic panel model instruments
the first-differenced lagged dependent with its lagged levels (Mileva 2007).
For robustness we also estimate our equation using Negative Binomial, and OLS models. For
these two models our dependent variable was the growth in fan count each week, which is
computed as the fan count for the current week minus the fan count for the previous week. While
estimating these models we included the previous week’s fan count value as a control variable
and dummy variables for firm and week. For the Negative Binomial model we had to shift he
dependent variable because this model accepts only positive values, and in our study there are
instances where the fan count drops across weeks, resulting in a negative value. Further, this
count model may not be the most appropriate for very large count values (as is the case for the
fan count in our study) that are heavily skewed.
Working paper. Please do no not redistribute or cite without author’s permission.
28
Results
Table 1 provides descriptive statistics for our variables, and Tables 2 and 3 provide correlations
for the variables at the construct and sub-construct levels respectively. We do not find
collinearity among the independent variables to be an issue.
Table 4 presents the results of the estimation from equation (1) that examines the impact of
online symbolic actions on fan count. The results reveal a positive and statistically significant
impact of symbolic actions reflecting credibility, and thus support H1. The coefficients for
symbolic actions reflecting professional organizing and organizational achievements are not
statistically significant. Hence we do not find initial support for H2 and H3. Further, contrary to
our expectations the results reveal a negative and statistically significant impact of symbolic
actions seeking opinions. Thus for H4 we find that the effect is in the opposite direction from
what we had predicted. We explain this result later in this section. With respect to our controls,
we find that contests are not associated with online social community growth, while posts with
content about the firm’s products or about promotions (that do not simultaneously exhibit any of
the for online symbolic actions hypothesized in this study), are associated with online social
community growth.
For comparative purposes, results from the dynamic panel model are show along with the results
from the OLS and the Negative Binomial models in Table 4. The R–square value from the OLS
model is 38%. The results from the OLS the Negative Binomial models for the four main
variables from our hypotheses are consistent with those of the dynamic panel model, except in
Working paper. Please do no not redistribute or cite without author’s permission.
29
the case of the Negative Binomial model for symbolic actions seeking opinions, which is found
to be not statistically significant.
To understand what is driving our results we examine the impact of each of the sub-constructs
associated with the four symbolic actions on fan count. The equation we estimate at the sub-
construct level is:
Fan Countit = β0ij + β1Fan Counti(t-1) + β2Capabilityit + β3Association With Expertsit + β4 Product
or Industry Knowledgeit + β5 Professional Processit + β5 Professional Structureit + β7 Firm
Accomplishmentit + β8 ProductAffirmationit + β9 Opinionit + β10Contestit + β11ProductOrSaleit
+uit (4)
Results from this estimation are shown in Table 5. Of the three sub-constructs that constitute
credibility, namely, personal capability, access to experts and product and industry knowledge
we find that product and industry knowledge (for example content that shows how to use a
product, broad educational information, and industry related news) is positively associated with
online social community growth. In addition, we find that of the two sub-constructs that
represent organizational achievement that is product affirmation and firm level accomplishments
such as awards and milestones, firm accomplishments are positively associated with online
social community growth. This finding provides partial support for H3. These two findings
indicate a form of preferential attachment wherein people want to associate online with firms that
they perceive to be knowledgeable and noteworthy in terms of achievements. None of the
Working paper. Please do no not redistribute or cite without author’s permission.
30
coefficients of the sub-constructs that constitute professional organizing were found to be
statistically significant. Thus analysis at the sub-construct level analysis lends no support for H2.
As with the previous model (at the construct level) we find that seeking opinions is negatively
associated with online social community growth. The volume of information that users on social
networking sites are exposed to exceeds their capacity to consume it (Weng Flammini,
Vespignani and Menczer 2012). Although questions posed on social media don’t necessarily
require a response, asking a question places additional demands on users’ limited time, and a
firm that asks too many questions may seem onerous to its community members. We attribute
the negative association between asking for opinions and online social community growth to the
fact that some firms are posing too many questions to their community members. We find that
some firms in our sample ask on average between 6 and 12 questions a week.
