SlideShare a Scribd company logo
LET ME TELL YOU A STORY: AN EXPLORATION OF THE COMPLIANCE-
GAINING EFFECTS OF NARRATIVE IDENTITIES IN ONLINE CROWDFUNDING
TEXTUAL APPEALS
by
Kenton Bruce Anderson
December 21, 2015
A dissertation submitted to the
Faculty of the Graduate School of
the University at Buffalo, State University of New York
in partial fulfillment of the requirements for the
degree of
Doctor of Philosophy
Department of Communication
ii
Acknowledgments
Dr. Gregory D. Saxton (Major Advisor), Dr. Michael A. Stefanone (Committee
Member), Dr. Joseph Woelfel (Committee Member), Dr. Tom Feeley, Rose Gryckiewicz,
Dr. Mark Frank, Aubrey Nye, Kelvin Anderson and my family, Edward J. Marinucci, Jr.,
Dr. Allison Zorsie-Shaw, Dr. Matthew Grizzard, Dr. Jacob Neiheisel, Emily Dolan, Dr.
Rosemarie Murray, Dr. Sean Dugan, Professor Paul Trent, Anne Slowe, Zed Ngoh &
family, Phil Teefy, Gil and Carmen Reynolds (Hot Glass Horizons), Jonathan Schwartz,
Dr. and Mrs. Ronald and Amelia Schwartz, Dr. Mary Flaherty Rogers, Dr. Wayne Xu,
Timothy Pruitt, Isabella Wilklow
iii
Table of Contents
Acknowledgements………………………………………………………………............. ii
Abstract…………………………………………………………………………............. vii
Chapter 1: Introduction........................................................................................................1
Background of the Problem ...........................................................................................1
Statement of the Problem...............................................................................................9
Purpose of the Study......................................................................................................9
Significance of the Study.............................................................................................11
Primary Research Questions........................................................................................11
Research Design...........................................................................................................12
Theoretical Framework................................................................................................14
Assumptions, Limitations, and Scope (Delimitations) ................................................15
Definition of Terms......................................................................................................16
Summary......................................................................................................................17
Chapter 2: Literature Review.............................................................................................19
Narrative as a Sequential Compliance-gaining Technique..........................................19
Donations, Non-profits, Online Giving, SNSs, P2P Lending and
Crowdfunding ..................................................................................................19
Narrative ......................................................................................................................22
Identity Claims, Trust, and Need.................................................................................25
Multiple Identity Claims..............................................................................................28
Time to Funding………………………………………………………………...……29
iv
Micronarratives............................................................................................................29
Why Measure Micronarratives?............................................................................ 31
Archetypes ............................................................................................................ 36
Relationship Between the Independent and Dependent Variables ..............................36
Goals ..........................................................................................................................37
Research Questions......................................................................................................39
Chapter 3: Research Method..............................................................................................41
Sample Selection Process ............................................................................................41
Collection Method ................................................................................................ 43
Coding................................................................................................................... 43
Structural Variables .....................................................................................................46
Measurement................................................................................................................46
Chapter 4: Coding..............................................................................................................50
Replicated Codes .........................................................................................................50
Raw Data Preparation ..................................................................................................51
Text Reading......................................................................................................... 51
Category Creation................................................................................................. 51
Uncoded or Overlapping Text .............................................................................. 52
Category Revision and Refinement ...................................................................... 52
Initial Coding Categories .............................................................................................53
Trustworthy........................................................................................................... 53
Successful ............................................................................................................. 53
v
Hardworking ......................................................................................................... 54
Economic Hardship............................................................................................... 54
Moral ................................................................................................................... 54
Religious ............................................................................................................... 55
Additional Coding Categories......................................................................................56
Needy ................................................................................................................... 57
Helpless................................................................................................................. 57
Victim ................................................................................................................... 57
Child-mentioned ................................................................................................... 58
Child-centered....................................................................................................... 58
Courageous ........................................................................................................... 58
Grateful ................................................................................................................. 59
Hero ................................................................................................................... 59
Hip ................................................................................................................... 60
Happy ................................................................................................................... 61
Chapter 5: Study 2 (Multivariate Analyses) Results..........................................................62
Chapter 6: Discussion, Conclusions, and Recommendations............................................70
Theoretical Implications ..............................................................................................72
Limitations and Future Research .................................................................................75
Practical Implications...................................................................................................76
Conclusion…………………………………………………………………………...77
Appendix 1: Tables............................................................................................................78
vi
Appendix 2: Figures...........................................................................................................85
References………………………………………………………………………………..90
vii
Abstract
This study assesses the compliance-gaining impact of specific micronarrative
identities on non-profit fundraising in the online environment. It examines the effects of
narratives in crowdfunding requests in the non-profit category of entities on the peer-to-
peer (P2P) crowdfunding website Indiegogo. The study has several goals: 1) distinguish
micronarrative identity claims from longer narratives and explore the compliance-gaining
influence mechanisms behind both simple, micronarrative identity claims and longer
narratives, by exploring previously studied and newly described identity claim categories
in the texts of non-profit donation requests; 2) explore the relevance of recent claims
about trust in the for-profit lending realm for their applicability to the realm of non-profit
donations; and 3) determine the relevance to nonprofit funding requesters of narratives
relaying need information. In study #1, in-depth inductive analyses were conducted to
identify the types of narratives nonprofit entrepreneurs are employing in their
crowdfunding requests, while in study #2 multivariate analyses were used to investigate
the relationship between the use of micro-narratives and the success of crowdfunding
requests. The study found narratives containing mentions of need, children and victim
terminology have a significant impact on funding success rates for non-profit lending
requests. In addition, hipster and hero narratives significantly affect the number of
funders. The study also supports previous research finding that the discrete number of
narratives is negatively correlated with funding success. By implication, having a
coherent narrative seems more important. Secondly, regarding other elements of the
campaign, such as making regular updates, adding photos and perk levels, and longer
viii
campaigns – it appears that providing current information and visual imagery leads to
greater funding success, as does providing more choices of perk levels. Finally, with the
narrative structure variables, longer campaigns help, as does having one’s request
professionally written or being low in illiteracy. These findings have potential
implications for requesters attempting to crowdsource funds on Internet lending sites.
ix
Keywords: crowdsourcing, crowdfunding, nonprofit fundraising, donations,
Indiegogo, micronarrative identity, compliance gaining, P2P lending, sequential
messages, narrative
Chapter 1: Introduction
Online funding is a growing phenomenon in the 21st
Century, as is evidenced by
multiple websites devoted entirely to funding individual needs, projects, and investments.
Some websites sponsor requests from for-profit entities, while others sponsor requests
from not-for-profit entities. These entities can include both institutions and individuals.
Some websites sponsor assorted requests from both for-profit and not-for-profit entities.
How these requests are structured to maximize their effectiveness is of importance to
both requesters and funders. The literature of self-presentation has explored this from
several approaches, including cross-cultural and network analysis (e.g. Rui & Stefanone,
2013a; Rui & Stefanone, 2013b). The story, or narrative, that requesters tell about
themselves to identify themselves to their audience is one part of the request’s structure
that may significantly impact their funding success or failure. This two-phase study
explores the use of narratives in the requests in order to decide what specific narrative
identities or other narrative elements are used in the more successful posts and,
conversely, which narrative identities or other elements requesters might be advised to
avoid using to ensure greater success.
Background of the Problem
In the postmodern era, the rise in the popularity of crowdsourcing and
crowdfunding is changing the arena of fundraising for both individuals and institutions
(Dholakia and Firat, 2006; Hibbert and Horne, 1996). Because individuals comprise 80%
of funders (Hibbert and Horne, 1996), however, it is especially important to explore
persuasion processes that impact individual funders. Consequently, institutional giving is
2
beyond the scope of this study. Discovering what we can know about individuals’
motivations for giving is therefore an important step in maximizing non-profit online
fundraising attempts, a topic of relevance not only to individuals, but also to institutions
and governments (Psacharopoulos and Nguyen, 1997). With a few important exceptions,
the empirical foundations of the online crowdfunding phenomenon have generally been
unexplored in the scientific literature (e.g. Herzenstein, Sonenshein, and Dholakia, 2011).
While researchers have begun exploring the impact of visual images in online
non-profit persuasive appeals (e.g. Anderson and Saxton, 2016), of particular relevance
to the present study are the persuasive factors in textual appeals that motivate individuals
to give to other individuals or institutions. Some communication scholars might assume
that such textual persuasive factors are limited primarily to argumentative elements such
as evidence, logic, cognition, nonverbal elements, or formal techniques as exemplified by
the currently conceptualization of sequential messages (e.g. Dillard, Hunter, and
Burgoon, 1984). However, it has become apparent to some communication theorists (e.g.
Green, Garst, and Brock, 2004) that an important source of data describing individual
persuasive communication processes has been largely ignored -- that of narrative -- and
this growing number of empirical researchers and theorists have begun to align
themselves with the postmodern focus upon narrative, or storytelling elements, that is
influencing empirical research in many other disciplines (see Reissman and Quinney,
2005).
Understanding the current state of the scientific exploration of persuasive textual
elements in online posts requires a beginning distinction between rhetorical and narrative
3
approaches to persuasion. In their seminal work, Green, Garst and Brock (2004, p. 162)
say persuasion research has primarily focused on rhetorical messages, rather than
narratives. They define rhetoric as: “primarily fact-based advocacy messages, such as
advertisements, speeches, and editorials, that contain arguments specifically designed to
sway a reader to a particular position (p. 162).” But narrative content -- which is
distinguished as containing characterization, plot elements and setting information -- is at
least as persuasively powerful as rhetorical “factual” statements, according to Strange and
Leung (1999) and Green and Brock (2000). This paradoxical state of the literature,
which has continued nearly unabated to the present -- as I will show in the following
chapter -- indicates the importance of studying narrative elements in persuasion in
addition to rhetorical elements.
Understanding the foregoing distinction between rhetorical and narrative
approaches to researching persuasion is only one step in discovering the persuasive
impact of narratives, however. In order to fully conceptualize the persuasive influence of
narratives, and more fully tie them to communication theory, it might be instructive to
show their relevance to a rather well-developed field of empirical research and theory in
communication that has previously not been seen by most scholars as connected to
narrative: the realm of sequential message compliance-gaining techniques. This
conceptualization of them as sequential compliance-gaining message techniques is based
upon the general consensus of narrative researchers on the central importance of their
sequential nature (Czarniawska, 1998; Herzenstein, et al. 2011; Martens, Jennings, and
Jennings, 2007; Merchant, Ford, and Sargeant, 2010; Riessman, 1993).
4
Business and marketing researchers, particularly those studying
entrepreneurialism and banking have been the primary explorers of persuasion and
funding as related to identity presentation. Given that funding is a major focus of activity
for not only business entities such as entrepreneurs, investors, and lenders, but also for
charitable organizations and donation requesters, researchers have generally assumed that
the two areas might offer pertinent insights for each other (e.g. Bennett, Foot, and Xenos,
2011; Boje, 1991; Dholakia and Firat, 2006; Fiol, 1989; Hjorth and Steyaert, 2005; Hytti,
2003; Makkonen, Aarikka-Stenroos and Olkkonen, 2012; Martens, Jennings, and
Jennings, 2007; Rhodes and Brown, 2005; Schembri, Merrilees, and Kristiansen, 2010).
However, some conflicting research indicates that there may be important differences
between effective fundraising for profit-making entities and effective fundraising for non-
profit donation seekers. It appears that the choice of self-identities or organizational
identities is a potentially problematic practice that may not be equivalent across both for-
profit and non-profit narrative fundraising attempts. Research to date has still left
unresolved the question of how these two groups might need to differentiate their
approaches to narrative self identity presentation, an issue the present study attempts to
help clarify.
Importantly, a preliminary question that presages the effectiveness of particular
appeals across for-profit and non-profit arenas is whether storytelling can actually help to
secure funding in non-profit settings. While a form of this question has been addressed
briefly in the business literature, which has observed such an effect in for-profit
fundraising (Herzenstein, Sonenshein, and Dholakia, 2011; Martens, Jennings, and
5
Jennings, 2007), this preliminary question has yet to be addressed in the non-profit
fundraising literature. The task for the current study is to determine a methodological
approach which might help shed first light on these foundational questions: Are stories
containing identity narratives related to the success of non-profit funding appeals? How?
Why? In chapter three, I will outline the methodological approach our limited sample of
identity-making on a real-life fundraising website allows us to use to flesh out the
connection between storytelling and funding success.
The scant existing empirical research on the impact of narrative elements in
online fundraising persuasiveness (Merchant, Ford, and Sargeant, 2010) has identified
several such elements that appear related to effective fundraising appeals, though this
research has to date only explored these elements as used in for-profit appeals and in
entrepreneurial settings (e.g. Herzenstein, Sonenshein, and Dholakia, 2011; Martens,
Jennings, and Jennings, 2007; Navis and Glynn, 2011). These elements derive from the
processes of individual and organizational identity creation engaged in by individual
requesters, entrepreneurs and impression managing entities. These seminal research
studies on funding requests in for-profit lending and entrepreneurial image management
present evidence that trust and risk concerns are influential in financial decision-making.
Consequently, narrative identities that address these concerns are predictive of effective
fundraising (Herzenstein, Sonenshein, and Dholakia, 2011) in the for-profit realm
explored in these studies. What remains unanswered is whether this model holds also for
non-profit fundraising, or whether different narrative elements will emerge as more
relevant in that arena -- particularly given that theory in charitable fundraising proposes
6
that trust and risk are less relevant to donors than are determination of need and
dedication to the cause (e.g. Bekker and Wiepking, 2011).
Theorists of non-profit fundraising propose that donor assessment of need and
belief in the cause espoused by the non-profit entity will each predict giving behavior
better than will trust or risk considerations (Bekkers and Wiepking, 2011). To date,
empirical support for this claim appears to be nonexistent. However, the recent study by
Herzenstein, Sonenshein, and Dholakia (2011) presents limited empirical support for the
relevance of trust-based narrative identities in for-profit lending considerations. There is a
paucity of research addressing this issue. Therefore, although this dataset doesn’t allow
us to measure level of belief in the cause -- either directly or by proxy -- we will explore
the relative effectiveness of trust-based narrative identities versus need-based narrative
identities in predicting the successful funding of the requests gathered in the present
sample set. Of particular significance will be what direction such potential relationships
exhibit, if found, and what other narrative variables besides trust might be relevant to
funding success.
Given the centrality of exploring narrative self-identities in the empirical literature
on narrative persuasion (see Green, Garst, and Brock, 2004; Herzenstein, Sonenshein,
and Dholakia, 2011), and the presumed use of them to present one’s best self (Goffman,
1959; Scott and Lane, 2000), the problem arises of whether to treat these self-identity
narratives as fictional or non-fictional accounts of personal reality. A distinction that
might be made at this point is whether there exists a practical difference between fictional
and nonfictional narrative accounts. Given the relative newness of the field of narrative
7
identity study, this problem has received little attention in empirical study, but narrative
theorists predict that the difference between fictional and nonfictional narratives is
largely irrelevant to the persuasive impact of the narratives (Bruner, 1986; Green et al.
2004; Rubin, 1994). In practice, it may be impossible to cleanly distinguish individual
self-identities as either purely non-fiction or purely fiction, but by current accounts this is
not a consideration that will determine the relative persuasiveness of the narrative
identities themselves and therefore this study does not propose to directly address this
problem.
At present, identity theory proposes that human beings expect coherence from
each other’s narrative identities (Polkinghorne, 1998, as related in Henriksen,
Holmegaard, Ulriksen, and Madsen, 2015). This also appears consistent with predictions
of exemplification theory and the identified victim effect (Polkinghorne, 1998, as related
in Henriksen, Holmegaard, Ulriksen, and Madsen, 2015). We would expect this to imply
that increasing the number of identities would lead to less coherence and therefore lower
funding success. However, extant research apparently shows the opposite effect. Previous
study of for-profit lending found increasing the number of narrative identities predicted
funding success (Herzenstein et al., 2011). What remains to be tested is whether this
effect is robust across domains. Will this hold in the non-profit realm? Will the inclusion
of multiple identities in a non-profit narrative appeal be linearly related to success? If so,
will the direction of the relationship be positive or negative? If the relationship is
negative, it should lend support for the above mentioned three models, but if the
relationship is positive, what alternate explanation can be found for the effect?
8
It is possible that identities contain another characteristic that predicts persuasion.
Several researchers have isolated narrative valence as a predictive factor (Herzenstein,
Sonenshein, and Dholakia, 2011; Merchant, Ford, and Sargeant, 2010; Parry and
Kempster, 2013). The degree to which a narrative identity conveys, enhances, or
provokes so-called “positive” or “negative” emotions has been shown to impact
effectiveness (Merchant, Ford, and Sargeant, 2010). However, none of this research has
yet addressed the question of whether this effect plays a role in the online persuasion
arena. This is another potential issue that the present study seeks to address.
Many narrative theorists (see Reissman and Quinney, 2005) agree that exploring
only the most complete narratives including elements of character, plot and setting is
essential for “best” narrative research. However, can we be certain that only complete
narrations affect behavior? Or can fragmentary, implied, or archetypal micro-narrative
identities/characterizations also do so? Toward this end, the present two-phase study has
inductively-derived a group of narrative identifiers contained in the narrative texts of our
sample. In order to better tease out as completely as possible all identities in our posts, we
have identified what we believe are the smallest units of narrative that can predict
actions. We call these units of characterization we find in our sample micronarratives. As
components comprised of only characterization and, in at least several cases, implied or
presumed emplotment (vis-à-vis their apparent conceptualization as oppositional to
another character or force) but primarily without reference to contextual settings, use of
these micronarrative identities allows us to respond to the related research-provoking
9
question asked by Green, Garst, and Brock (2004): What are the boundaries of the power
of context-free narrative?
Statement of the Problem
Recent research indicates that specific narrative identities are related to funding
success in the fast-growing online for-profit crowdfunding phenomenon. Little is known,
however, about whether specific self-identity narratives have an effect upon the success
of online non-profit crowdfunding appeals, and if they do have an effect, which ones
have an effect, and what that effect will be. There is thus a gap in the knowledge about
the nature of the textual narrative identities that are most effective in securing nonprofit
crowdfunding online.
Purpose of the Study
The purpose of this two-phase qualitative and quantitative study is to explore a
random sample of funding requests retrieved from a major online crowdsourcing website,
Indiegogo (IGG), in order to explore the narrative elements used in the requests and the
relative successfulness of each. In study #1, I conduct in-depth inductive analyses to
identify the types of narratives nonprofit entrepreneurs are employing in their
crowdfunding requests, while in study #2 I use multivariate analyses to quantitatively
investigate the relationship between the use of micro-narratives and the success of
crowdfunding requests. I ran a series of regression analyses to determine precise
associations between these specific, inductively generated narrative identities and several
measures of funding success, including both metric and binary data. In order to achieve
the highest explanatory power possible, this study conceptualizes five distinct ratio-level
10
dependent variables (DVs): amount raised, percentage of goal reached, number of
funders, time to funding, and number of narrative identities. It also conceptualizes one
binary DV: goal fulfillment, which measures whether or not the requester received
complete funding of their stated fiscal goal.
The independent variables (IVs) and controls form another subset of data looked
at in the study. The IVs are inductively determined narrative identities, based in part upon
Herzenstein, Sonenshein, & Dholakia’s (2011) list of categories, derived from the trust
model of Mayer, Davis, and Schoorman (1995). The original list included narrative
identities of trustworthy, successful, hardworking, economic hardship, moral, and
religious. The present study presents an additional group of inductively generated
identities including needy, helpless, victim, child mentioned, child-centered, courageous,
grateful, hero, hip, and happy. The extent to which each identity contributes to a
fundraising campaign’s success or failure is one of the central questions this study seeks
to address.
Items controlled for include both non-structural and structural elements found on
the webpages. Non-structural items are those items that have no part in the narratives
themselves, but are instead merely items competing with narratives for conscious or
unconscious consideration by the funders. They include measures of campaign length in
seconds and numbers of, respectively, updates, comments, funders, gallery images, perk
levels, Tweets, and Facebook likes. Structural items are those elements of narratives that
are not identity claims precisely, but instead help describe other aspects of the narratives
that may influence success. They include measures of reading ease, grade level, auto
11
read index, average grade level, word count, sentence count, and errata—both metric
counts of total instances and binary determination of errata as present or not present.
Errata can include instances of grammatical, spelling, punctuation, syntactical, and
idiomatic errors as well as non-standard language usage.
Significance of the Study
This study is significant in that it attempts to address the gap in the knowledge
about which narrative identities contribute to a fundraising campaign’s success and
failure and how and why they might do so. Its aim is to increase the understanding of
sequential messaging compliance-gaining attempts in the non-profit online lending realm,
specifically the presentation of individuals’ narrative identities. This increased
understanding might serve to better inform theoretical models of giving, fundraising, and
website management, as well as help those individuals or organizations desiring funding
to make better decisions about how they present themselves in the online crowdsourcing
environment. Given the study’s contribution to advancing theory, the results may also
generalize to other fundraising venues as well. Government and large charitable
institutions perhaps could use the findings to improve their own fundraising policies and
practices.
Primary Research Questions
Therefore, I ask the research questions, “What aspects of narrative identity
presentation will relate to non-profit funding success?” and “What will be the direction of
each relationship?”
12
Research Design
This study uses a two-phase design in which a qualitative inductive analysis (i.e.
coding) is performed in Study 1, followed by a quantitative analysis in Study 2 – in this
case using two types of regressions – as outlined by Lee (1999). This design is
recommended as having great potential for enhancing research, though its weakness is
that some researchers may apply it in an unbalanced way if they have greater facility with
the methods used in one or the other of its phases. The two-phase design is less complex
than a formal mixed methodology design, which contains multiple qualitative and
quantitative methods (Lee, 1999). By combining multiple qualitative and quantitative
methods, a mixed methodology design, in contrast to a two-phase design, could increase
confidence in inferences drawn from findings (Lee, 1999); however, this was not a
primary concern for this exploratory study, so the two-phase method was chosen instead.
This study has one qualitative method (coding analysis) and one overarching quantitative
method, regression analyses. Although two types of regressions – ordinary least squares
regression and logistical regression – are herein used in multiple iterations, all are
examples of one overarching kind of analysis used at the same point in the research
process, so I consider them as one quantitative methodology overall, instead of multiple
kinds.
As laid out in Lieblich et al. (1998), Study 1 takes the approach of a categorical
analysis, abstracting words or sections of words as fitting into particular categories,
derived from a coding strategy. This coding strategy is subject to limitations including the
precept that narrative accounts are multiplicitous, as each narrator has many voices and
13
self-representations. Narrative analysis is not necessarily a fact-finding mission. It
doesn’t assume that people narrate objective facts about their lives. Rather, as proposed
by Donald Spence (1982), it assumes that truth in narrative is a reconstruction of
experience, not an historical record of actual facts. People construct meaning for and
about their lives making linkages between events or aspects of life and how they
understand and interpret these. Philosophers foundational to narrative analysis include
Ricoeur, Heidegger, Husserl, Dilthey, Wittgenstein, Bakhtim, Lyotard, MacIntyre, and
Gadamer (Josselson, 2004).
This dissertation also assumes the perspective of Mikhail Bakhtin (1981) that the
narrative self is dynamic and never fully finished, so that self-narratives will be, by their
nature, inconsistent, resisting closure, and comprised of many voices dialoguing. Such
selves are more than just making themes; they are also layering and connecting themes to
each other (Josselson, 2004). Josselson lists the following 4 stages of analysis: overall
reading for structure, multiple readings for identifying voices, developing a Gestalt for
coherence, and touching base with theoretical literature. Being aware of the researcher’s
own presuppositions is also required.
Parry and Kempster (2013) assert that their mixed methods approach is narrative,
rather than positivist, but is called narrative positivism, and more precisely aesthetic
narrative positivism. They cite evidence that due to the increased exploration of narrative
in the social sciences, boundaries between science and narrative are blurring. In their
method (as in mine), narrative is the data. Their “intention is for positivism to assist the
narrative rather than for narrative being of assistance to positivism.” (p. 6) My study does
14
not concern itself with the interiority of the subject (as would be expected from the life
story identity model of McAdams, 1993—informed by Erikson, 1963, and James’s “I”
and “Me” distinction, 1963), but rather with the exteriority or social claims made. In
Smith and Sparke’s (2008) terminology, my study is “thin individual” and “thick social
relational” in its focus or “spotlight,” since I am looking primarily for evidence
supporting social processing of individual claims, rather than the subjective germination
of such claims, presented as they are within a social sphere.
My coding process in Study 1 began in the manner of Boyatzis’s (1998) template
(codebook) approach, as outlined by Crabtree and Miller (1999). Codes replicated from
the Herzenstein et al. (2011) study were outlined a priori based on preliminary analysis of
text, rather than inductively from the actual sample, then other codes were added as they
were inductively derived from the actual sample.
Theoretical Framework
My underlying epistemological assumptions are grounded primarily in the anti-
positivist Symbolic Interactionism of Mead (1963) and the Chicago school of sociology,
with branches stretching into the communication field through Galileo theory (Woelfel,
2013; Woelfel, 1992; Woelfel & Fink, 1988). My approach to knowledge production
therefore reflects a healthy skepticism of pure positivism, because my approach does not
presuppose any absolute “scientific truth” is attainable. Rather, my approach presumes all
scientific fact, including any derived herefrom, is mutable, and subject to verification by
the community of other scientists, who will, presumably, change or add to it. Such
15
scientific fact also assumedly bears the marks of cultural bias, ethnocentrism, and
historical milieu.
My research approach cannot be claimed as purely narrative, as it does not meet
all the formal criteria for narrative research laid out by Riessman and Quinney (2005),
calling for in-depth and extensive, dialogical narrative collection, full exploration of
emplotment, and full consideration of context or setting. Given, however, that I attempt
to clearly identify my biases and philosophical approach, I believe I am approaching a
quasi-narrative approach, in which I attempt to include as many narrative elements as
practicable and applicable, rather than a pseudo-narrative approach in which I would
attempt to obfuscate my admittedly partial, perhaps even impoverished use of narrative
elements by claiming to do otherwise.
Assumptions, Limitations, and Scope (Delimitations)
The self-evident truths I assume for the purposes of this study are: first, I assume
that the random sample gathered is indeed representative of the online funding appeals
posted in the non-profit section of the Indiegogo.com website during the six month period
between April 1, 2012 and September 30, 2012. Second, I assume that the six measures
of funding success actually capture the effect. Third, I assume that relationships such as
might be found in our regression analysis do not predict causality. Fourth, I assume that,
should those potential relationships be found to be robust, further research – including
laboratory experimentation – would be necessary to establish causality.
One limitation of the study over which I have no control is the choice of
instrumentation for data-gathering. The only practical way of gathering the needed raw
16
data appeared to be Python language-driven databases. Any shortcomings of this method
of data gathering were thus beyond my control. Another limitation of this study over
which I have no control are my unconscious biases, which might have significantly
affected my coding scheme development and implementation. It is possible that my
cultural heritage as a Caucasian male of a certain age may have caused me to overlook
some possible alternate codes or explanations for effects I found. In addition, it is also
possible that, although attempting to be as precise as possible in my language choices, I
may have inadvertently over- or understated some relationships, claims, or observed
narrative phenomena. Any of these might be potential weaknesses of my study.
The scope of this study included a comprehensive dataset gathered over more than
a six-month period of all non-profit funding requests on a major crowdfunding website,
Indiegogo.com. The population was ultimately reduced to a six-month period; then a
random sample of 200 posts was selected from this population. This one-time sample of
200 posts served as the basis for all the measurements described in this study. Given the
random nature of my sample from the targeted population, I feel confident that the results
attain generalizability to other periods and places beyond that population gather from
which I took my sample. Our choice to deliberately delimit our population gather to a six-
month period was done for simplicity and to ensure that all initial posts of my final
sample of requests would be fully retrievable.
Definition of Terms
Several terms may cause confusion among even highly experienced researchers in
this topic area. Therefore, I will define the following terms. Narrative identity is that
17
process theorized to occur within every consciously aware human being wherein she
engages in mental deliberation about, identification of, and presentation of -- either to
herself, others, or both -- her nature, function and place within the mental historical
projection she creates of her own lifeline. It theoretically is the driving force for
establishing personal meaning and significance in one’s life. Inductive coding is the
process of reading text repeatedly, with the aim of distilling conceptual elements that can
be categorized into identifiable groupings or schemas. Positivism is a philosophy that
asserts that science is capable of positively identifying universal truth in the objective
world using a carefully applied scientific method, including observation,
experimentation, and theory testing. Crowdsourcing is the activity of mass mediated
efforts to accomplish goals held in common by online participants. Crowdfunding is a
type of crowdsourcing specifically geared toward raising money, particularly in
transactions between individual requesters and potential masses of funders.
Micronarratives are short, perhaps fragmentary story elements including traces of one or
more of the following: characterization, plot, and setting. Emplotment is the process of
including within a narrative any elements of plot development, particularly conflicts and
resolutions arising from incidents and actions taking place in the story. Context-free
signifies the absence of all indicators of setting or “place” where story action occurs.
Summary
In this chapter, I previewed the background of the research problem and identified
the gap in the literature that I will address in this study. I also outlined the purpose of the
study and its significance for the field of study, and proffered two formal research
18
questions that I shall explore in the study. I then briefly previewed the methodology I use
in the study and laid out my underlying philosophical framework, assumptions,
limitations, and scope and closed with a preview of definitions I use in the study. In
Chapter 2, I will next review the studies relevant to the present two-phase study, showing
that each fails to solve the research problem I have set out to guide this study. In this way
I will seek to establish the relevance of the present study for the discipline at large.
19
Chapter 2: Literature Review
From the days of Aristotle’s Rhetoric and the works of Cicero to modern courts of
law, scholars, specialists and laypersons have all faced the need to explore and
understand persuasion. Contemporary empirical persuasion research includes compliance
gaining – the study of strategies used to induce others to perform actions, rather than
change their opinions, beliefs, or emotions. While the actions requested are often in the
nature of purchase requests by for-profit entities, there is also a sizable literature
addressing their use in nonprofit fundraising efforts.
Narrative as a Sequential Compliance-gaining Technique
But what techniques influence compliance? To date, much of the literature has
focused on power (French & Raven, 1960), compliance tactics and strategies (Marwell
and Schmitt, 1967), and sequential effects (Cialdini, 1984). Less well developed,
however, is the study of narratives: sequentially organized (Reissman, 1993; Herzenstein,
Sonenstein, and Dholakia, 2011) persuasive words, phrases, and ideas that create a
coherent logic, identity, or story. Thus, while some researchers have theorized that Homo
sapiens is a story-telling animal (Gottschall, 2012; Gottschall and Wilson, 2005),
surprisingly, in the compliance-gaining literature, the existing research on the effects of
narrative is rather limited.
Donations, Non-profits, Online Giving, SNSs, P2P Lending and Crowdfunding
Civilized society is dependent on donations (Zhuang & Saxton, 2014). Of high
interest to individuals, scientists, organizations, and governments, therefore, is
discovering which techniques garner more donations and why. Donations are especially
20
interesting as a measure of compliance gaining. This is particularly true in the arena of
non-profit online funding requests (e.g. Guo & Saxton, 2013; Hackler & Saxton, 2007;
Nah & Saxton, 2012; Saxton, 2005; Saxton & Benson, 2005; Saxton & Guo, 2011;
Saxton, Guo, and Brown, 2007; Saxton, Neely, & Guo, 2014; Saxton, Oh, & Kishore,
2013; Saxton & Wang, 2013). Contemporary research on compliance gaining and
narrative has focused on print media, while electronic media has been less studied. This
study therefore extends the research somewhat by analyzing online requester narratives, a
sequential form of influence on compliance (Herzenstein et al., 2011; Reissman, 1993;
Marten, Jennings, and Jennings, 2007), to discover what narratives predict donor
behavior, and see how this may be accounted for by current narrative theory.
Online giving is an important and growing industry. According to the 2013
Blackbaud Index, online giving accounted for 6.4% of all charitable giving, excluding
grants, in 2013. The Index also reports that overall giving rose 4.0% in August 2014
compared with the same month in 2013, but that online giving rose 12.7% in August
2014 compared to August 2013. Thus, 2014 was the second year in a row in which online
giving had increased in double digits over the previous year. The growth of funding
online for non-profit organizations is ongoing and shows no sign of abating soon
(Blackbaud, 2014).
Traditionally, large nonprofit entities have become the repositories of much
charitable giving. Currently, however, there is an online crowdsourcing fundraising
revolution occurring in which cumbersome corporate-like charitable entities are
competing head-to-head with massive numbers of individuals directly asking for help,
21
peer-to-peer (P2P), from their financially able individual peers (Satorius and Pollard,
2010). This new world of online crowdfunding has given individuals an alternative
funding resource that appears to offer advantages of speed, simplicity, and greater
personal involvement and satisfaction for all involved. Dholakia and Firat (2006) mention
that this diffused marketing signified by the postmodern progress away from centralized
organizational “management” toward the participatory and co-creative elimination of the
marketing “membrane” is somewhat like a neural network constantly working to
“re(de)construct” itself (p. 151).
Given the ubiquity of online interactivity, person-to-person (P2P) online social
networking websites have interested researchers for at least the past ten years (Trammell,
Williams, Postelnicu, and Landreville, 2006). Research has been conducted on P2P
interaction conducted on social networking sites and political candidate websites and
blogs, for instance (Trammell et al., 2006). Moreover, as interactive crowdsourcing sites
have expanded, they have become increasingly important arenas to find and measure the
outcomes of real-world compliance-gaining attempts.
Chaffee and Rapp (2012) describe P2P lending conducted through websites that
offer individual investors the chance to lend funds to individual borrowers and promise
lower borrower interest rates and higher investor return rates. Such websites have
generated significant media attention as well as regulatory concerns at both the state and
federal levels. Recently, narrative effectiveness in for-profit funding requests has begun
to be addressed in entrepreneurial research (Boje, 1991; Martens, Jennings, and Jennings,
2007; Navis and Glynn, 2011), network research (Bennett, Foot, and Xenos, 2011), and
22
organizational research (Herzenstein et al., 2011). Not yet addressed, however, is
narrative effectiveness in not-for-profit donation requests. This study will therefore
explore the impact of narrative in the online environment of not-for-profit, P2P funding
requests.
Narrative
In their seminal work, Gottschall and Wilson (2005) argue that humans by their
nature are storytellers. Whether this claim is justified or not, considerable research has
explored the relevance of storytelling in business. Stories, or “narratives” have recently
been the focus of research in entrepreneurialism, for example (e.g. Martens et al., 2007;
Herzenstein et al., 2011; Navis and Glynn, 2011). To explain how narratives lead to
compliance-gaining success in entrepreneurial efforts, Martens et al., (2007) argue that
narratives work in three ways to help capital acquisition. They
1. help comprehension by simplifying explanation of the offering into a coherent
unity.
2. explain reasoning behind offering so investors see its nature and value.
3. generate commitment and interest when they connect “to broader contextual
narratives” and make distinctive, original offerings appear sound.
In these ways, storytelling reduces uncertainty, motivates involvement, and mobilizes
capital. Martens et al. (2007) present their seminal study as “the first systematic, large-
sample test of the overarching claim that effective storytelling can facilitate external
resource acquisition.” They develop and test a theory of how narrative content affects
lending behavior in entrepreneur development requests. They explore first the notion of
23
whether stories help resource providers and, if so, in what ways. They postulate that
simplification of packaging and elaboration of reasoning are two such ways and they
propose that connecting to broader contextual narratives from the relevant field,
marketplace, or pool of competitors is a third way that narratives increase interest and
participation. So far, these postulates have not been formally tested for their
generalizability to the non-profit online donation context.
Martens et al. (2007) mention that the narrative form allows the narrators to
construct a sort of play in which they as lead characters describe and understand
themselves. Stories relay facts, but in a memorable way that unifies disconnected parts
into a unified whole. Moreover, stories relay these facts in a sequenced way using
historical perspective to increase the certainty of the proposed progression. Martens and
colleagues further propose that this process reduces uncertainty and leads to greater
facilitation of resource acquisition. We should expect this role to be less important in
non-profit than for-profit, but our study is not designed to compare the two roles thus
here. Martens and colleagues also see a major role of narrative to be reducing information
asymmetry, since they share information with the inquisitors. In Study 1 of this
dissertation, I first explore whether narratives do, indeed, include self-identity elements.
Then I categorize those identities. I will follow the convention established by the seminal
study of Green and Brock (2000) in describing these instantiations of self-identity as
public narratives, specifically external public narratives, to which more than one person