Once again, for comparative purposes, results from the dynamic panel model are show along
with the results from the OLS model in Table 5. The R–square value from the OLS model is
39%. The results from the OLS model for the sub-constructs associated with our hypotheses and
for the two control variables are consistent with those of the dynamic panel model.
To be able to better interpret our coefficients we estimate our model using the natural log of the
dependent variable. Given the skewed nature of our dependent variable this improves the fit for
the OLS model. A normally distributed dependent variable, however is not an assumption for the
dynamic panel model. The equations we estimate at the construct and sub-construct levels are
Working paper. Please do no not redistribute or cite without author’s permission.
31
Ln(Fan Count) it = β0ij + β1Fan Counti(t-1) + β2Credibilityit + β3Professional Organizingit +
β4Organizational Achievementit + β5Opinionit + β5Contestit + β7 ProductOrSaleit + uit (5)
Ln(Fan Count) it = β0ij + β1Fan Counti(t-1) + β2Capabilityit + β3Association With Expertsit + β4
Product or Industry Knowledgeit + β5 Professional Processit + β5 Professional Structureit + β7
Firm Accomplishmentit + β8 ProductAffirmationit + β9 Opinionit + β10Contestit +
β11ProductOrSaleit +uit (6)
Estimates for the models at the construct level, that is, equation 5, are shown in Tables 6.
Estimates for the models at the sub-construct level, that is equation 6, are shown in Table 7. The
significance and the direction of the coefficients are the same as the estimates for the model
represented by equation 1. We find that a unit increase in online symbolic actions reflecting
product and industry knowledge is associated with an average increase in fan count by 0.2%3,
Similarly a unit increase in online symbolic actions showing firm achievements is associated
with an average increase in fan count by 1.3%. However, a unit increase in online symbolic
actions seeking opinions is associated with an average decrease of fans by 0.2%. Mangers we
talked to feel that if using online symbolic actions contributes to even a small (5%) increment in
their online population growth, it is worth pursuing. In one entrepreneur’s words, “Every little
increase matters.” This information gives us an organizational benchmark to evaluate magnitude
of effect for symbolic action on online social community size.
3 computed as 100(exp(b)-1)
Working paper. Please do no not redistribute or cite without author’s permission.
32
Contributions
Our study makes three contributions to research. To begin with, the study defines online
symbolic actions and describes how they are performed. The study identifies which symbolic
actions from previous literature translate effectively to the social domain and also suggests some
additional online symbolic actions that are relevant to social media sites. In addition, the study
empirically examines the impact of various types of online symbolic actions on online social
community growth. Furthermore, the study suggests a theoretical model whereby legitimacy is
created via social networking sites.
Our empirical analysis points to three important findings. First, online symbolic actions that
show product and industry knowledge, which is a component of online symbolic actions
reflecting credibility, are associated with online social community growth. Next, online symbolic
actions that show firm accomplishments, which is a component of online symbolic actions
reflecting organizational achievements, are also associated with online social community growth.
Last, contrary to our predictions we find that online symbolic actions that seek opinions have a
negative impact on online social community growth. The first two findings indicate a form of
preferential attachment at work in the dynamics of social networking sites. The third finding can
be attributed to an over use of questions on social networking sites.
This study is relevant to managers and timely because both new ventures and incumbents are
rapidly adopting the use of social networking sites. While several social networking analytics
sites provide entrepreneurs with up to the minute social networking sites statistics such as:
number of fans and engagement level, the analytics sites leave it to the firm to determine what
Working paper. Please do no not redistribute or cite without author’s permission.
33
leads to better or worse performance on these metrics. By associating online symbolic actions
with online social community growth, the study provides one possible path to growing online
social communities and to legitimacy. The latter facilitates resource acquisition, which is critical
to the survival of new ventures. Conversely, the study also provides some understanding about
what kinds of online actions will lead to a decline in online social community size. Our study has
wide applicability as it encompasses online B2C new-ventures from various industries that are
product focused.
Limitations and Directions for Future Research
One limitation of this study is that it does not consider the quality of the posts by the firm. Apart
from the fact that quality of posts is a subjective measure that is hard to define and measure, the
volume of data in this study would make it difficult to systematically hand code the quality of
each post.
Another limitation is that while it is conceivable that some firms purchase ‘likes’ or fans on
Facebook and thus increase the size of their online social community. Our study is unable to
control for this directly. However, to the extent that this may be an ongoing practice by a
particular firm, the practice can be thought of as a time invariant fixed effect that gets washed
out. The reason that this may be an ongoing practice rather than a one-time activity is that a
sudden surge in fans would seem suspect. Further, Facebook has never permitted the purchase or
sale of Facebook Likes, and their website indicates that detection of such activity could result in
action against the firm by Facebook. This may be a deterrent, since the firms in our study belong
to the Flash sales industry that relies heavily on social networking sites marketing due to their
Working paper. Please do no not redistribute or cite without author’s permission.
34
constantly changing inventory
In terms of future research, there are several interesting research questions related to the creation
of legitimacy through the use of social networking sites. For instance, what impact does user
generated content have on legitimacy, what types of user generated content leads to legitimacy
and what actions can the firm take to stimulate such user generated content, are promising
questions for future research.
Working paper. Please do no not redistribute or cite without author’s permission.
35
Figure 1: Symbolic action and social diffusion work together to grow the online social
community
Working paper. Please do no not redistribute or cite without author’s permission.
36
Figure 2: Online symbolic actions (construct and sub-construct level) that lead to online social
community growth
Working paper. Please do no not redistribute or cite without author’s permission.
37
Figure 3: Process for qualifying workers and rating posts
Working paper. Please do no not redistribute or cite without author’s permission. 38
Working paper. Please do no not redistribute or cite without author’s permission. 39
Working paper. Please do no not redistribute or cite without author’s permission. 40
Working paper. Please do no not redistribute or cite without author’s permission. 41
Working paper. Please do no not redistribute or cite without author’s permission.
42
Working paper. Please do no not redistribute or cite without author’s permission.
43
Working paper. Please do no not redistribute or cite without author’s permission.
44
Working paper. Please do no not redistribute or cite without author’s permission.
45
Working paper. Please do no not redistribute or cite without author’s permission.
46
References:
Aral, S., D. Walker. 2011. Creating social contagion through viral product design: A randomized
trial of peer influence in networks. Management Science, 57(9) 1623-1639.
Arellano, M., S. Bond. 1991. Some tests of specification for panel data: Monte Carlo evidence
and an application to employment equations. The Review of Economic Studies 58 277-297.
Aldrich, H. E., C. M. Fiol. 1994 Fools rush in? The institutional context of industry creation.
Academy of Management Review 19 645–670.
Bagozzi, R. P., Y. Yi. 1991. Multitrait-multimethod matrices in consumer research. Journal of
Consumer Research 17:4 426-439.
Bagwell, L. S., D. B. Bernheim. 1996. Veblen effects in a theory of conspicuous consumption.
American Economic Review 86 349–373.
Brown, A. D. (1994). Politics, Symbolic Action and Myth Making in Pursuit of Legitimacy. Organization Studies (Walter De Gruyter Gmbh & Co. KG.), 15(6), 861.
Buhrmester, M., T. Kwang, S. D. Gosling. 2011. Amazon's Mechanical Turk : A New Source of
inexpensive, yet high-quality, data? Perspectives on Psychological Science 6: 3.
Citibank and GFK Roper. 2009. Facebook, Twitter and LinkedIn not seen as helpful to build
small businesses http://www.prnewswire.com/news-releases/citibank-survey-reveals-small-
businesses-not-joining-social-media-conversation-63837487.html, accessed Feb 2012.
Cocotas. 2012. Flash sales will be a $6 billion market by 2015. Business Insider, BI Intelligence.
http://articles.businessinsider.com/2012-01-23/research/30646020_1_flash-sales-e-commerce-
market#ixzz1rxbfG8oW accessed on April 13, 2012.
Cohen, J. 1960. A coefficient of agreement for nominal scales. Educational and Psychological
Measurement 20 7–46.
Collins English Dictionary. Complete & unabridged 10th Edition. HarperCollins Publishers.