at a time is exposed.
24
Martens et al. (2007) also theorize that complexity will reduce the effectiveness of
narratives, a position consistent with both the heuristic-systematic model (Chaiken, 1980)
and the elaboration likelihood model (Wegener and Petty, 2001). Both of these models
predict people’s attention will remain focused on the main features of the messages, when
they have high capacity and motive for processing. Committing personal or corporate
funds in the face of risk of loss would seem a time when motive for processing would be
high. They found a diminishing return on complexity of messaging. This extends in a
way to the number of narrative identities themselves used. Based on the social cognition
research of Brewer, Weber, and Carini (1995), Martens and colleagues predicted tapping
too “many categories of social identification” would likely lead to “recognition
confusion.” When this is combined with low levels of familiarity with the social
categories, adding more identities leads to diminishing returns. Finally, they raised a call
for future mixed methodology studies in order to demonstrate that storytelling is more
than rhetoric and is effective in producing outcomes. However, their study did not
examine this effect in the non-profit arena to see if the same results will still hold. The
present two-phase study attempts to partially address this call for mixed methodology
study by determining if narrative identity self-presentation can predict funding success.
According to Herzenstein et al. (2011), narrative gives meaning to events
mentioned by the narrator. This autobiographical window also can be used to reveal the
authors’ self concept and identity, at the discretion of the authors, giving the authors the
opportunity to manage their image or identity. This is analogous to the process of
strategic impression management (Leary and Kowalski, 1990), wherein presenters choose
25
to portray identities most likely to gain support. Narrative is also the best method for
showing character development, according to the seminal work of Green and Brock
(2000). Isolating the character development aspect of narrative, as Green and Brock have
done, lends power to my conceptualization of micronarratives, developed later in this
chapter.
Navis and Glynn (2011) built upon the findings of Martens et al (2007) and
codified them in more detail. They explored how entrepreneurs are marked by newness, a
liability until legitimacy is acquired by endorsements, which require some conformity.
This results in a tension between legitimacy and distinctiveness. They proposed that
identities of entrepreneurs are favorably judged when their distinctiveness is legitimate.
They also predicted that these judgments are context-dependent and mediated by
narratives of identity which give ‘institutional primes’ and ‘equivocal cues’ when
investors try to make sense of offerings. Thus, an organizational identity is part of an
organizational “theory of being.” Their model to date has only been tested in
entrepreneurial for-profit settings.
Identity Claims, Trust, and Need
According to Herzenstein et al. (2011), identity claims are an important aspect of
narrative self-presentation used in public discourse. Such claims help establish the “who
we are” and “what we do” facts about the narrating entity. They might include, for
example, short verbal descriptions of the requester as needy, community service-oriented,
hardworking, or trustworthy. Herzenstein and colleagues found that narrative identities
describing trustworthiness appear to account for a significant amount of success in for-
26
profit lending situations. Specifically, they coded narratives containing descriptions of the
requester that fit one or more of the following trust-based categories: trustworthy,
successful, hard-working, economic hardship, moral, and religious. These are factors
predicted by the trust model based in the assumption that for-profit investors seek to
minimize risk as part of their investment strategy, so they are especially influenced by
indicators that they are more likely to be repaid, such as trustworthy, moral, and hard-
working external public self-identities. In the Herzenstein et al. (2011) study, trustworthy
and hardworking narratives significantly and positively affected funding, while religious
marginally affected it negatively. No other factors impacted funding success. Loan rates
were reduced also by trustworthy and hardworking, but by none of the other factors.
However, a different cognitive process in non-profit giving is presumed in this
dissertation, based in non-profit management literature predicting that, because
repayment is not a consideration and because empathy, altruism, and belief in the cause
are more relevant cognitive factors for donor deliberation, need, rather than trust, will be
the more important consideration for donors (Bekkers and Wiepking, 2011). The
prediction of the present two-phase study is that trust will not be a relevant concern for
donors.
In lending situations, trustworthiness is usually assumed to be of great importance
to lenders. According to Mayer, Davis, and Schoorman (1995), trust is the willingness of
one party to extend resources to another while at the same time assuming risk in the
transaction. Risk is thus a primary element distinguishing trust from such constructs as
confidence, for instance. Another aspect of lender decision-making is the effect of
27
narratives on lender deliberation. Herzenstein et al. (2011) find that identity claims can
affect the willingness of lenders to commit resources to borrowers, in part because of
what they communicate about borrower integrity, ability, and benevolence, a prediction
these researchers borrow from the trust model of Mayer, Davis, and Schoorman (1995).
Herzenstein and colleagues found support for this trust model by analyzing for-profit
narratives on the Prosper.com P2P lending website, but their study does not attempt to
determine whether trust factors also predict donations in the non-profit realm, as might be
predicted if we assume donations include elements of risk to the giver. The current study
explores this specific question to see if any evidence can be found linking their for-profit
model to the non-profit realm or if, conversely, the two realms are different in this
respect.
According to the seminal study by Sargeant and Woodliffe (2007), no empirical
studies have been conducted exploring the importance of variables other than trust in
building loyalty or commitment in nonprofit donations. They also observed that risk of
the organization losing out if they stop giving appears to prevent some people from
giving in the first place. This risk of harm to the organization is in direct contrast to
concerns about risk to themselves. They also found that trust does predict commitment to
non-profit entities. Appel & Mara (2013), moreover, found perceived trustworthiness of
characters in narratives affects persuasiveness. In their seminal study, Sargeant, Ford, and
West (2006) wrote the first marketing model based on empirical measurement of giver
perceptions and their impact on donations. The literature has previously measured the
effects of intrinsic and extrinsic factors as a way to distinguish givers and non-givers, but
28
these factors don’t help explain levels of giving as much as perceptual determinants (such
as trust) can. Trust and commitment are also explored by Sargeant, Ford, and West
(2006) as mediators, with trust (and commitment, indirectly) appearing to relate only to
the perceptions of donors about whether the beneficiaries will benefit, rather than to any
trust that potential benefits will accrue to donors themselves. While these three studies
indicate that trust may be a concern in non-profit donation behavior, none empirically
addresses the question of whether need is also a concern in the non-profit arena. I propose
to do that in the present study. I also propose to address the lack of empirical exploration
of other narratives besides trust for their impact on fundraising success, by identifying
other narratives and measuring their impact.
Multiple Identity Claims
Exemplification theory (Zillman and Brosius, 2000) and the identified victim
effect (Small, Lowenstein, and Slovic, 2007) both postulate that brief stories about single
individuals are more persuasive than those referring to multiple individuals. What is
uncertain, however, is whether these results will obtain when multiple identities, rather
than multiple individuals, are referred to in a persuasive narrative appeal. Herzenstein et
al. (2011) found that increasing the number of identity claims appeared to reduce
uncertainty and increase trust, but also predicted a lower tendency for repayment. They
theorized there might be an implicit tendency for narrative requesters to increase number
of identities mentioned when they were less desirable borrowers and the choice to do so
will also predict loan default. They did not, however, explore whether this same
prediction holds in the non-profit field, where we suspect need, rather than trust, is the
29
driving force in decision-making. This study will consequently explore the relationship
between the number of narrative identities used in self-presentation and non-profit
fundraising success.
Time to Funding
Recently, Anderson and Saxton (2016, in press) operationalized the success of
online nonprofit fundraising as time to funding, and found that images of children,
husbands, and items of prosperity significantly impacted the speed of online funding of
individual loan requests on Kiva.com. The present two-phase study seeks to explore the
relevance of their explanatory model of online non-profit loan requests to online
individual donation requests as well. To do so, the present two-phase study measured the
time to funding in seconds and looked for an impact of presented narrative identities upon
speed to funding.
Micronarratives
The use of a narrative approach to focus on characterizations to the exclusion of
either setting or plot is problematical, according to Riessman and Quinney (2005). They
critique narrative research that does not use narrative elements fully and richly as not
“good enough” narrative study, perhaps not even truly narrative research. Martens et al.
(2007) also propose a list of essential narrative elements including subject, goal, forces,
and implicit or explicit event sequencing which create plot. However, dogmatic
prescription that narrative research must necessarily include all of its potential elements
in order to be well-done is inconsistent with Georgakopoulou’s (2006, p. 6) model
30
supporting “late modern theorizing of the micro-, or small, unofficial, fragmented and/or
non-hegemonic social practices as crucial sites of activity.”
In this dissertation, therefore, based as it is on a replication and extension of the
identity coding categories inductively derived by Herzenstein et al. (2011), I argue that
this pure approach to narrative research is too constrictive and may even be
counterproductive. Such pure narrative research, so defined, would not allow one to
analyze the smallest units of narrative that can predict compliance behavior, such as that
indicated by Herzenstein and colleagues’ significant results. The instances of self-identity
inductively coded by Herzenstein and colleagues have utility in spite of not containing all
three elements of narrative (plot, setting, and character). In fact, as Herzenstein and
colleagues demonstrated, the use of several of them correlated significantly with funding
success indicators. In my model, I will present these identities found by Herzenstein and
colleagues, along with several of my own inductively derived from my sample, as
narrative fragments we will call micronarratives. This is a term we are coining here for
the first time as a means of recognizing, distinguishing, and legitimizing identity forms
which we believe are narrative in nature or content, though not themselves formal or full
narratives. They are, by nature, narration, although not complete narratives. They may
tell merely an aspect of a story, rather than full sections, parts, or components, or an
interaction of those, as is usually stipulated for a narrative. Thus, a component of
characterization, or a component of a journey without characterization, are still narration
in style, content, and method, but less of complete narratives. This appears consistent
with Ibarra and Barbulescu’s (2010) model of identity management as well. As I describe
31
them, micronarratives may contain more or less coherence, given their lack of detail, but
they may also contain the potential for multiple combinations of or interactions among
narrative components such as plot development, setting, and agency/characterization. In
order to offer more support for this choice of terminology, I offer the following
justification, based in part in the narrative literature itself.
Why Measure Micronarratives?
When Ezzy’s seminal study (1998) first proposed that scholars integrate Mead’s
(1934) Social Interactionist approach with the hermeneutic narrative approach of Ricoeur
(1984), one of his intentions was to correct the tendency of some scholars to see identity
as either a substance or an illusion. He recommended instead seeing identity as a process,
using the relativity theory of Einstein (1920) to elaborate Mead’s statements about the
temporality of identity, and explained how, “In the same way that, according to the
theory of relativity, the measurements of an object depend upon the spatio-temporal
framework from which it is observed, the meaning a ‘minded organism’ gives to a
particular event depends on the spatio-temporal-social-interpretive framework from
which it is interpreted (Ezzy, 1998, p. 241).”
Similarly, I argue, micronarratives offer a chance to measure constructs in a more
discrete manner than do fuller narratives. The degree of narrativity contained in an
utterance or narrative unit can be determined individually, based on its spatio-temporal
framework, rather than by an overt reference to all three narrative elements at once.
Narrations with fewer than all of the above story elements might be usefully termed
32
micronarratives. They may stand in place of character development or be merely
character description or they may merely imply one or more story elements.
I am not the first researcher to study the discrete narrative elements that we here
call micronarratives. Herzenstein et al. (2011) do so in the study I attempt to replicate and
extend here, though without addressing the philosophical concerns mentioned by
Riessman and Quinney (2005). Czarniawska (1998: 17) also identifies “minimal
narratives,” and Bennett, Foot, and Xenos (2011) assert the value of frame analysis for
simplifying more complex discourse and enabling researchers to “pick up narrative
fragments that may reside outside of fully formed stories (p. 223-4).” They note, too, that
narratives “often become simplified, sloganized, and cued by images that can reside
outside of people (p. 223-4).”
Boje (1991) also mentions, “A finding in sociolinguistic studies is that stories are
brief and fragmented across extended and interrupted discourse, but this has been ignored
in organization stories (p. 109).” This can occur in the same way that “researchers can
‘unpack’ very brief enactments in dialogue to discover the reality underlying the
linguistic enactments (Boje, 1991, 110).”
Boje (1991) also describes “terse storytelling” as the most abbreviated a story can
be and still be a story. “It can be so brief that the performance is barely distinguishable
from other nonperformance utterances (p. 115).” It is described as using an “abbreviated
code.” At times, the terseness can be so great that there appears no story at all to the
uninitiated. Boje describes cases in which single words can represent entire stories with
just a prolonged, significant pause following them. Boje emphasizes that, “First, it is not
33
the fact that the story is terse and abbreviated that counts; it is the fact that the teller picks
one aspect to abbreviate (‘You know the rest of the story,’) and another to accentuate (p.
124).”
Georgakopoulou (2006) refers to small stories. Micronarratives as I have
conceptualized them are probably at times smaller than small stories, though they may
perhaps be situated within “small stories” and include the category of “identities,” though
they are not limited to this category.
Micronarratives are more specific than “repertoire” elements, in the
conceptualization of the seminal Ibarra and Barbulescu (2010) model. The category of
repertoire elements is more inclusive and less specific than micronarratives as it can hold
unstructured story fragments, such as micronarratives, along with delivery aspects as
well.
Ezzy (1998) also adds more useful perspective on our micronarrative
conceptualization with his description of Saas’s (1988) hermeneutics of suspicion, which
“argues that self-knowledge is always only ‘approximate, tentative and indirect’ and
often influenced by ‘only partially specifiable themes and backgrounds that exist at
various levels of implicit and explicit awareness.’” (p. 249) Micronarratives thus can be
any one or more of the essential components (plot—temporal sequence, character,
setting, etc.) of an actual or potentially larger story and can include any part of,
implication of, or derived meaning from actual or implied full-length stories, enacted or
still-potential.
Given these precedents, I propose that the term micronarrative is not only an
34
acceptable, but also a useful way to describe story elements that serve a narrative
purpose, even if the story is only implied or inferential. Plot sequence may not be made
explicit, and there may be no other elements, such as conflict, but character is crystalized;
other story elements can sometimes be inferred, even when not direct or implied.
In practice, it appears that the findings of both Herzenstein et al.’s (2011) study
and the present one (as I will show in my chapters 4 and 5) show some reasonable
connectedness between these micronarratives in real life settings and their impact on
social outcomes. Even though they are fragmentary and small, these micronarratives do
indeed tell at least part of a story, it appears. While Ezzy (1998) argues that temporality is
integral to narrative identity, he does not mention the degree of temporality that must
accrue to these accounts. I argue, therefore, that not only may the degree of temporality
be flexible, it may actually be implicit, understood, or evidentially absent as well, as long
as it is present by proxy in some form. This temporality or plot development may be as
simple as, for instance, socially-constructed, conflict-derived personality
characterizations such as those found in cultural archetypes or in binary/oppositional
traits, moods, or chemical reactions. Narrative can thus include limited
conceptualizations; the problem for the present two-phase research study is: how small or
fragmentary can those concepts be and still be narratives? I suspect they may be as small
as the smallest unit of language, provided they include sufficient identity elements,
however invisible. They may therefore be fully narration, yet not be full-length
narratives.
35
Elliott and Wattanasuwan’s (1998) discussion of identity lends power to the
present micronarrative model as well. Their exploration of the importance of deep
meaning, personal meaning, and trust in brand effectiveness can be used to demonstrate
that even as small a symbolic element as a brand name or logo still functions in a
narrative capacity to tell part or all of a story.
I have argued that these characterizations we call micronarratives may exhibit a
sort of implied emplotedness, in that their meanings presuppose actions by the character
in the context of time. They are thus at times the recognizable symbol of an implied
conflict history or plot. The meaning projected by a micronarrative is resolved into a
unity that the reader perceives as a character, but which is also, as the Classical Greek
origins of that word kharakter suggest, a distinctive mark—such as a facial dimple --
which in the Einsteinian sense of Reimannian space, may be plotted even in its negative
space, the part not seen on the surface, but hidden in the warped folds of the experiential
symbol itself.
Our disinclination to explore the contextual factors of setting (along with our
decision to explore emplotment only as the implied conflict measure) can also be seen as
consistent with the narrative research model proposed by Green, Garst, and Brock (2004).
Their model predicts that even merely implied narrative elements can affect attitude
change. They allow for separating out of discrete contextual factors by measuring
context-free assertions in the search for persuasion effectiveness or attitude change and
call for further exploration of the boundaries of the influence of context-free narrative
36
implication. This allowance parallels how I frame the present research, as an analysis of
the effects of implied content, over and above overt content.
Archetypes
Caldwell (2010) explores the identifiability of archetypes specifically derived
from consumer models, rather than derived from more general human theories of
motivation such as proposed by Jung (1975) and Mark and Pearson (2001). Ibarra and
Barbulescu (2010) also bring up the relevance of heroic/mythic conceptualizations in
their framework. In order to assess the relevance of archetypal identities on donor
behavior, I searched our sample for archetypes, using the model of Mark and Pearson
(2001) as our guide. My inductive analysis only pulled out one readily apparent
mythic/archetypal theme, hero, which perhaps says more about the lack of their overt use
than it does any significant success or failure of the Mark and Pearson (2001) model
itself. Perhaps other archetypes could in future be identified by implication, rather than
predominantly using specific word cues, as we largely did to identify hero. Also, merely
identifying whether a term or concept has been included in a post gives little indication of
how it is used. This is a concern of seminal researchers such as Hibbert and Horne (1996)
as well. For instance, just because one mythic character is identified, does that mean it
has been presented in a meaningful way? The limited occurrence of easily identifiable
archetypes in my sample prevents me from analyzing them as fully as I would like.
Relationship Between the Independent and Dependent Variables
The goal of the present two-phase study is to explore the efficacy of real-world
persuasion attempts, in particular persuasiveness in the online prosocial donation arena.
37
The persuasion attempts looked at are high stakes, real-life requests for funding from
both individuals and companies, some made with high-dollar resources, some made with
nearly none. These are not artificially set up laboratory experiments with little
applicability to real-life persuasion, but ongoing persuasive campaigns for action.
Specifically, this study explores the persuasive effects of narrative identities, as
evidenced in textual funding requests in the online environment of non-profit donations.
Its goal is to determine whether there are any relationships between specific independent
variables (which revolve around narrative elements in text) and specific dependent
variables (which measure funding success in several ways).
The compliance-gaining dependent variable (DV) is broadly defined here as the
successful attempt to obtain a commitment of funding from another person, group or
organization. This is in keeping with the seminal study by Vaara (2002) calling for more
study of the phenomena of success and failure in narrative studies, which has been
understudied in the social science literature – an especially important issue, given that
such narratives are an important source of empirical data for scholars. More narrowly,
the most relevant success measures were determined to be: the continuous measure of
overall amount funded, the binary measure of whether or not the requester’s goal was
reached, the bounded measure of percentage of goal reached, and the continuous
measures of time to funding and number of funders.
Goals
The present two-phase study seeks to expand the narrative identity concept
proffered by Herzenstein et al. (2011) to the nonprofit realm using micronarrative
38
identities to see which narrative identities relate to higher funding, full funding, and a
higher percentage of the funding goal met. It also adds to the list of narrative identities
explored in previous literature by expanding Herzenstein and colleagues’ identity claim
categories to include some additional identities inductively generated from its unique
dataset. For example, preliminary research and data coding for the current study brought
to light additional narratives such as “jobs,” “needs,” and “I’m the best fit,” although
final coding eliminated or consolidated some of these categories.
This two-phase study also attempts to address the relevance of trust factors in
non-profit donations in the following way. It replicates Herzenstein et al.’s (2011) use of
Mayer et al.’s (1995) three theoretical borrower characteristics of integrity, ability, and
benevolence to structure its coding categories for trust-risk identity claims. Overall, this
study attempts to determine whether the trust factor appears relevant to the non-profit
fundraising world, or whether factors implying need predict success better.
Based on the empirical precedents and theoretical models in the preceding studies,
the two studies presented here explore the aforementioned narrative identities and several
previously undeveloped ones to shed additional light on: 1) what stories people are telling
about hardship, personal disaster, and personal responsibility; 2) which are more
successful; and 3) how the results fit with predictions from current theory. The
contribution to the communication literature that the study offers is to identify differences
between compliance-gaining results in nonprofit donations and in for-profit lending
within a specific theoretical framework.
39
This study overcomes previous limitations of compliance-gaining research by
exploring the real-time interactions among actual requesters and donors in the ubiquitous
posts requesting online non-profit donations, an innovation based on calls for expanding
the generalizability of entrepreneurial narrative research findings to other financial
contexts (Martens et al., 2007). Using automated data collection, social media resources,
and statistical analysis methods, this study design also innovates by measuring actual
online persuasion processes as they occur.
Research Questions
Given that related findings on the effects of narrative choice in self-presentation
indicate that the story the words tell will significantly impact the perception of
trustworthiness, it can be predicted that certain narrative frameworks or “stories” will
lead to more complete funding than others. Trust has been identified as an important
indicator of success in for-profit funding requests, given the risk involved for investors. It
would also seem that, even though risk may not be as important a consideration for
nonprofit donors, it will still be somewhat of a concern for non-profit lenders, albeit
probably less so. To date, narrative research has not explored the relevance of needy
identities. If, as the for-profit model from the financial research suggests, lenders are
motivated differently from donors, we should see that need considerations will more
strongly account for funding in contexts that are less driven by trust considerations since
payback is not at stake. Therefore, we ask the research questions: “What narrative self-
identities are used by request posters in the online nonprofit fundraising?” and “What
40
aspects of narrative will be related to non-profit funding success?” and “What will be the
direction of such relationships?”
41
Chapter 3: Research Method
Sample Selection Process
Initially, the choice of potential websites suitable for gathering sufficient data on
nonprofit crowdfunding success in real time was narrowed down between
Gofundme.com (GFM) and Indiegogo.com (IGG). Both are highly popular online
crowdfunding websites that offer the general public opportunities to donate money to
either persons or organizations that can be either profit seeking or not-for-profit. A spot
check of GFM showed 40 pages of nonprofits on January 23, 2013 at 9:23AM EST.
There were ten posts per page. A brief check of every other page or so found three videos
total. None of the pages used the standard video arrow on their thumbnails.
At the time of the preliminary gather on IGG, that site also showed 40 pages of
“verified nonprofits,” but with 9 entries per page. The first five categories of nonprofits
were: animals, community, education, environment, and health. The overwhelming
majority had embedded videos. Based on this preliminary count, there appeared to be
substantially more videos on IGG than on GFM, so to maximize data collection efforts
(some of which data is planned for use in future analyses), the decision was made to use
IGG instead of GFM for this study. Complete data collection began February 22, 2013
and ended October 10, 2013, gathering all that were labeled Verified Non-Profit (VNP).
Thus, the URL added dates in the final data were February 22, 2013 through October 10,
2013. (February to October was nearly an eight-month period.) That added up to 2764
cases, which, when all duplicates were removed, yielded 2750 non-duplicates (unique
campaigns). During analysis, cases were identified by their URL and gathered date.
42
Cases gathered early had the longest data insertion lag. Any that did not have key data –
Funding Total, End Date, etc., were deleted.
In preparation for Study 1, this researcher started looking at the data set in
February and gathered any 2012 campaigns that were still running. However, when ready
to analyze the pages of these preexisting campaigns, the researcher would have had
search for them online by URL, since the pages themselves would have been posted
before our collection began. It was determined that if IGG was not reliable in keeping old
campaigns, we might not have access to all those from that year. To check this, the lead
researcher emailed IGG, who replied that they do not, as a rule, delete old campaigns
from their server history. However, the strategy was chosen to delete any that started
before Feb 22, 2013, just as a safeguard against such criticism. (That eliminated about
25% of the original data.)
The data gather was completed on Oct 14th
. This end gather date decision was
somewhat arbitrary, based primarily on convenience, so we decided to narrow the
population into an even 6 months gather, by using only the campaigns that both started
and finished within the 6 month period from April 1 to September 30, which fully
comprises the 2nd
and 3rd
quarters of 2013. The usable population was narrowed to
certified nonprofit postings on IGG during the complete six months from April 1, 2013
through September 30, 2013. Thus, the final gather started in April on any campaign that
started and ended in this period. (About 46% started before that.) This yielded 1247
(actual final count) campaigns from which a random sample of 200 cases was taken.
43
Collection Method
First, preliminary data was hand-collected in near-daily grabs of online data using
Python script to gather data from webpages of donation requesters on the IGG website.
Originally, IGG showed several advantages over the competing website, GFM, in that the
former hosts significantly more nonprofit requests and accompanying videos. Also, GFM
doesn’t directly host all its videos and offers little “hint” that a video is available. When
called, the workers at the GFM helpdesk said to check YouTube for their videos, since all
videos must be uploaded there to work on their site. But they offered no way to easily
search for those videos. Conversely, since IGG hosts videos using Vimeo, Python script
could be used to mechanically gather all the videos.
All elements of borrower post webpages were downloaded, including text,
pictures, symbols, and videos. Later, long-term collection efforts were mechanized using
electronic databases. Snapshots of webpages from the IGG crowdfunding website were
collected, as were all separable components from each page, so that data could be
analyzed according to the dependent variables indicating funding success: amount
funded, whether or not funding goals were reached, and percentage of goal reached. Also
included were data on controls, including those indicating attention gained: likes, hits,
etc. Data on the independent narrative variables were also collected.
Coding
The present study replicates the conditions of Herzenstein et al.’s (2011) study of
narrative influence in the online borrowing and lending environment in order to extend
their findings about narrative effectiveness from the for-profit sector (as found on
44
Prosper.com) to the not-for-profit donation request environment found on IGG. The
“identity claims” originally presented by Herzenstein and others were: trustworthy,
successful, hardworking, economic hardship, moral, and religious. Herzenstein and
colleagues inductively determined those narrative identities after reading about a third of
their gathered narratives. Then, by analyzing the same narratives, two research assistants
verified the list was exhaustive. Ten more research assistants coded all the data in the set,
assigning a “1” to any narrative containing any version of each specific identity.
Otherwise, the narrative was coded with a “0” for each absent identity. Finally two RAs
read the listings, first independently, then with discussion to come up with a unified code.
The present Study 1 used a similar inductive process to determine additional
identities. The researcher started with a preliminary textual analysis of 100 targeted posts
on IGG and categorized each instance of identity presentation. The list for this inductive
search start with those specified by Herzenstein et al. (2011): trustworthy, successful,
hardworking, economic hardship, moral, and religious. Preliminary analysis by the
present author resulted in a few additional micro-narrative identities, including need, fun,
helpless, and victim. This addition comprises a unique contribution to the literature of
non-profit P2P lending analysis. (See Table 1 for actual examples of each narrative
identity.)
Given that the original Herzenstein et al. (2011) study did not offer an extensive
description of each item in its coding system, Study 1 attempted an approximate
replication of the coding itself in the following manner. Any form of the word trust was
taken to infer an identity based on trustworthiness. Terms such as “innocent” were
45
determined to be analogous to trustworthy and so coded. Any form of the word success
was a signal of an identity based on successfulness. Any use of the exact term “working
hard” or “hard work,” or an extensive description of laborious effort was assigned a value
of hardworking. Terms such as “less fortunate” and “financial difficulty” described
aspects of economic hardship. Moral narratives include references to doing good, helping
others, or community contributions. Any reference to religions, doctrines, or deities was
termed a religious identity. Need identities contained any form of the word “need.” Fun
narratives included “enjoy” words, “adventure” words, and any other words that
expressed subjective states of apparent hormonal pleasure (e.g. “enrichment of the soul”).
“Love” and “fondness” elements were coded in the fun narrative identity category, since
they can be understood to describe pleasant feelings. Terms such as “homelessness” or
“overwhelming odds” connote helpless. Descriptions such as “fleeing genocide” were
coded as victim identities, conversely. This means that several of the identities such as
need narratives are being scored according to specific word use, rather than overall
meaning, which may present other semantic problems. Overall, however, this seemed the
most relevant way to code some categories of narratives.
Other identities were also found to exist. For instance, posts on IGG are grouped
according to activities and interests such as childcare, healthcare, pet care, teaching, or
environmental. Given our earlier research into the effectiveness or nonprofit requests on
Kiva.org (Anderson & Saxton, 2016), which showed the relevance of babies in online
fundraising appeals, we coded for child-oriented identities here as well.
In addition to types of identities, the Herzenstein et al. (2011) study also found the
46
number of identities used correlated with an increase in the amount loaned, but also with
reduced repayment. Therefore, the present study also counted how many identities were
used in each posting. The number of identities was then added to the list of IVs.
Structural Variables
In addition to narrative identities, it is possible that other textual factors affecting
funding can be measured. In this vein, for Study 2 I created a listing of additional factors
we titled structural variables that included various measures of readability, length, and
errata. Readability measures include reading ease and grade level; length measures
included word and sentence lengths; errata measures syntax, grammar, idiom, spelling
and obvious punctuation errors. In cases where grammar rules are more vague or “best
judgment” is called for in online grammar sites, I coded for consistency, rather than
absolute “correct” or “incorrect” values. Thus, for instance, personal titles coming after
the name should not be capitalized unless all in the post are (thus, an apparent stylistic or
emphatic choice) and a Perk element once titled should be titled throughout the post,
particularly “Thank You” gifts. In general, improper or inconsistent hyphenation was not
counted as an error. “Ands” beginning sentences were considered errors; “buts” were not.
Comma errors or splices are counted.
Measurement
The overall dependent variable is the ability of the poster to acquire funding for
the project. The time to success measure is finite, since all projects must be funded in 28
days in order to succeed. Therefore, we did not use speed to funding as the primary
measure of success, although I did use it as a secondary potential measure of success.
47
Instead, we operationalized funding success as the continuous or binary measures of the
actual final amount raised, the achievement of goal or not (binary), the percentage of
goal achieved (bounded), and the number of funders who contributed. To measure the
effects of our independent variables, we conducted simple linear regressions of the
continuous dependent variables amount raised and number of funders, a separate
logistical regression (LOGIT) with the binary variable goal reached, and a simple linear
regression of the bounded variable percentage of goal achieved.
The independent variables include type of micronarrative identity claims and
number of narrative identity claims. These independent variables are categorical and
ratio, respectively.
An additional group of independent variables was added called structural
variables. These are an exploration of other text factors besides narrative identities that
we controlled for that we speculated might affect persuasive campaign outcomes. These
include Flesch-Kincaid measures of reading ease, grade level, reading index, average
grade level, word count, and sentence count, as well as two measures of proofreading
errors. The latter included a continuous and a binary measure of the amount of
proofreading errors.
The dependent variables are mostly continuous or ratio level, particularly those
indicating funding success, such as amount raised, percentage of goal reached, and
funder count. Amount requested was also controlled for, since success might be
determined in part by that amount, which might have skewed our results.
48
In order to choose the control variables, correlation tests were run to see which
potential controls correlated with both the micro narratives and the dependent variables
and thus could be possible confounds. In the first model, the linear regression with DV
amount raised, seven potential confounds were found to be correlated with both the DV
and with at least one of either the micro narratives or the narrative structures. These were
updates, comments, funders, perk levels, Twitter count, and Facebook count and were
therefore controlled for. In the second model, the logistical regression with the DV
success, four possible confounds were isolated that correlated with both the DV and with
at least one of either the micro narratives or the narrative structures and were therefore
controlled for. These were campaign length, comments, funders, and Facebook count. In
the third model, the linear regression with the DV Percent of Goal Reached, the same
four possible confounds were isolated as in the previous model (campaign length,
comments, funders, and Facebook count) and were likewise controlled for.
As a way of empirically testing the hypotheses regarding the related, but
differently measured outcomes denoting funding success, two different types of models
were employed. First, a multiple regression was used to capture the specific effects of
micro narratives and narrative structures on the amount raised, a continuous measure
between zero and upward infinity. Next, a second multiple regression was used to capture
the specific effects of these micro narratives and narrative structures on percent of goal
raised, a bounded variable with the limited range between 0 and 1. Third, a logistical
regression model (LOGIT) was used to capture the specific effects of these micro
narratives and narrative structures on goal reached, a binary measure of whether or not a
49
campaign achieved its donation goal. The LOGIT model accommodates an otherwise
non-linear spread of data by transforming it to logs, which allow the plotting of a
relatively linear regression. Fourth, two simple OLS regressions were run using number
of narrative identities as predictor variables against amount raised and percentage of goal
raised. An additional LOGIT regression was run using the number of narrative identities
to predict goal reached.
Because another possible indication of the success of a text persuasion appeal in
the present context might also be the number of funders who responded to the appeal, we
decided to also explore any correlation between the number of funders and each specific
narrative identity and cluster of identities.
50
Chapter 4: Study 1 (Inductive Coding)
In study 1, I conduct in-depth inductive analyses to identify the types of self-
identity narratives nonprofit entrepreneurs employ in their crowdfunding requests. The
method of data assignment into categories as used in Study 1 is a qualitative approach.
According to Thomas (2006, p. 238), the method of inductive coding refers to
“approaches that primarily use detailed readings of raw data to derive concepts, themes,
or a model through interpretations made from the raw data by an evaluator or researcher.”
The particular method of coding used in this study is an inductive assessment of specific
occurrences in the data set, leading to assignment of instances in separate categories of
identity narratives. To accomplish this, the researcher conducted a detailed analysis of
text in the sampled funding requests on IGG, deriving concepts, themes, or terminology
which indicated a thematic direction to the identity claimed, either overtly or subtly,
allowing it to be assigned a binary code as present or not. The inductive method is
consistent with the grounded theory of Strauss and Corbin (1998) who describe this
method as beginning with a study area and coaxing the theory out of the data, rather than
beginning with a theory and looking for data which fit it (Thomas, 2006).
Replicated Codes
Given the goal of this study to replicate some aspects of Herzenstein et al.’s
(2011) research findings that specific narrative identities predicted funding success in for-
profit online fundraising, I first developed coding schema for each of these previous
variables. Since the analysis by Herzenstein and colleagues did not offer a detailed
explanation of how instances of text were assigned to specific coding categories, I
51
developed a coding scheme to identify these previous codes, as well as new ones that
were conceived during the textual analysis. To do this, the five basic procedures of
inductive coding described by Thomas (2006) were followed: 1) Raw Data Preparation;
2) Text Reading; 3) Category Creation; 4) Text Overlapping; and 5) Category Revision
and Refinement.
Raw Data Preparation
The raw data files were first collected and saved to a relational database using
Python programming language script. Thus, all raw data script was formatted in a
standard font and size and saved in table form, with rows comprising each separate
webpage and columns comprising each category of webpage elements. The latter
elements include not only textual items, but also images, sound files, URLs and other
background items tangential to the current study. With an eye toward future analysis,
backups of all original files were saved in various locations.
Text Reading
Once the textual elements were collected in the database, a close reading of the
material was performed to gain familiarity with the content and understand the themes
and events it contained.
Category Creation
Once an understanding of the basic elements was obtained, this researcher began
to identify and create categories and themes. Initially, the evaluation group of six
categories from Herzenstein et al. (2011) was identified as existing in at least some of the
posts. Additional categories were derived after repeated readings of the text. The
52
categories were generally created from instances of phrasing or apparent meanings in
actual segments of text. No analytical computer software was used for the actual coding
or creating of categories. Instead, the primary coder individually determined all new
categories.
Uncoded or Overlapping Text
Given the qualitative nature of the inductive coding process, some text may
contain multiple themes, while other text may contain none of relevance to the analysis at
hand. This was the case in the present dataset. Thus, some text contained overlapping
themes, while other text contained no coded themes.
Category Revision and Refinement
Throughout the coding process, category names, descriptions, and content were
changed, revised, and refined until the final coding categories were finalized. For
example, the category titled hip at one point was conceived as youthful but ultimately
renamed as hip to help distinguish it from other categories that contain language focusing
on children and youth. In addition, the kinds of hip language elements that were found
are not necessarily the exclusive domain of young people. This author can think of
several late-night talk show hosts who use several of the phrases, even though they are
well past middle-aged. Such phrases include “make some noise for,” and iterations of
“awesome,” and “cool,” for example. For such reasons, it was determined to rename the
category hip instead. In this way, category names were progressively analyzed, re-
considered, and changed when needed.
1 13 16 Kenton Bruce Anderson PhD Dissertation (7)-1
1 13 16 Kenton Bruce Anderson PhD Dissertation (7)-1
1 13 16 Kenton Bruce Anderson PhD Dissertation (7)-1
1 13 16 Kenton Bruce Anderson PhD Dissertation (7)-1
1 13 16 Kenton Bruce Anderson PhD Dissertation (7)-1
1 13 16 Kenton Bruce Anderson PhD Dissertation (7)-1
1 13 16 Kenton Bruce Anderson PhD Dissertation (7)-1
1 13 16 Kenton Bruce Anderson PhD Dissertation (7)-1
1 13 16 Kenton Bruce Anderson PhD Dissertation (7)-1
1 13 16 Kenton Bruce Anderson PhD Dissertation (7)-1
1 13 16 Kenton Bruce Anderson PhD Dissertation (7)-1
1 13 16 Kenton Bruce Anderson PhD Dissertation (7)-1
1 13 16 Kenton Bruce Anderson PhD Dissertation (7)-1
1 13 16 Kenton Bruce Anderson PhD Dissertation (7)-1
1 13 16 Kenton Bruce Anderson PhD Dissertation (7)-1
1 13 16 Kenton Bruce Anderson PhD Dissertation (7)-1
1 13 16 Kenton Bruce Anderson PhD Dissertation (7)-1
1 13 16 Kenton Bruce Anderson PhD Dissertation (7)-1
1 13 16 Kenton Bruce Anderson PhD Dissertation (7)-1
1 13 16 Kenton Bruce Anderson PhD Dissertation (7)-1
1 13 16 Kenton Bruce Anderson PhD Dissertation (7)-1
1 13 16 Kenton Bruce Anderson PhD Dissertation (7)-1
1 13 16 Kenton Bruce Anderson PhD Dissertation (7)-1
1 13 16 Kenton Bruce Anderson PhD Dissertation (7)-1
1 13 16 Kenton Bruce Anderson PhD Dissertation (7)-1
1 13 16 Kenton Bruce Anderson PhD Dissertation (7)-1
1 13 16 Kenton Bruce Anderson PhD Dissertation (7)-1
1 13 16 Kenton Bruce Anderson PhD Dissertation (7)-1
1 13 16 Kenton Bruce Anderson PhD Dissertation (7)-1
1 13 16 Kenton Bruce Anderson PhD Dissertation (7)-1
1 13 16 Kenton Bruce Anderson PhD Dissertation (7)-1
1 13 16 Kenton Bruce Anderson PhD Dissertation (7)-1
1 13 16 Kenton Bruce Anderson PhD Dissertation (7)-1
1 13 16 Kenton Bruce Anderson PhD Dissertation (7)-1
1 13 16 Kenton Bruce Anderson PhD Dissertation (7)-1
1 13 16 Kenton Bruce Anderson PhD Dissertation (7)-1
1 13 16 Kenton Bruce Anderson PhD Dissertation (7)-1
1 13 16 Kenton Bruce Anderson PhD Dissertation (7)-1
1 13 16 Kenton Bruce Anderson PhD Dissertation (7)-1
1 13 16 Kenton Bruce Anderson PhD Dissertation (7)-1
1 13 16 Kenton Bruce Anderson PhD Dissertation (7)-1
1 13 16 Kenton Bruce Anderson PhD Dissertation (7)-1
1 13 16 Kenton Bruce Anderson PhD Dissertation (7)-1
1 13 16 Kenton Bruce Anderson PhD Dissertation (7)-1
1 13 16 Kenton Bruce Anderson PhD Dissertation (7)-1
1 13 16 Kenton Bruce Anderson PhD Dissertation (7)-1
1 13 16 Kenton Bruce Anderson PhD Dissertation (7)-1
1 13 16 Kenton Bruce Anderson PhD Dissertation (7)-1
1 13 16 Kenton Bruce Anderson PhD Dissertation (7)-1
1 13 16 Kenton Bruce Anderson PhD Dissertation (7)-1