Accessed 02 Aug. 2013. Dictionary.com http://dictionary.reference.com/browse/credibility.
Comscore. 2011. It’s a Social World: Top 10 Need-to-Knows About Social Networking and
Where It’s Headed
http://www.comscore.com/Insights/Presentations_and_Whitepapers/2011/it_is_a_social_world_t
op_10_need-to-knows_about_social_networking accessed on Aug 29, 2013.
Comscore. 2012. The Power of like
2: How social marketing works. White Paper
http://www.comscore.com/Insights/Presentations_and_Whitepapers/2012/The_Power_of_Like_2_How_Social_Mar
keting_Works. Accessed on Sept 1, 2013.
Working paper. Please do no not redistribute or cite without author’s permission.
47
DiMaggio P. J., W. W. Powell. 1983. The iron cage revisited: Institutional isomorphism and
collective rationality in organizational fields. Administrative Sociological Review 48 147-160.
Duan, W., B. Gu, A. Whinston. 2008. Do online reviews matter? An empirical investigation of
panel data. Decision Support Systems 45:4 1007-1016.
Ellison, N. B., D. Boyd. 2013. Sociality through social network sites. In Dutton, W. H. (Ed.), The
Oxford Handbook of Internet Studies. Oxford: Oxford University Press, 151-172.
Goh, K. Y., Heng, C. S., & Lin, Z. 2013. Social media brand community and consumer behavior:
Quantifying the relative impact of user-and marketer-generated content. Information Systems
Research, 24(1), 88-107.
Higgins, M. C., R. Gulati. 2003. Getting off to a good start: The effects of upper echelon
affiliations on underwriter prestige. Organization Science 14 244–264.
Holtz-Eakin, D., W. K. Newey, H. S. Rosen. 1988. Estimating vector autoregressions with panel
data. Econometrica 56 1371–1395. <this is cited in the footer of one of our tables>
Human, S. E., K. G. Provan. 2000. Legitimacy building in the evolution of small-firm
multilateral networks: A comparative study of success and demise. Administrative Science
Quarterly 45(2) 327-365.
Johnson. 2012. The state of corporate social networking sites. Useful Social networking sites.
Kaplan A, M., M. Haenlein. 2010 Users of the world, unite! The challenges and opportunities of
Social networking sites. Business Horizons Volume 53, Issue 1, January–February 2010 59–68.
King, W. R., C. Z. Liu, M. H. Haney, J. He. 2007. Method effects in IS survey research: An
assessment and recommendations. Communications of the Association for Information Systems
20:30.
Kline, T. J. B., L. M. Sulsky, S. D. Rever-Moriyama. 2000. Common method variance and
specification errors: A practical approach to detection. Journal of Psychology 134:4
401.
Kostova, T., S. Zaheer. 1999. Organizational legitimacy under conditions of complexity: The
case of the multinational enterprise. Academy of Management review 64-81.
Kvalseth, T. O. 1989. Note on Cohen’s kappa. Psychological Reports 65 223–226.
Law, K. S., Wong, C. S., Mobley, W., H. 1998. Toward a taxonomy of multidimensional
constructs. Academy of Management Review 23 741–755.
Lievens, F., S. Highhouse. 2003. The relation of instrumental and symbolic attributes to a
company’s attractiveness as an employer. Personnel Psychology 56 75–102.
Working paper. Please do no not redistribute or cite without author’s permission.
48
Lipsman, A., G. Mudd, M. Rich, S. Bruich. 2011. The Power of like: How brands reach and
influence fans through social media marketing. comScore White Paper.
Lipsman, A., G. Mudd, C. Aquino, P. Kemp. 2012. The Power of like2: How social marketing
works. comScore White Paper
Lounsbury, M., M. Glynn. 2001. Cultural entrepreneurship: Stories, legitimacy, and the
acquisition of resources. Strategic Management Journal 22 545–564.
Meyer, J. W., B. Rowan. 1977. Institutionalized organizations: Formal structure as myth and
ceremony. American journal of sociology 340-363
Meyer, J. W., & Rowan, B. 1991. Institutionalized organizations: Formal structure as myth and
ceremony. In W. W. Powell & P. J. DiMaggio (Eds.), The new institutionalism in organizational
analysis: 41-62. Chicago: University of Chicago Press.