More Related Content

Similar to 1 13 16 Kenton Bruce Anderson PhD Dissertation (7)-1

dissertation_mcnally_FINAL[6-26]
dissertation_mcnally_FINAL[6-26]dissertation_mcnally_FINAL[6-26]
dissertation_mcnally_FINAL[6-26]Dr. Michael McNally
 
David gonzales the art of solving problems-lateral thinking
David gonzales  the art of solving problems-lateral thinkingDavid gonzales  the art of solving problems-lateral thinking
David gonzales the art of solving problems-lateral thinkingarvind_k_mishra
 
Thesis ms llq
Thesis ms llqThesis ms llq
Thesis ms llq
Liangqun Lu
 
Felicia oropeza thesis_final_draft_ready_to_go_qin
Felicia oropeza thesis_final_draft_ready_to_go_qinFelicia oropeza thesis_final_draft_ready_to_go_qin
Felicia oropeza thesis_final_draft_ready_to_go_qinfeoropeza
 
Forlorn and scarred - A situation analysis of child sexual abuse
Forlorn and scarred - A situation analysis of child sexual abuseForlorn and scarred - A situation analysis of child sexual abuse
Forlorn and scarred - A situation analysis of child sexual abuse
UNICEF Europe & Central Asia
 
2020 international survey v04
2020 international survey v042020 international survey v04
2020 international survey v04
otaval
 
Evaluating what Mind, Brain, and Education has taught us about teaching and l...
Evaluating what Mind, Brain, and Education has taught us about teaching and l...Evaluating what Mind, Brain, and Education has taught us about teaching and l...
Evaluating what Mind, Brain, and Education has taught us about teaching and l...
Conexiones: The Learning Sciences Platform
 
FIFTY SHADES OF FREED
FIFTY SHADES OF FREEDFIFTY SHADES OF FREED
FIFTY SHADES OF FREEDlboustany
 
Dr. Karen Deller RPL Thesis
Dr. Karen Deller RPL ThesisDr. Karen Deller RPL Thesis
Dr. Karen Deller RPL Thesis
CIMAP
 
Deller rpl thesis
Deller rpl thesisDeller rpl thesis
Deller rpl thesis
Linda Meyer
 
An research paper
An research paperAn research paper
An research paperDanny Steve
 
Neuroinflammatory Alterations via CD-36 in Traumatic Brain Injury
Neuroinflammatory Alterations via CD-36 in Traumatic Brain InjuryNeuroinflammatory Alterations via CD-36 in Traumatic Brain Injury
Neuroinflammatory Alterations via CD-36 in Traumatic Brain InjuryDiana G. Hernandez O. Ph.D.
 
Global Medical Cures™ | Emerging & Re-Emerging Infectious Diseases
 Global Medical Cures™ | Emerging & Re-Emerging Infectious Diseases Global Medical Cures™ | Emerging & Re-Emerging Infectious Diseases
Global Medical Cures™ | Emerging & Re-Emerging Infectious Diseases
Global Medical Cures™
 
TJ_Murphy_Epistemology_Final_Paper
TJ_Murphy_Epistemology_Final_PaperTJ_Murphy_Epistemology_Final_Paper
TJ_Murphy_Epistemology_Final_PaperTimothy J. Murphy
 

Similar to 1 13 16 Kenton Bruce Anderson PhD Dissertation (7)-1 (20)

dissertation_mcnally_FINAL[6-26]
dissertation_mcnally_FINAL[6-26]dissertation_mcnally_FINAL[6-26]
dissertation_mcnally_FINAL[6-26]
 
David gonzales the art of solving problems-lateral thinking
David gonzales  the art of solving problems-lateral thinkingDavid gonzales  the art of solving problems-lateral thinking
David gonzales the art of solving problems-lateral thinking
 
Thesis ms llq
Thesis ms llqThesis ms llq
Thesis ms llq
 
rosario_phd_thesis
rosario_phd_thesisrosario_phd_thesis
rosario_phd_thesis
 
Felicia oropeza thesis_final_draft_ready_to_go_qin
Felicia oropeza thesis_final_draft_ready_to_go_qinFelicia oropeza thesis_final_draft_ready_to_go_qin
Felicia oropeza thesis_final_draft_ready_to_go_qin
 
Forlorn and scarred - A situation analysis of child sexual abuse
Forlorn and scarred - A situation analysis of child sexual abuseForlorn and scarred - A situation analysis of child sexual abuse
Forlorn and scarred - A situation analysis of child sexual abuse
 
Wiggins dissertationfinal
Wiggins dissertationfinalWiggins dissertationfinal
Wiggins dissertationfinal
 
2020 international survey v04
2020 international survey v042020 international survey v04
2020 international survey v04
 
Evaluating what Mind, Brain, and Education has taught us about teaching and l...
Evaluating what Mind, Brain, and Education has taught us about teaching and l...Evaluating what Mind, Brain, and Education has taught us about teaching and l...
Evaluating what Mind, Brain, and Education has taught us about teaching and l...
 
FIFTY SHADES OF FREED
FIFTY SHADES OF FREEDFIFTY SHADES OF FREED
FIFTY SHADES OF FREED
 
Dr. Karen Deller RPL Thesis
Dr. Karen Deller RPL ThesisDr. Karen Deller RPL Thesis
Dr. Karen Deller RPL Thesis
 
Deller rpl thesis
Deller rpl thesisDeller rpl thesis
Deller rpl thesis
 
Hssttx1
Hssttx1Hssttx1
Hssttx1
 
Tanya's dissertation
Tanya's dissertationTanya's dissertation
Tanya's dissertation
 
1455406
14554061455406
1455406
 
An research paper
An research paperAn research paper
An research paper
 
Nofi paper on sa
 Nofi paper on sa Nofi paper on sa
Nofi paper on sa
 
Neuroinflammatory Alterations via CD-36 in Traumatic Brain Injury
Neuroinflammatory Alterations via CD-36 in Traumatic Brain InjuryNeuroinflammatory Alterations via CD-36 in Traumatic Brain Injury
Neuroinflammatory Alterations via CD-36 in Traumatic Brain Injury
 
Global Medical Cures™ | Emerging & Re-Emerging Infectious Diseases
 Global Medical Cures™ | Emerging & Re-Emerging Infectious Diseases Global Medical Cures™ | Emerging & Re-Emerging Infectious Diseases
Global Medical Cures™ | Emerging & Re-Emerging Infectious Diseases
 
TJ_Murphy_Epistemology_Final_Paper
TJ_Murphy_Epistemology_Final_PaperTJ_Murphy_Epistemology_Final_Paper
TJ_Murphy_Epistemology_Final_Paper
 

1 13 16 Kenton Bruce Anderson PhD Dissertation (7)-1

  • 1. LET ME TELL YOU A STORY: AN EXPLORATION OF THE COMPLIANCE- GAINING EFFECTS OF NARRATIVE IDENTITIES IN ONLINE CROWDFUNDING TEXTUAL APPEALS by Kenton Bruce Anderson December 21, 2015 A dissertation submitted to the Faculty of the Graduate School of the University at Buffalo, State University of New York in partial fulfillment of the requirements for the degree of Doctor of Philosophy Department of Communication
  • 2. ii Acknowledgments Dr. Gregory D. Saxton (Major Advisor), Dr. Michael A. Stefanone (Committee Member), Dr. Joseph Woelfel (Committee Member), Dr. Tom Feeley, Rose Gryckiewicz, Dr. Mark Frank, Aubrey Nye, Kelvin Anderson and my family, Edward J. Marinucci, Jr., Dr. Allison Zorsie-Shaw, Dr. Matthew Grizzard, Dr. Jacob Neiheisel, Emily Dolan, Dr. Rosemarie Murray, Dr. Sean Dugan, Professor Paul Trent, Anne Slowe, Zed Ngoh & family, Phil Teefy, Gil and Carmen Reynolds (Hot Glass Horizons), Jonathan Schwartz, Dr. and Mrs. Ronald and Amelia Schwartz, Dr. Mary Flaherty Rogers, Dr. Wayne Xu, Timothy Pruitt, Isabella Wilklow
  • 3. iii Table of Contents Acknowledgements………………………………………………………………............. ii Abstract…………………………………………………………………………............. vii Chapter 1: Introduction........................................................................................................1 Background of the Problem ...........................................................................................1 Statement of the Problem...............................................................................................9 Purpose of the Study......................................................................................................9 Significance of the Study.............................................................................................11 Primary Research Questions........................................................................................11 Research Design...........................................................................................................12 Theoretical Framework................................................................................................14 Assumptions, Limitations, and Scope (Delimitations) ................................................15 Definition of Terms......................................................................................................16 Summary......................................................................................................................17 Chapter 2: Literature Review.............................................................................................19 Narrative as a Sequential Compliance-gaining Technique..........................................19 Donations, Non-profits, Online Giving, SNSs, P2P Lending and Crowdfunding ..................................................................................................19 Narrative ......................................................................................................................22 Identity Claims, Trust, and Need.................................................................................25 Multiple Identity Claims..............................................................................................28 Time to Funding………………………………………………………………...……29
  • 4. iv Micronarratives............................................................................................................29 Why Measure Micronarratives?............................................................................ 31 Archetypes ............................................................................................................ 36 Relationship Between the Independent and Dependent Variables ..............................36 Goals ..........................................................................................................................37 Research Questions......................................................................................................39 Chapter 3: Research Method..............................................................................................41 Sample Selection Process ............................................................................................41 Collection Method ................................................................................................ 43 Coding................................................................................................................... 43 Structural Variables .....................................................................................................46 Measurement................................................................................................................46 Chapter 4: Coding..............................................................................................................50 Replicated Codes .........................................................................................................50 Raw Data Preparation ..................................................................................................51 Text Reading......................................................................................................... 51 Category Creation................................................................................................. 51 Uncoded or Overlapping Text .............................................................................. 52 Category Revision and Refinement ...................................................................... 52 Initial Coding Categories .............................................................................................53 Trustworthy........................................................................................................... 53 Successful ............................................................................................................. 53
  • 5. v Hardworking ......................................................................................................... 54 Economic Hardship............................................................................................... 54 Moral ................................................................................................................... 54 Religious ............................................................................................................... 55 Additional Coding Categories......................................................................................56 Needy ................................................................................................................... 57 Helpless................................................................................................................. 57 Victim ................................................................................................................... 57 Child-mentioned ................................................................................................... 58 Child-centered....................................................................................................... 58 Courageous ........................................................................................................... 58 Grateful ................................................................................................................. 59 Hero ................................................................................................................... 59 Hip ................................................................................................................... 60 Happy ................................................................................................................... 61 Chapter 5: Study 2 (Multivariate Analyses) Results..........................................................62 Chapter 6: Discussion, Conclusions, and Recommendations............................................70 Theoretical Implications ..............................................................................................72 Limitations and Future Research .................................................................................75 Practical Implications...................................................................................................76 Conclusion…………………………………………………………………………...77 Appendix 1: Tables............................................................................................................78
  • 7. vii Abstract This study assesses the compliance-gaining impact of specific micronarrative identities on non-profit fundraising in the online environment. It examines the effects of narratives in crowdfunding requests in the non-profit category of entities on the peer-to- peer (P2P) crowdfunding website Indiegogo. The study has several goals: 1) distinguish micronarrative identity claims from longer narratives and explore the compliance-gaining influence mechanisms behind both simple, micronarrative identity claims and longer narratives, by exploring previously studied and newly described identity claim categories in the texts of non-profit donation requests; 2) explore the relevance of recent claims about trust in the for-profit lending realm for their applicability to the realm of non-profit donations; and 3) determine the relevance to nonprofit funding requesters of narratives relaying need information. In study #1, in-depth inductive analyses were conducted to identify the types of narratives nonprofit entrepreneurs are employing in their crowdfunding requests, while in study #2 multivariate analyses were used to investigate the relationship between the use of micro-narratives and the success of crowdfunding requests. The study found narratives containing mentions of need, children and victim terminology have a significant impact on funding success rates for non-profit lending requests. In addition, hipster and hero narratives significantly affect the number of funders. The study also supports previous research finding that the discrete number of narratives is negatively correlated with funding success. By implication, having a coherent narrative seems more important. Secondly, regarding other elements of the campaign, such as making regular updates, adding photos and perk levels, and longer
  • 8. viii campaigns – it appears that providing current information and visual imagery leads to greater funding success, as does providing more choices of perk levels. Finally, with the narrative structure variables, longer campaigns help, as does having one’s request professionally written or being low in illiteracy. These findings have potential implications for requesters attempting to crowdsource funds on Internet lending sites.
  • 9. ix Keywords: crowdsourcing, crowdfunding, nonprofit fundraising, donations, Indiegogo, micronarrative identity, compliance gaining, P2P lending, sequential messages, narrative
  • 10. Chapter 1: Introduction Online funding is a growing phenomenon in the 21st Century, as is evidenced by multiple websites devoted entirely to funding individual needs, projects, and investments. Some websites sponsor requests from for-profit entities, while others sponsor requests from not-for-profit entities. These entities can include both institutions and individuals. Some websites sponsor assorted requests from both for-profit and not-for-profit entities. How these requests are structured to maximize their effectiveness is of importance to both requesters and funders. The literature of self-presentation has explored this from several approaches, including cross-cultural and network analysis (e.g. Rui & Stefanone, 2013a; Rui & Stefanone, 2013b). The story, or narrative, that requesters tell about themselves to identify themselves to their audience is one part of the request’s structure that may significantly impact their funding success or failure. This two-phase study explores the use of narratives in the requests in order to decide what specific narrative identities or other narrative elements are used in the more successful posts and, conversely, which narrative identities or other elements requesters might be advised to avoid using to ensure greater success. Background of the Problem In the postmodern era, the rise in the popularity of crowdsourcing and crowdfunding is changing the arena of fundraising for both individuals and institutions (Dholakia and Firat, 2006; Hibbert and Horne, 1996). Because individuals comprise 80% of funders (Hibbert and Horne, 1996), however, it is especially important to explore persuasion processes that impact individual funders. Consequently, institutional giving is
  • 11. 2 beyond the scope of this study. Discovering what we can know about individuals’ motivations for giving is therefore an important step in maximizing non-profit online fundraising attempts, a topic of relevance not only to individuals, but also to institutions and governments (Psacharopoulos and Nguyen, 1997). With a few important exceptions, the empirical foundations of the online crowdfunding phenomenon have generally been unexplored in the scientific literature (e.g. Herzenstein, Sonenshein, and Dholakia, 2011). While researchers have begun exploring the impact of visual images in online non-profit persuasive appeals (e.g. Anderson and Saxton, 2016), of particular relevance to the present study are the persuasive factors in textual appeals that motivate individuals to give to other individuals or institutions. Some communication scholars might assume that such textual persuasive factors are limited primarily to argumentative elements such as evidence, logic, cognition, nonverbal elements, or formal techniques as exemplified by the currently conceptualization of sequential messages (e.g. Dillard, Hunter, and Burgoon, 1984). However, it has become apparent to some communication theorists (e.g. Green, Garst, and Brock, 2004) that an important source of data describing individual persuasive communication processes has been largely ignored -- that of narrative -- and this growing number of empirical researchers and theorists have begun to align themselves with the postmodern focus upon narrative, or storytelling elements, that is influencing empirical research in many other disciplines (see Reissman and Quinney, 2005). Understanding the current state of the scientific exploration of persuasive textual elements in online posts requires a beginning distinction between rhetorical and narrative
  • 12. 3 approaches to persuasion. In their seminal work, Green, Garst and Brock (2004, p. 162) say persuasion research has primarily focused on rhetorical messages, rather than narratives. They define rhetoric as: “primarily fact-based advocacy messages, such as advertisements, speeches, and editorials, that contain arguments specifically designed to sway a reader to a particular position (p. 162).” But narrative content -- which is distinguished as containing characterization, plot elements and setting information -- is at least as persuasively powerful as rhetorical “factual” statements, according to Strange and Leung (1999) and Green and Brock (2000). This paradoxical state of the literature, which has continued nearly unabated to the present -- as I will show in the following chapter -- indicates the importance of studying narrative elements in persuasion in addition to rhetorical elements. Understanding the foregoing distinction between rhetorical and narrative approaches to researching persuasion is only one step in discovering the persuasive impact of narratives, however. In order to fully conceptualize the persuasive influence of narratives, and more fully tie them to communication theory, it might be instructive to show their relevance to a rather well-developed field of empirical research and theory in communication that has previously not been seen by most scholars as connected to narrative: the realm of sequential message compliance-gaining techniques. This conceptualization of them as sequential compliance-gaining message techniques is based upon the general consensus of narrative researchers on the central importance of their sequential nature (Czarniawska, 1998; Herzenstein, et al. 2011; Martens, Jennings, and Jennings, 2007; Merchant, Ford, and Sargeant, 2010; Riessman, 1993).
  • 13. 4 Business and marketing researchers, particularly those studying entrepreneurialism and banking have been the primary explorers of persuasion and funding as related to identity presentation. Given that funding is a major focus of activity for not only business entities such as entrepreneurs, investors, and lenders, but also for charitable organizations and donation requesters, researchers have generally assumed that the two areas might offer pertinent insights for each other (e.g. Bennett, Foot, and Xenos, 2011; Boje, 1991; Dholakia and Firat, 2006; Fiol, 1989; Hjorth and Steyaert, 2005; Hytti, 2003; Makkonen, Aarikka-Stenroos and Olkkonen, 2012; Martens, Jennings, and Jennings, 2007; Rhodes and Brown, 2005; Schembri, Merrilees, and Kristiansen, 2010). However, some conflicting research indicates that there may be important differences between effective fundraising for profit-making entities and effective fundraising for non- profit donation seekers. It appears that the choice of self-identities or organizational identities is a potentially problematic practice that may not be equivalent across both for- profit and non-profit narrative fundraising attempts. Research to date has still left unresolved the question of how these two groups might need to differentiate their approaches to narrative self identity presentation, an issue the present study attempts to help clarify. Importantly, a preliminary question that presages the effectiveness of particular appeals across for-profit and non-profit arenas is whether storytelling can actually help to secure funding in non-profit settings. While a form of this question has been addressed briefly in the business literature, which has observed such an effect in for-profit fundraising (Herzenstein, Sonenshein, and Dholakia, 2011; Martens, Jennings, and
  • 14. 5 Jennings, 2007), this preliminary question has yet to be addressed in the non-profit fundraising literature. The task for the current study is to determine a methodological approach which might help shed first light on these foundational questions: Are stories containing identity narratives related to the success of non-profit funding appeals? How? Why? In chapter three, I will outline the methodological approach our limited sample of identity-making on a real-life fundraising website allows us to use to flesh out the connection between storytelling and funding success. The scant existing empirical research on the impact of narrative elements in online fundraising persuasiveness (Merchant, Ford, and Sargeant, 2010) has identified several such elements that appear related to effective fundraising appeals, though this research has to date only explored these elements as used in for-profit appeals and in entrepreneurial settings (e.g. Herzenstein, Sonenshein, and Dholakia, 2011; Martens, Jennings, and Jennings, 2007; Navis and Glynn, 2011). These elements derive from the processes of individual and organizational identity creation engaged in by individual requesters, entrepreneurs and impression managing entities. These seminal research studies on funding requests in for-profit lending and entrepreneurial image management present evidence that trust and risk concerns are influential in financial decision-making. Consequently, narrative identities that address these concerns are predictive of effective fundraising (Herzenstein, Sonenshein, and Dholakia, 2011) in the for-profit realm explored in these studies. What remains unanswered is whether this model holds also for non-profit fundraising, or whether different narrative elements will emerge as more relevant in that arena -- particularly given that theory in charitable fundraising proposes
  • 15. 6 that trust and risk are less relevant to donors than are determination of need and dedication to the cause (e.g. Bekker and Wiepking, 2011). Theorists of non-profit fundraising propose that donor assessment of need and belief in the cause espoused by the non-profit entity will each predict giving behavior better than will trust or risk considerations (Bekkers and Wiepking, 2011). To date, empirical support for this claim appears to be nonexistent. However, the recent study by Herzenstein, Sonenshein, and Dholakia (2011) presents limited empirical support for the relevance of trust-based narrative identities in for-profit lending considerations. There is a paucity of research addressing this issue. Therefore, although this dataset doesn’t allow us to measure level of belief in the cause -- either directly or by proxy -- we will explore the relative effectiveness of trust-based narrative identities versus need-based narrative identities in predicting the successful funding of the requests gathered in the present sample set. Of particular significance will be what direction such potential relationships exhibit, if found, and what other narrative variables besides trust might be relevant to funding success. Given the centrality of exploring narrative self-identities in the empirical literature on narrative persuasion (see Green, Garst, and Brock, 2004; Herzenstein, Sonenshein, and Dholakia, 2011), and the presumed use of them to present one’s best self (Goffman, 1959; Scott and Lane, 2000), the problem arises of whether to treat these self-identity narratives as fictional or non-fictional accounts of personal reality. A distinction that might be made at this point is whether there exists a practical difference between fictional and nonfictional narrative accounts. Given the relative newness of the field of narrative
  • 16. 7 identity study, this problem has received little attention in empirical study, but narrative theorists predict that the difference between fictional and nonfictional narratives is largely irrelevant to the persuasive impact of the narratives (Bruner, 1986; Green et al. 2004; Rubin, 1994). In practice, it may be impossible to cleanly distinguish individual self-identities as either purely non-fiction or purely fiction, but by current accounts this is not a consideration that will determine the relative persuasiveness of the narrative identities themselves and therefore this study does not propose to directly address this problem. At present, identity theory proposes that human beings expect coherence from each other’s narrative identities (Polkinghorne, 1998, as related in Henriksen, Holmegaard, Ulriksen, and Madsen, 2015). This also appears consistent with predictions of exemplification theory and the identified victim effect (Polkinghorne, 1998, as related in Henriksen, Holmegaard, Ulriksen, and Madsen, 2015). We would expect this to imply that increasing the number of identities would lead to less coherence and therefore lower funding success. However, extant research apparently shows the opposite effect. Previous study of for-profit lending found increasing the number of narrative identities predicted funding success (Herzenstein et al., 2011). What remains to be tested is whether this effect is robust across domains. Will this hold in the non-profit realm? Will the inclusion of multiple identities in a non-profit narrative appeal be linearly related to success? If so, will the direction of the relationship be positive or negative? If the relationship is negative, it should lend support for the above mentioned three models, but if the relationship is positive, what alternate explanation can be found for the effect?
  • 17. 8 It is possible that identities contain another characteristic that predicts persuasion. Several researchers have isolated narrative valence as a predictive factor (Herzenstein, Sonenshein, and Dholakia, 2011; Merchant, Ford, and Sargeant, 2010; Parry and Kempster, 2013). The degree to which a narrative identity conveys, enhances, or provokes so-called “positive” or “negative” emotions has been shown to impact effectiveness (Merchant, Ford, and Sargeant, 2010). However, none of this research has yet addressed the question of whether this effect plays a role in the online persuasion arena. This is another potential issue that the present study seeks to address. Many narrative theorists (see Reissman and Quinney, 2005) agree that exploring only the most complete narratives including elements of character, plot and setting is essential for “best” narrative research. However, can we be certain that only complete narrations affect behavior? Or can fragmentary, implied, or archetypal micro-narrative identities/characterizations also do so? Toward this end, the present two-phase study has inductively-derived a group of narrative identifiers contained in the narrative texts of our sample. In order to better tease out as completely as possible all identities in our posts, we have identified what we believe are the smallest units of narrative that can predict actions. We call these units of characterization we find in our sample micronarratives. As components comprised of only characterization and, in at least several cases, implied or presumed emplotment (vis-à-vis their apparent conceptualization as oppositional to another character or force) but primarily without reference to contextual settings, use of these micronarrative identities allows us to respond to the related research-provoking
  • 18. 9 question asked by Green, Garst, and Brock (2004): What are the boundaries of the power of context-free narrative? Statement of the Problem Recent research indicates that specific narrative identities are related to funding success in the fast-growing online for-profit crowdfunding phenomenon. Little is known, however, about whether specific self-identity narratives have an effect upon the success of online non-profit crowdfunding appeals, and if they do have an effect, which ones have an effect, and what that effect will be. There is thus a gap in the knowledge about the nature of the textual narrative identities that are most effective in securing nonprofit crowdfunding online. Purpose of the Study The purpose of this two-phase qualitative and quantitative study is to explore a random sample of funding requests retrieved from a major online crowdsourcing website, Indiegogo (IGG), in order to explore the narrative elements used in the requests and the relative successfulness of each. In study #1, I conduct in-depth inductive analyses to identify the types of narratives nonprofit entrepreneurs are employing in their crowdfunding requests, while in study #2 I use multivariate analyses to quantitatively investigate the relationship between the use of micro-narratives and the success of crowdfunding requests. I ran a series of regression analyses to determine precise associations between these specific, inductively generated narrative identities and several measures of funding success, including both metric and binary data. In order to achieve the highest explanatory power possible, this study conceptualizes five distinct ratio-level
  • 19. 10 dependent variables (DVs): amount raised, percentage of goal reached, number of funders, time to funding, and number of narrative identities. It also conceptualizes one binary DV: goal fulfillment, which measures whether or not the requester received complete funding of their stated fiscal goal. The independent variables (IVs) and controls form another subset of data looked at in the study. The IVs are inductively determined narrative identities, based in part upon Herzenstein, Sonenshein, & Dholakia’s (2011) list of categories, derived from the trust model of Mayer, Davis, and Schoorman (1995). The original list included narrative identities of trustworthy, successful, hardworking, economic hardship, moral, and religious. The present study presents an additional group of inductively generated identities including needy, helpless, victim, child mentioned, child-centered, courageous, grateful, hero, hip, and happy. The extent to which each identity contributes to a fundraising campaign’s success or failure is one of the central questions this study seeks to address. Items controlled for include both non-structural and structural elements found on the webpages. Non-structural items are those items that have no part in the narratives themselves, but are instead merely items competing with narratives for conscious or unconscious consideration by the funders. They include measures of campaign length in seconds and numbers of, respectively, updates, comments, funders, gallery images, perk levels, Tweets, and Facebook likes. Structural items are those elements of narratives that are not identity claims precisely, but instead help describe other aspects of the narratives that may influence success. They include measures of reading ease, grade level, auto
  • 20. 11 read index, average grade level, word count, sentence count, and errata—both metric counts of total instances and binary determination of errata as present or not present. Errata can include instances of grammatical, spelling, punctuation, syntactical, and idiomatic errors as well as non-standard language usage. Significance of the Study This study is significant in that it attempts to address the gap in the knowledge about which narrative identities contribute to a fundraising campaign’s success and failure and how and why they might do so. Its aim is to increase the understanding of sequential messaging compliance-gaining attempts in the non-profit online lending realm, specifically the presentation of individuals’ narrative identities. This increased understanding might serve to better inform theoretical models of giving, fundraising, and website management, as well as help those individuals or organizations desiring funding to make better decisions about how they present themselves in the online crowdsourcing environment. Given the study’s contribution to advancing theory, the results may also generalize to other fundraising venues as well. Government and large charitable institutions perhaps could use the findings to improve their own fundraising policies and practices. Primary Research Questions Therefore, I ask the research questions, “What aspects of narrative identity presentation will relate to non-profit funding success?” and “What will be the direction of each relationship?”