Mileva, E. 2007. Using Arellano-Bond dynamic panel GMM estimators in Stata. Economic
Department, Fordhan University.
Mzinga, Inc., Babson Executive Education. 2009. Survey: Social software in business.
http://mouriz.files.wordpress.com/2010/02/mzingababson-socialsoftwaresurvey.pdf, accessed on
July 18, 2013.
Oldham, G. R., N. L.Rotchford. 1983. Relationships between office characteristics and employee
reactions: A study of the physical environment. Administrative Science Quarterly 28 542–556.
Oliver, C., 1991"Strategic responses to institutional processes," Academy of Management Review. 16: 145-179,
Rafaeli, A., I. Vilnai-Yavetz. 2004. Instrumentality, aesthetics and symbolism of physical
artifacts as triggers of emotion. Theoretical Issues in Ergonomics Science 5 91–112.
Rao, H. 1994. The social construction of reputation: certification contests, legitimation and the
survival of organizations in the American automobile industry: 1895-1912. Strategic
Management Journal 15 29-44.
Ruef, M., W. R. Scott. 1998). A multidimensional model of organizational legitimacy: Hospital
survival in changing institutional environments. Administrative Science Quarterly 877-904.
Salancik, G. R. 1984. On priming, consistency, and order effects in job attitude assessment:
with a note on current research. Journal of Management 10 250-254.
Santos, F.M., K. M. Eisenhardt. 2009. Constructing markets and shaping boundaries:
Entrepreneurial power in nascent fields. Academy of Management Journal 52:4 643-671.
Scott, W. R., & Meyer, J. W. 1991. The organization of societal sectors. In W. W. Powell & P. J.
Working paper. Please do no not redistribute or cite without author’s permission.
49
DiMaggio (Eds.), The new institutionalism in organizational analysis: 108-140. Chicago:
University of Chicago Press.
Segars, A. H., V. Grover. 1998. Strategic Information systems planning success: An
investigation of the construct and its measurement. MIS Quarterly 22 139-
163.
Shane, S. A. (2008). The illusions of entrepreneurship: The costly myths that entrepreneurs,
investors, and policy makers live by. Yale University Press
Starr J. A., I. C. MacMillan. 1990. Resource cooptation via social contracting: Resource
acquisition strategies for new ventures. Strategic Management Journal Special Issue: Corporate
Entrepreneurship (Summer, 1990), 11 79-92.
Stinchcombe, A. L. 1965. Social structure and organizations. In J. G. March (Ed.), Handbook of organizations: 142-193.
Chicago: Rand-McNally.
Suchman, M. C. 1995. Managing legitimacy: Strategic and institutional approaches. Academy of
Management Review 20 571–610.
Suddaby, R., R. Greenwood. 2005. Rhetorical strategies of legitimacy. Administrative science
quarterly 50(1) 35-67.
Sudman, S., N. M. Bradburn, N. Schwarz. 1996. Thinking about answers: The application
of cognitive processes to survey methodology. Jossey-Bass.
Tourangeau, R., K. A. Rasinski. 1988. Cognitive processes underlying context effects in
attitude measurement. Psychological Bulletin 103 299-314.
Weng, L. Flammini, A. Vespignani, A. Menczer, F. Competition among memes in a world with
limited attention. Science Reports 2012/03/29/online VL-2 Macmillan Publishers Limited.
http://dx.doi.org/10.1038/srep00335
Wooldridge J. M. 2002. Econometric analysis of cross section and panel data. The MIT Press:
London.
Zaheer A., G. Soda. 2009. Network evolution: The origins of structural holes. Administrative
Science Quarterly 54(1) 1-31.
Zimmerman, M.A., G.J. Zeitz. 2002. Beyond survival: Achieving new venture growth by
building legitimacy. Academy of Management Review 27(3) 414-431.
Zott C., Q. N. Huy. 2007. How entrepreneurs use symbolic management to acquire resources.
Administrative Science Quarterly 52 70–105
Recommended