  • 21. 12 Research Design This study uses a two-phase design in which a qualitative inductive analysis (i.e. coding) is performed in Study 1, followed by a quantitative analysis in Study 2 – in this case using two types of regressions – as outlined by Lee (1999). This design is recommended as having great potential for enhancing research, though its weakness is that some researchers may apply it in an unbalanced way if they have greater facility with the methods used in one or the other of its phases. The two-phase design is less complex than a formal mixed methodology design, which contains multiple qualitative and quantitative methods (Lee, 1999). By combining multiple qualitative and quantitative methods, a mixed methodology design, in contrast to a two-phase design, could increase confidence in inferences drawn from findings (Lee, 1999); however, this was not a primary concern for this exploratory study, so the two-phase method was chosen instead. This study has one qualitative method (coding analysis) and one overarching quantitative method, regression analyses. Although two types of regressions – ordinary least squares regression and logistical regression – are herein used in multiple iterations, all are examples of one overarching kind of analysis used at the same point in the research process, so I consider them as one quantitative methodology overall, instead of multiple kinds. As laid out in Lieblich et al. (1998), Study 1 takes the approach of a categorical analysis, abstracting words or sections of words as fitting into particular categories, derived from a coding strategy. This coding strategy is subject to limitations including the precept that narrative accounts are multiplicitous, as each narrator has many voices and
  • 22. 13 self-representations. Narrative analysis is not necessarily a fact-finding mission. It doesn’t assume that people narrate objective facts about their lives. Rather, as proposed by Donald Spence (1982), it assumes that truth in narrative is a reconstruction of experience, not an historical record of actual facts. People construct meaning for and about their lives making linkages between events or aspects of life and how they understand and interpret these. Philosophers foundational to narrative analysis include Ricoeur, Heidegger, Husserl, Dilthey, Wittgenstein, Bakhtim, Lyotard, MacIntyre, and Gadamer (Josselson, 2004). This dissertation also assumes the perspective of Mikhail Bakhtin (1981) that the narrative self is dynamic and never fully finished, so that self-narratives will be, by their nature, inconsistent, resisting closure, and comprised of many voices dialoguing. Such selves are more than just making themes; they are also layering and connecting themes to each other (Josselson, 2004). Josselson lists the following 4 stages of analysis: overall reading for structure, multiple readings for identifying voices, developing a Gestalt for coherence, and touching base with theoretical literature. Being aware of the researcher’s own presuppositions is also required. Parry and Kempster (2013) assert that their mixed methods approach is narrative, rather than positivist, but is called narrative positivism, and more precisely aesthetic narrative positivism. They cite evidence that due to the increased exploration of narrative in the social sciences, boundaries between science and narrative are blurring. In their method (as in mine), narrative is the data. Their “intention is for positivism to assist the narrative rather than for narrative being of assistance to positivism.” (p. 6) My study does
  • 23. 14 not concern itself with the interiority of the subject (as would be expected from the life story identity model of McAdams, 1993—informed by Erikson, 1963, and James’s “I” and “Me” distinction, 1963), but rather with the exteriority or social claims made. In Smith and Sparke’s (2008) terminology, my study is “thin individual” and “thick social relational” in its focus or “spotlight,” since I am looking primarily for evidence supporting social processing of individual claims, rather than the subjective germination of such claims, presented as they are within a social sphere. My coding process in Study 1 began in the manner of Boyatzis’s (1998) template (codebook) approach, as outlined by Crabtree and Miller (1999). Codes replicated from the Herzenstein et al. (2011) study were outlined a priori based on preliminary analysis of text, rather than inductively from the actual sample, then other codes were added as they were inductively derived from the actual sample. Theoretical Framework My underlying epistemological assumptions are grounded primarily in the anti- positivist Symbolic Interactionism of Mead (1963) and the Chicago school of sociology, with branches stretching into the communication field through Galileo theory (Woelfel, 2013; Woelfel, 1992; Woelfel & Fink, 1988). My approach to knowledge production therefore reflects a healthy skepticism of pure positivism, because my approach does not presuppose any absolute “scientific truth” is attainable. Rather, my approach presumes all scientific fact, including any derived herefrom, is mutable, and subject to verification by the community of other scientists, who will, presumably, change or add to it. Such
  • 24. 15 scientific fact also assumedly bears the marks of cultural bias, ethnocentrism, and historical milieu. My research approach cannot be claimed as purely narrative, as it does not meet all the formal criteria for narrative research laid out by Riessman and Quinney (2005), calling for in-depth and extensive, dialogical narrative collection, full exploration of emplotment, and full consideration of context or setting. Given, however, that I attempt to clearly identify my biases and philosophical approach, I believe I am approaching a quasi-narrative approach, in which I attempt to include as many narrative elements as practicable and applicable, rather than a pseudo-narrative approach in which I would attempt to obfuscate my admittedly partial, perhaps even impoverished use of narrative elements by claiming to do otherwise. Assumptions, Limitations, and Scope (Delimitations) The self-evident truths I assume for the purposes of this study are: first, I assume that the random sample gathered is indeed representative of the online funding appeals posted in the non-profit section of the Indiegogo.com website during the six month period between April 1, 2012 and September 30, 2012. Second, I assume that the six measures of funding success actually capture the effect. Third, I assume that relationships such as might be found in our regression analysis do not predict causality. Fourth, I assume that, should those potential relationships be found to be robust, further research – including laboratory experimentation – would be necessary to establish causality. One limitation of the study over which I have no control is the choice of instrumentation for data-gathering. The only practical way of gathering the needed raw
  • 25. 16 data appeared to be Python language-driven databases. Any shortcomings of this method of data gathering were thus beyond my control. Another limitation of this study over which I have no control are my unconscious biases, which might have significantly affected my coding scheme development and implementation. It is possible that my cultural heritage as a Caucasian male of a certain age may have caused me to overlook some possible alternate codes or explanations for effects I found. In addition, it is also possible that, although attempting to be as precise as possible in my language choices, I may have inadvertently over- or understated some relationships, claims, or observed narrative phenomena. Any of these might be potential weaknesses of my study. The scope of this study included a comprehensive dataset gathered over more than a six-month period of all non-profit funding requests on a major crowdfunding website, Indiegogo.com. The population was ultimately reduced to a six-month period; then a random sample of 200 posts was selected from this population. This one-time sample of 200 posts served as the basis for all the measurements described in this study. Given the random nature of my sample from the targeted population, I feel confident that the results attain generalizability to other periods and places beyond that population gather from which I took my sample. Our choice to deliberately delimit our population gather to a six- month period was done for simplicity and to ensure that all initial posts of my final sample of requests would be fully retrievable. Definition of Terms Several terms may cause confusion among even highly experienced researchers in this topic area. Therefore, I will define the following terms. Narrative identity is that
  • 26. 17 process theorized to occur within every consciously aware human being wherein she engages in mental deliberation about, identification of, and presentation of -- either to herself, others, or both -- her nature, function and place within the mental historical projection she creates of her own lifeline. It theoretically is the driving force for establishing personal meaning and significance in one’s life. Inductive coding is the process of reading text repeatedly, with the aim of distilling conceptual elements that can be categorized into identifiable groupings or schemas. Positivism is a philosophy that asserts that science is capable of positively identifying universal truth in the objective world using a carefully applied scientific method, including observation, experimentation, and theory testing. Crowdsourcing is the activity of mass mediated efforts to accomplish goals held in common by online participants. Crowdfunding is a type of crowdsourcing specifically geared toward raising money, particularly in transactions between individual requesters and potential masses of funders. Micronarratives are short, perhaps fragmentary story elements including traces of one or more of the following: characterization, plot, and setting. Emplotment is the process of including within a narrative any elements of plot development, particularly conflicts and resolutions arising from incidents and actions taking place in the story. Context-free signifies the absence of all indicators of setting or “place” where story action occurs. Summary In this chapter, I previewed the background of the research problem and identified the gap in the literature that I will address in this study. I also outlined the purpose of the study and its significance for the field of study, and proffered two formal research
  • 27. 18 questions that I shall explore in the study. I then briefly previewed the methodology I use in the study and laid out my underlying philosophical framework, assumptions, limitations, and scope and closed with a preview of definitions I use in the study. In Chapter 2, I will next review the studies relevant to the present two-phase study, showing that each fails to solve the research problem I have set out to guide this study. In this way I will seek to establish the relevance of the present study for the discipline at large.
  • 28. 19 Chapter 2: Literature Review From the days of Aristotle’s Rhetoric and the works of Cicero to modern courts of law, scholars, specialists and laypersons have all faced the need to explore and understand persuasion. Contemporary empirical persuasion research includes compliance gaining – the study of strategies used to induce others to perform actions, rather than change their opinions, beliefs, or emotions. While the actions requested are often in the nature of purchase requests by for-profit entities, there is also a sizable literature addressing their use in nonprofit fundraising efforts. Narrative as a Sequential Compliance-gaining Technique But what techniques influence compliance? To date, much of the literature has focused on power (French & Raven, 1960), compliance tactics and strategies (Marwell and Schmitt, 1967), and sequential effects (Cialdini, 1984). Less well developed, however, is the study of narratives: sequentially organized (Reissman, 1993; Herzenstein, Sonenstein, and Dholakia, 2011) persuasive words, phrases, and ideas that create a coherent logic, identity, or story. Thus, while some researchers have theorized that Homo sapiens is a story-telling animal (Gottschall, 2012; Gottschall and Wilson, 2005), surprisingly, in the compliance-gaining literature, the existing research on the effects of narrative is rather limited. Donations, Non-profits, Online Giving, SNSs, P2P Lending and Crowdfunding Civilized society is dependent on donations (Zhuang & Saxton, 2014). Of high interest to individuals, scientists, organizations, and governments, therefore, is discovering which techniques garner more donations and why. Donations are especially
  • 29. 20 interesting as a measure of compliance gaining. This is particularly true in the arena of non-profit online funding requests (e.g. Guo & Saxton, 2013; Hackler & Saxton, 2007; Nah & Saxton, 2012; Saxton, 2005; Saxton & Benson, 2005; Saxton & Guo, 2011; Saxton, Guo, and Brown, 2007; Saxton, Neely, & Guo, 2014; Saxton, Oh, & Kishore, 2013; Saxton & Wang, 2013). Contemporary research on compliance gaining and narrative has focused on print media, while electronic media has been less studied. This study therefore extends the research somewhat by analyzing online requester narratives, a sequential form of influence on compliance (Herzenstein et al., 2011; Reissman, 1993; Marten, Jennings, and Jennings, 2007), to discover what narratives predict donor behavior, and see how this may be accounted for by current narrative theory. Online giving is an important and growing industry. According to the 2013 Blackbaud Index, online giving accounted for 6.4% of all charitable giving, excluding grants, in 2013. The Index also reports that overall giving rose 4.0% in August 2014 compared with the same month in 2013, but that online giving rose 12.7% in August 2014 compared to August 2013. Thus, 2014 was the second year in a row in which online giving had increased in double digits over the previous year. The growth of funding online for non-profit organizations is ongoing and shows no sign of abating soon (Blackbaud, 2014). Traditionally, large nonprofit entities have become the repositories of much charitable giving. Currently, however, there is an online crowdsourcing fundraising revolution occurring in which cumbersome corporate-like charitable entities are competing head-to-head with massive numbers of individuals directly asking for help,
  • 30. 21 peer-to-peer (P2P), from their financially able individual peers (Satorius and Pollard, 2010). This new world of online crowdfunding has given individuals an alternative funding resource that appears to offer advantages of speed, simplicity, and greater personal involvement and satisfaction for all involved. Dholakia and Firat (2006) mention that this diffused marketing signified by the postmodern progress away from centralized organizational “management” toward the participatory and co-creative elimination of the marketing “membrane” is somewhat like a neural network constantly working to “re(de)construct” itself (p. 151). Given the ubiquity of online interactivity, person-to-person (P2P) online social networking websites have interested researchers for at least the past ten years (Trammell, Williams, Postelnicu, and Landreville, 2006). Research has been conducted on P2P interaction conducted on social networking sites and political candidate websites and blogs, for instance (Trammell et al., 2006). Moreover, as interactive crowdsourcing sites have expanded, they have become increasingly important arenas to find and measure the outcomes of real-world compliance-gaining attempts. Chaffee and Rapp (2012) describe P2P lending conducted through websites that offer individual investors the chance to lend funds to individual borrowers and promise lower borrower interest rates and higher investor return rates. Such websites have generated significant media attention as well as regulatory concerns at both the state and federal levels. Recently, narrative effectiveness in for-profit funding requests has begun to be addressed in entrepreneurial research (Boje, 1991; Martens, Jennings, and Jennings, 2007; Navis and Glynn, 2011), network research (Bennett, Foot, and Xenos, 2011), and
  • 31. 22 organizational research (Herzenstein et al., 2011). Not yet addressed, however, is narrative effectiveness in not-for-profit donation requests. This study will therefore explore the impact of narrative in the online environment of not-for-profit, P2P funding requests. Narrative In their seminal work, Gottschall and Wilson (2005) argue that humans by their nature are storytellers. Whether this claim is justified or not, considerable research has explored the relevance of storytelling in business. Stories, or “narratives” have recently been the focus of research in entrepreneurialism, for example (e.g. Martens et al., 2007; Herzenstein et al., 2011; Navis and Glynn, 2011). To explain how narratives lead to compliance-gaining success in entrepreneurial efforts, Martens et al., (2007) argue that narratives work in three ways to help capital acquisition. They 1. help comprehension by simplifying explanation of the offering into a coherent unity. 2. explain reasoning behind offering so investors see its nature and value. 3. generate commitment and interest when they connect “to broader contextual narratives” and make distinctive, original offerings appear sound. In these ways, storytelling reduces uncertainty, motivates involvement, and mobilizes capital. Martens et al. (2007) present their seminal study as “the first systematic, large- sample test of the overarching claim that effective storytelling can facilitate external resource acquisition.” They develop and test a theory of how narrative content affects lending behavior in entrepreneur development requests. They explore first the notion of
  • 32. 23 whether stories help resource providers and, if so, in what ways. They postulate that simplification of packaging and elaboration of reasoning are two such ways and they propose that connecting to broader contextual narratives from the relevant field, marketplace, or pool of competitors is a third way that narratives increase interest and participation. So far, these postulates have not been formally tested for their generalizability to the non-profit online donation context. Martens et al. (2007) mention that the narrative form allows the narrators to construct a sort of play in which they as lead characters describe and understand themselves. Stories relay facts, but in a memorable way that unifies disconnected parts into a unified whole. Moreover, stories relay these facts in a sequenced way using historical perspective to increase the certainty of the proposed progression. Martens and colleagues further propose that this process reduces uncertainty and leads to greater facilitation of resource acquisition. We should expect this role to be less important in non-profit than for-profit, but our study is not designed to compare the two roles thus here. Martens and colleagues also see a major role of narrative to be reducing information asymmetry, since they share information with the inquisitors. In Study 1 of this dissertation, I first explore whether narratives do, indeed, include self-identity elements. Then I categorize those identities. I will follow the convention established by the seminal study of Green and Brock (2000) in describing these instantiations of self-identity as public narratives, specifically external public narratives, to which more than one person at a time is exposed.
  • 33. 24 Martens et al. (2007) also theorize that complexity will reduce the effectiveness of narratives, a position consistent with both the heuristic-systematic model (Chaiken, 1980) and the elaboration likelihood model (Wegener and Petty, 2001). Both of these models predict people’s attention will remain focused on the main features of the messages, when they have high capacity and motive for processing. Committing personal or corporate funds in the face of risk of loss would seem a time when motive for processing would be high. They found a diminishing return on complexity of messaging. This extends in a way to the number of narrative identities themselves used. Based on the social cognition research of Brewer, Weber, and Carini (1995), Martens and colleagues predicted tapping too “many categories of social identification” would likely lead to “recognition confusion.” When this is combined with low levels of familiarity with the social categories, adding more identities leads to diminishing returns. Finally, they raised a call for future mixed methodology studies in order to demonstrate that storytelling is more than rhetoric and is effective in producing outcomes. However, their study did not examine this effect in the non-profit arena to see if the same results will still hold. The present two-phase study attempts to partially address this call for mixed methodology study by determining if narrative identity self-presentation can predict funding success. According to Herzenstein et al. (2011), narrative gives meaning to events mentioned by the narrator. This autobiographical window also can be used to reveal the authors’ self concept and identity, at the discretion of the authors, giving the authors the opportunity to manage their image or identity. This is analogous to the process of strategic impression management (Leary and Kowalski, 1990), wherein presenters choose
  • 34. 25 to portray identities most likely to gain support. Narrative is also the best method for showing character development, according to the seminal work of Green and Brock (2000). Isolating the character development aspect of narrative, as Green and Brock have done, lends power to my conceptualization of micronarratives, developed later in this chapter. Navis and Glynn (2011) built upon the findings of Martens et al (2007) and codified them in more detail. They explored how entrepreneurs are marked by newness, a liability until legitimacy is acquired by endorsements, which require some conformity. This results in a tension between legitimacy and distinctiveness. They proposed that identities of entrepreneurs are favorably judged when their distinctiveness is legitimate. They also predicted that these judgments are context-dependent and mediated by narratives of identity which give ‘institutional primes’ and ‘equivocal cues’ when investors try to make sense of offerings. Thus, an organizational identity is part of an organizational “theory of being.” Their model to date has only been tested in entrepreneurial for-profit settings. Identity Claims, Trust, and Need According to Herzenstein et al. (2011), identity claims are an important aspect of narrative self-presentation used in public discourse. Such claims help establish the “who we are” and “what we do” facts about the narrating entity. They might include, for example, short verbal descriptions of the requester as needy, community service-oriented, hardworking, or trustworthy. Herzenstein and colleagues found that narrative identities describing trustworthiness appear to account for a significant amount of success in for-
  • 35. 26 profit lending situations. Specifically, they coded narratives containing descriptions of the requester that fit one or more of the following trust-based categories: trustworthy, successful, hard-working, economic hardship, moral, and religious. These are factors predicted by the trust model based in the assumption that for-profit investors seek to minimize risk as part of their investment strategy, so they are especially influenced by indicators that they are more likely to be repaid, such as trustworthy, moral, and hard- working external public self-identities. In the Herzenstein et al. (2011) study, trustworthy and hardworking narratives significantly and positively affected funding, while religious marginally affected it negatively. No other factors impacted funding success. Loan rates were reduced also by trustworthy and hardworking, but by none of the other factors. However, a different cognitive process in non-profit giving is presumed in this dissertation, based in non-profit management literature predicting that, because repayment is not a consideration and because empathy, altruism, and belief in the cause are more relevant cognitive factors for donor deliberation, need, rather than trust, will be the more important consideration for donors (Bekkers and Wiepking, 2011). The prediction of the present two-phase study is that trust will not be a relevant concern for donors. In lending situations, trustworthiness is usually assumed to be of great importance to lenders. According to Mayer, Davis, and Schoorman (1995), trust is the willingness of one party to extend resources to another while at the same time assuming risk in the transaction. Risk is thus a primary element distinguishing trust from such constructs as confidence, for instance. Another aspect of lender decision-making is the effect of
  • 36. 27 narratives on lender deliberation. Herzenstein et al. (2011) find that identity claims can affect the willingness of lenders to commit resources to borrowers, in part because of what they communicate about borrower integrity, ability, and benevolence, a prediction these researchers borrow from the trust model of Mayer, Davis, and Schoorman (1995). Herzenstein and colleagues found support for this trust model by analyzing for-profit narratives on the Prosper.com P2P lending website, but their study does not attempt to determine whether trust factors also predict donations in the non-profit realm, as might be predicted if we assume donations include elements of risk to the giver. The current study explores this specific question to see if any evidence can be found linking their for-profit model to the non-profit realm or if, conversely, the two realms are different in this respect. According to the seminal study by Sargeant and Woodliffe (2007), no empirical studies have been conducted exploring the importance of variables other than trust in building loyalty or commitment in nonprofit donations. They also observed that risk of the organization losing out if they stop giving appears to prevent some people from giving in the first place. This risk of harm to the organization is in direct contrast to concerns about risk to themselves. They also found that trust does predict commitment to non-profit entities. Appel & Mara (2013), moreover, found perceived trustworthiness of characters in narratives affects persuasiveness. In their seminal study, Sargeant, Ford, and West (2006) wrote the first marketing model based on empirical measurement of giver perceptions and their impact on donations. The literature has previously measured the effects of intrinsic and extrinsic factors as a way to distinguish givers and non-givers, but
  • 37. 28 these factors don’t help explain levels of giving as much as perceptual determinants (such as trust) can. Trust and commitment are also explored by Sargeant, Ford, and West (2006) as mediators, with trust (and commitment, indirectly) appearing to relate only to the perceptions of donors about whether the beneficiaries will benefit, rather than to any trust that potential benefits will accrue to donors themselves. While these three studies indicate that trust may be a concern in non-profit donation behavior, none empirically addresses the question of whether need is also a concern in the non-profit arena. I propose to do that in the present study. I also propose to address the lack of empirical exploration of other narratives besides trust for their impact on fundraising success, by identifying other narratives and measuring their impact. Multiple Identity Claims Exemplification theory (Zillman and Brosius, 2000) and the identified victim effect (Small, Lowenstein, and Slovic, 2007) both postulate that brief stories about single individuals are more persuasive than those referring to multiple individuals. What is uncertain, however, is whether these results will obtain when multiple identities, rather than multiple individuals, are referred to in a persuasive narrative appeal. Herzenstein et al. (2011) found that increasing the number of identity claims appeared to reduce uncertainty and increase trust, but also predicted a lower tendency for repayment. They theorized there might be an implicit tendency for narrative requesters to increase number of identities mentioned when they were less desirable borrowers and the choice to do so will also predict loan default. They did not, however, explore whether this same prediction holds in the non-profit field, where we suspect need, rather than trust, is the
  • 38. 29 driving force in decision-making. This study will consequently explore the relationship between the number of narrative identities used in self-presentation and non-profit fundraising success. Time to Funding Recently, Anderson and Saxton (2016, in press) operationalized the success of online nonprofit fundraising as time to funding, and found that images of children, husbands, and items of prosperity significantly impacted the speed of online funding of individual loan requests on Kiva.com. The present two-phase study seeks to explore the relevance of their explanatory model of online non-profit loan requests to online individual donation requests as well. To do so, the present two-phase study measured the time to funding in seconds and looked for an impact of presented narrative identities upon speed to funding. Micronarratives The use of a narrative approach to focus on characterizations to the exclusion of either setting or plot is problematical, according to Riessman and Quinney (2005). They critique narrative research that does not use narrative elements fully and richly as not “good enough” narrative study, perhaps not even truly narrative research. Martens et al. (2007) also propose a list of essential narrative elements including subject, goal, forces, and implicit or explicit event sequencing which create plot. However, dogmatic prescription that narrative research must necessarily include all of its potential elements in order to be well-done is inconsistent with Georgakopoulou’s (2006, p. 6) model
  • 39. 30 supporting “late modern theorizing of the micro-, or small, unofficial, fragmented and/or non-hegemonic social practices as crucial sites of activity.” In this dissertation, therefore, based as it is on a replication and extension of the identity coding categories inductively derived by Herzenstein et al. (2011), I argue that this pure approach to narrative research is too constrictive and may even be counterproductive. Such pure narrative research, so defined, would not allow one to analyze the smallest units of narrative that can predict compliance behavior, such as that indicated by Herzenstein and colleagues’ significant results. The instances of self-identity inductively coded by Herzenstein and colleagues have utility in spite of not containing all three elements of narrative (plot, setting, and character). In fact, as Herzenstein and colleagues demonstrated, the use of several of them correlated significantly with funding success indicators. In my model, I will present these identities found by Herzenstein and colleagues, along with several of my own inductively derived from my sample, as narrative fragments we will call micronarratives. This is a term we are coining here for the first time as a means of recognizing, distinguishing, and legitimizing identity forms which we believe are narrative in nature or content, though not themselves formal or full narratives. They are, by nature, narration, although not complete narratives. They may tell merely an aspect of a story, rather than full sections, parts, or components, or an interaction of those, as is usually stipulated for a narrative. Thus, a component of characterization, or a component of a journey without characterization, are still narration in style, content, and method, but less of complete narratives. This appears consistent with Ibarra and Barbulescu’s (2010) model of identity management as well. As I describe
  • 40. 31 them, micronarratives may contain more or less coherence, given their lack of detail, but they may also contain the potential for multiple combinations of or interactions among narrative components such as plot development, setting, and agency/characterization. In order to offer more support for this choice of terminology, I offer the following justification, based in part in the narrative literature itself. Why Measure Micronarratives? When Ezzy’s seminal study (1998) first proposed that scholars integrate Mead’s (1934) Social Interactionist approach with the hermeneutic narrative approach of Ricoeur (1984), one of his intentions was to correct the tendency of some scholars to see identity as either a substance or an illusion. He recommended instead seeing identity as a process, using the relativity theory of Einstein (1920) to elaborate Mead’s statements about the temporality of identity, and explained how, “In the same way that, according to the theory of relativity, the measurements of an object depend upon the spatio-temporal framework from which it is observed, the meaning a ‘minded organism’ gives to a particular event depends on the spatio-temporal-social-interpretive framework from which it is interpreted (Ezzy, 1998, p. 241).” Similarly, I argue, micronarratives offer a chance to measure constructs in a more discrete manner than do fuller narratives. The degree of narrativity contained in an utterance or narrative unit can be determined individually, based on its spatio-temporal framework, rather than by an overt reference to all three narrative elements at once. Narrations with fewer than all of the above story elements might be usefully termed
  • 41. 32 micronarratives. They may stand in place of character development or be merely character description or they may merely imply one or more story elements. I am not the first researcher to study the discrete narrative elements that we here call micronarratives. Herzenstein et al. (2011) do so in the study I attempt to replicate and extend here, though without addressing the philosophical concerns mentioned by Riessman and Quinney (2005). Czarniawska (1998: 17) also identifies “minimal narratives,” and Bennett, Foot, and Xenos (2011) assert the value of frame analysis for simplifying more complex discourse and enabling researchers to “pick up narrative fragments that may reside outside of fully formed stories (p. 223-4).” They note, too, that narratives “often become simplified, sloganized, and cued by images that can reside outside of people (p. 223-4).” Boje (1991) also mentions, “A finding in sociolinguistic studies is that stories are brief and fragmented across extended and interrupted discourse, but this has been ignored in organization stories (p. 109).” This can occur in the same way that “researchers can ‘unpack’ very brief enactments in dialogue to discover the reality underlying the linguistic enactments (Boje, 1991, 110).” Boje (1991) also describes “terse storytelling” as the most abbreviated a story can be and still be a story. “It can be so brief that the performance is barely distinguishable from other nonperformance utterances (p. 115).” It is described as using an “abbreviated code.” At times, the terseness can be so great that there appears no story at all to the uninitiated. Boje describes cases in which single words can represent entire stories with just a prolonged, significant pause following them. Boje emphasizes that, “First, it is not
  • 42. 33 the fact that the story is terse and abbreviated that counts; it is the fact that the teller picks one aspect to abbreviate (‘You know the rest of the story,’) and another to accentuate (p. 124).” Georgakopoulou (2006) refers to small stories. Micronarratives as I have conceptualized them are probably at times smaller than small stories, though they may perhaps be situated within “small stories” and include the category of “identities,” though they are not limited to this category. Micronarratives are more specific than “repertoire” elements, in the conceptualization of the seminal Ibarra and Barbulescu (2010) model. The category of repertoire elements is more inclusive and less specific than micronarratives as it can hold unstructured story fragments, such as micronarratives, along with delivery aspects as well. Ezzy (1998) also adds more useful perspective on our micronarrative conceptualization with his description of Saas’s (1988) hermeneutics of suspicion, which “argues that self-knowledge is always only ‘approximate, tentative and indirect’ and often influenced by ‘only partially specifiable themes and backgrounds that exist at various levels of implicit and explicit awareness.’” (p. 249) Micronarratives thus can be any one or more of the essential components (plot—temporal sequence, character, setting, etc.) of an actual or potentially larger story and can include any part of, implication of, or derived meaning from actual or implied full-length stories, enacted or still-potential. Given these precedents, I propose that the term micronarrative is not only an
  • 43. 34 acceptable, but also a useful way to describe story elements that serve a narrative purpose, even if the story is only implied or inferential. Plot sequence may not be made explicit, and there may be no other elements, such as conflict, but character is crystalized; other story elements can sometimes be inferred, even when not direct or implied. In practice, it appears that the findings of both Herzenstein et al.’s (2011) study and the present one (as I will show in my chapters 4 and 5) show some reasonable connectedness between these micronarratives in real life settings and their impact on social outcomes. Even though they are fragmentary and small, these micronarratives do indeed tell at least part of a story, it appears. While Ezzy (1998) argues that temporality is integral to narrative identity, he does not mention the degree of temporality that must accrue to these accounts. I argue, therefore, that not only may the degree of temporality be flexible, it may actually be implicit, understood, or evidentially absent as well, as long as it is present by proxy in some form. This temporality or plot development may be as simple as, for instance, socially-constructed, conflict-derived personality characterizations such as those found in cultural archetypes or in binary/oppositional traits, moods, or chemical reactions. Narrative can thus include limited conceptualizations; the problem for the present two-phase research study is: how small or fragmentary can those concepts be and still be narratives? I suspect they may be as small as the smallest unit of language, provided they include sufficient identity elements, however invisible. They may therefore be fully narration, yet not be full-length narratives.
  • 44. 35 Elliott and Wattanasuwan’s (1998) discussion of identity lends power to the present micronarrative model as well. Their exploration of the importance of deep meaning, personal meaning, and trust in brand effectiveness can be used to demonstrate that even as small a symbolic element as a brand name or logo still functions in a narrative capacity to tell part or all of a story. I have argued that these characterizations we call micronarratives may exhibit a sort of implied emplotedness, in that their meanings presuppose actions by the character in the context of time. They are thus at times the recognizable symbol of an implied conflict history or plot. The meaning projected by a micronarrative is resolved into a unity that the reader perceives as a character, but which is also, as the Classical Greek origins of that word kharakter suggest, a distinctive mark—such as a facial dimple -- which in the Einsteinian sense of Reimannian space, may be plotted even in its negative space, the part not seen on the surface, but hidden in the warped folds of the experiential symbol itself. Our disinclination to explore the contextual factors of setting (along with our decision to explore emplotment only as the implied conflict measure) can also be seen as consistent with the narrative research model proposed by Green, Garst, and Brock (2004). Their model predicts that even merely implied narrative elements can affect attitude change. They allow for separating out of discrete contextual factors by measuring context-free assertions in the search for persuasion effectiveness or attitude change and call for further exploration of the boundaries of the influence of context-free narrative
  • 45. 36 implication. This allowance parallels how I frame the present research, as an analysis of the effects of implied content, over and above overt content. Archetypes Caldwell (2010) explores the identifiability of archetypes specifically derived from consumer models, rather than derived from more general human theories of motivation such as proposed by Jung (1975) and Mark and Pearson (2001). Ibarra and Barbulescu (2010) also bring up the relevance of heroic/mythic conceptualizations in their framework. In order to assess the relevance of archetypal identities on donor behavior, I searched our sample for archetypes, using the model of Mark and Pearson (2001) as our guide. My inductive analysis only pulled out one readily apparent mythic/archetypal theme, hero, which perhaps says more about the lack of their overt use than it does any significant success or failure of the Mark and Pearson (2001) model itself. Perhaps other archetypes could in future be identified by implication, rather than predominantly using specific word cues, as we largely did to identify hero. Also, merely identifying whether a term or concept has been included in a post gives little indication of how it is used. This is a concern of seminal researchers such as Hibbert and Horne (1996) as well. For instance, just because one mythic character is identified, does that mean it has been presented in a meaningful way? The limited occurrence of easily identifiable archetypes in my sample prevents me from analyzing them as fully as I would like. Relationship Between the Independent and Dependent Variables The goal of the present two-phase study is to explore the efficacy of real-world persuasion attempts, in particular persuasiveness in the online prosocial donation arena.
  • 46. 37 The persuasion attempts looked at are high stakes, real-life requests for funding from both individuals and companies, some made with high-dollar resources, some made with nearly none. These are not artificially set up laboratory experiments with little applicability to real-life persuasion, but ongoing persuasive campaigns for action. Specifically, this study explores the persuasive effects of narrative identities, as evidenced in textual funding requests in the online environment of non-profit donations. Its goal is to determine whether there are any relationships between specific independent variables (which revolve around narrative elements in text) and specific dependent variables (which measure funding success in several ways). The compliance-gaining dependent variable (DV) is broadly defined here as the successful attempt to obtain a commitment of funding from another person, group or organization. This is in keeping with the seminal study by Vaara (2002) calling for more study of the phenomena of success and failure in narrative studies, which has been understudied in the social science literature – an especially important issue, given that such narratives are an important source of empirical data for scholars. More narrowly, the most relevant success measures were determined to be: the continuous measure of overall amount funded, the binary measure of whether or not the requester’s goal was reached, the bounded measure of percentage of goal reached, and the continuous measures of time to funding and number of funders. Goals The present two-phase study seeks to expand the narrative identity concept proffered by Herzenstein et al. (2011) to the nonprofit realm using micronarrative
  • 47. 38 identities to see which narrative identities relate to higher funding, full funding, and a higher percentage of the funding goal met. It also adds to the list of narrative identities explored in previous literature by expanding Herzenstein and colleagues’ identity claim categories to include some additional identities inductively generated from its unique dataset. For example, preliminary research and data coding for the current study brought to light additional narratives such as “jobs,” “needs,” and “I’m the best fit,” although final coding eliminated or consolidated some of these categories. This two-phase study also attempts to address the relevance of trust factors in non-profit donations in the following way. It replicates Herzenstein et al.’s (2011) use of Mayer et al.’s (1995) three theoretical borrower characteristics of integrity, ability, and benevolence to structure its coding categories for trust-risk identity claims. Overall, this study attempts to determine whether the trust factor appears relevant to the non-profit fundraising world, or whether factors implying need predict success better. Based on the empirical precedents and theoretical models in the preceding studies, the two studies presented here explore the aforementioned narrative identities and several previously undeveloped ones to shed additional light on: 1) what stories people are telling about hardship, personal disaster, and personal responsibility; 2) which are more successful; and 3) how the results fit with predictions from current theory. The contribution to the communication literature that the study offers is to identify differences between compliance-gaining results in nonprofit donations and in for-profit lending within a specific theoretical framework.
  • 48. 39 This study overcomes previous limitations of compliance-gaining research by exploring the real-time interactions among actual requesters and donors in the ubiquitous posts requesting online non-profit donations, an innovation based on calls for expanding the generalizability of entrepreneurial narrative research findings to other financial contexts (Martens et al., 2007). Using automated data collection, social media resources, and statistical analysis methods, this study design also innovates by measuring actual online persuasion processes as they occur. Research Questions Given that related findings on the effects of narrative choice in self-presentation indicate that the story the words tell will significantly impact the perception of trustworthiness, it can be predicted that certain narrative frameworks or “stories” will lead to more complete funding than others. Trust has been identified as an important indicator of success in for-profit funding requests, given the risk involved for investors. It would also seem that, even though risk may not be as important a consideration for nonprofit donors, it will still be somewhat of a concern for non-profit lenders, albeit probably less so. To date, narrative research has not explored the relevance of needy identities. If, as the for-profit model from the financial research suggests, lenders are motivated differently from donors, we should see that need considerations will more strongly account for funding in contexts that are less driven by trust considerations since payback is not at stake. Therefore, we ask the research questions: “What narrative self- identities are used by request posters in the online nonprofit fundraising?” and “What
  • 49. 40 aspects of narrative will be related to non-profit funding success?” and “What will be the direction of such relationships?”
  • 50. 41 Chapter 3: Research Method Sample Selection Process Initially, the choice of potential websites suitable for gathering sufficient data on nonprofit crowdfunding success in real time was narrowed down between Gofundme.com (GFM) and Indiegogo.com (IGG). Both are highly popular online crowdfunding websites that offer the general public opportunities to donate money to either persons or organizations that can be either profit seeking or not-for-profit. A spot check of GFM showed 40 pages of nonprofits on January 23, 2013 at 9:23AM EST. There were ten posts per page. A brief check of every other page or so found three videos total. None of the pages used the standard video arrow on their thumbnails. At the time of the preliminary gather on IGG, that site also showed 40 pages of “verified nonprofits,” but with 9 entries per page. The first five categories of nonprofits were: animals, community, education, environment, and health. The overwhelming majority had embedded videos. Based on this preliminary count, there appeared to be substantially more videos on IGG than on GFM, so to maximize data collection efforts (some of which data is planned for use in future analyses), the decision was made to use IGG instead of GFM for this study. Complete data collection began February 22, 2013 and ended October 10, 2013, gathering all that were labeled Verified Non-Profit (VNP). Thus, the URL added dates in the final data were February 22, 2013 through October 10, 2013. (February to October was nearly an eight-month period.) That added up to 2764 cases, which, when all duplicates were removed, yielded 2750 non-duplicates (unique campaigns). During analysis, cases were identified by their URL and gathered date.
  • 51. 42 Cases gathered early had the longest data insertion lag. Any that did not have key data – Funding Total, End Date, etc., were deleted. In preparation for Study 1, this researcher started looking at the data set in February and gathered any 2012 campaigns that were still running. However, when ready to analyze the pages of these preexisting campaigns, the researcher would have had search for them online by URL, since the pages themselves would have been posted before our collection began. It was determined that if IGG was not reliable in keeping old campaigns, we might not have access to all those from that year. To check this, the lead researcher emailed IGG, who replied that they do not, as a rule, delete old campaigns from their server history. However, the strategy was chosen to delete any that started before Feb 22, 2013, just as a safeguard against such criticism. (That eliminated about 25% of the original data.) The data gather was completed on Oct 14th . This end gather date decision was somewhat arbitrary, based primarily on convenience, so we decided to narrow the population into an even 6 months gather, by using only the campaigns that both started and finished within the 6 month period from April 1 to September 30, which fully comprises the 2nd and 3rd quarters of 2013. The usable population was narrowed to certified nonprofit postings on IGG during the complete six months from April 1, 2013 through September 30, 2013. Thus, the final gather started in April on any campaign that started and ended in this period. (About 46% started before that.) This yielded 1247 (actual final count) campaigns from which a random sample of 200 cases was taken.
  • 52. 43 Collection Method First, preliminary data was hand-collected in near-daily grabs of online data using Python script to gather data from webpages of donation requesters on the IGG website. Originally, IGG showed several advantages over the competing website, GFM, in that the former hosts significantly more nonprofit requests and accompanying videos. Also, GFM doesn’t directly host all its videos and offers little “hint” that a video is available. When called, the workers at the GFM helpdesk said to check YouTube for their videos, since all videos must be uploaded there to work on their site. But they offered no way to easily search for those videos. Conversely, since IGG hosts videos using Vimeo, Python script could be used to mechanically gather all the videos. All elements of borrower post webpages were downloaded, including text, pictures, symbols, and videos. Later, long-term collection efforts were mechanized using electronic databases. Snapshots of webpages from the IGG crowdfunding website were collected, as were all separable components from each page, so that data could be analyzed according to the dependent variables indicating funding success: amount funded, whether or not funding goals were reached, and percentage of goal reached. Also included were data on controls, including those indicating attention gained: likes, hits, etc. Data on the independent narrative variables were also collected. Coding The present study replicates the conditions of Herzenstein et al.’s (2011) study of narrative influence in the online borrowing and lending environment in order to extend their findings about narrative effectiveness from the for-profit sector (as found on
  • 53. 44 Prosper.com) to the not-for-profit donation request environment found on IGG. The “identity claims” originally presented by Herzenstein and others were: trustworthy, successful, hardworking, economic hardship, moral, and religious. Herzenstein and colleagues inductively determined those narrative identities after reading about a third of their gathered narratives. Then, by analyzing the same narratives, two research assistants verified the list was exhaustive. Ten more research assistants coded all the data in the set, assigning a “1” to any narrative containing any version of each specific identity. Otherwise, the narrative was coded with a “0” for each absent identity. Finally two RAs read the listings, first independently, then with discussion to come up with a unified code. The present Study 1 used a similar inductive process to determine additional identities. The researcher started with a preliminary textual analysis of 100 targeted posts on IGG and categorized each instance of identity presentation. The list for this inductive search start with those specified by Herzenstein et al. (2011): trustworthy, successful, hardworking, economic hardship, moral, and religious. Preliminary analysis by the present author resulted in a few additional micro-narrative identities, including need, fun, helpless, and victim. This addition comprises a unique contribution to the literature of non-profit P2P lending analysis. (See Table 1 for actual examples of each narrative identity.) Given that the original Herzenstein et al. (2011) study did not offer an extensive description of each item in its coding system, Study 1 attempted an approximate replication of the coding itself in the following manner. Any form of the word trust was taken to infer an identity based on trustworthiness. Terms such as “innocent” were
  • 54. 45 determined to be analogous to trustworthy and so coded. Any form of the word success was a signal of an identity based on successfulness. Any use of the exact term “working hard” or “hard work,” or an extensive description of laborious effort was assigned a value of hardworking. Terms such as “less fortunate” and “financial difficulty” described aspects of economic hardship. Moral narratives include references to doing good, helping others, or community contributions. Any reference to religions, doctrines, or deities was termed a religious identity. Need identities contained any form of the word “need.” Fun narratives included “enjoy” words, “adventure” words, and any other words that expressed subjective states of apparent hormonal pleasure (e.g. “enrichment of the soul”). “Love” and “fondness” elements were coded in the fun narrative identity category, since they can be understood to describe pleasant feelings. Terms such as “homelessness” or “overwhelming odds” connote helpless. Descriptions such as “fleeing genocide” were coded as victim identities, conversely. This means that several of the identities such as need narratives are being scored according to specific word use, rather than overall meaning, which may present other semantic problems. Overall, however, this seemed the most relevant way to code some categories of narratives. Other identities were also found to exist. For instance, posts on IGG are grouped according to activities and interests such as childcare, healthcare, pet care, teaching, or environmental. Given our earlier research into the effectiveness or nonprofit requests on Kiva.org (Anderson & Saxton, 2016), which showed the relevance of babies in online fundraising appeals, we coded for child-oriented identities here as well. In addition to types of identities, the Herzenstein et al. (2011) study also found the
  • 55. 46 number of identities used correlated with an increase in the amount loaned, but also with reduced repayment. Therefore, the present study also counted how many identities were used in each posting. The number of identities was then added to the list of IVs. Structural Variables In addition to narrative identities, it is possible that other textual factors affecting funding can be measured. In this vein, for Study 2 I created a listing of additional factors we titled structural variables that included various measures of readability, length, and errata. Readability measures include reading ease and grade level; length measures included word and sentence lengths; errata measures syntax, grammar, idiom, spelling and obvious punctuation errors. In cases where grammar rules are more vague or “best judgment” is called for in online grammar sites, I coded for consistency, rather than absolute “correct” or “incorrect” values. Thus, for instance, personal titles coming after the name should not be capitalized unless all in the post are (thus, an apparent stylistic or emphatic choice) and a Perk element once titled should be titled throughout the post, particularly “Thank You” gifts. In general, improper or inconsistent hyphenation was not counted as an error. “Ands” beginning sentences were considered errors; “buts” were not. Comma errors or splices are counted. Measurement The overall dependent variable is the ability of the poster to acquire funding for the project. The time to success measure is finite, since all projects must be funded in 28 days in order to succeed. Therefore, we did not use speed to funding as the primary measure of success, although I did use it as a secondary potential measure of success.
  • 56. 47 Instead, we operationalized funding success as the continuous or binary measures of the actual final amount raised, the achievement of goal or not (binary), the percentage of goal achieved (bounded), and the number of funders who contributed. To measure the effects of our independent variables, we conducted simple linear regressions of the continuous dependent variables amount raised and number of funders, a separate logistical regression (LOGIT) with the binary variable goal reached, and a simple linear regression of the bounded variable percentage of goal achieved. The independent variables include type of micronarrative identity claims and number of narrative identity claims. These independent variables are categorical and ratio, respectively. An additional group of independent variables was added called structural variables. These are an exploration of other text factors besides narrative identities that we controlled for that we speculated might affect persuasive campaign outcomes. These include Flesch-Kincaid measures of reading ease, grade level, reading index, average grade level, word count, and sentence count, as well as two measures of proofreading errors. The latter included a continuous and a binary measure of the amount of proofreading errors. The dependent variables are mostly continuous or ratio level, particularly those indicating funding success, such as amount raised, percentage of goal reached, and funder count. Amount requested was also controlled for, since success might be determined in part by that amount, which might have skewed our results.
  • 57. 48 In order to choose the control variables, correlation tests were run to see which potential controls correlated with both the micro narratives and the dependent variables and thus could be possible confounds. In the first model, the linear regression with DV amount raised, seven potential confounds were found to be correlated with both the DV and with at least one of either the micro narratives or the narrative structures. These were updates, comments, funders, perk levels, Twitter count, and Facebook count and were therefore controlled for. In the second model, the logistical regression with the DV success, four possible confounds were isolated that correlated with both the DV and with at least one of either the micro narratives or the narrative structures and were therefore controlled for. These were campaign length, comments, funders, and Facebook count. In the third model, the linear regression with the DV Percent of Goal Reached, the same four possible confounds were isolated as in the previous model (campaign length, comments, funders, and Facebook count) and were likewise controlled for. As a way of empirically testing the hypotheses regarding the related, but differently measured outcomes denoting funding success, two different types of models were employed. First, a multiple regression was used to capture the specific effects of micro narratives and narrative structures on the amount raised, a continuous measure between zero and upward infinity. Next, a second multiple regression was used to capture the specific effects of these micro narratives and narrative structures on percent of goal raised, a bounded variable with the limited range between 0 and 1. Third, a logistical regression model (LOGIT) was used to capture the specific effects of these micro narratives and narrative structures on goal reached, a binary measure of whether or not a
  • 58. 49 campaign achieved its donation goal. The LOGIT model accommodates an otherwise non-linear spread of data by transforming it to logs, which allow the plotting of a relatively linear regression. Fourth, two simple OLS regressions were run using number of narrative identities as predictor variables against amount raised and percentage of goal raised. An additional LOGIT regression was run using the number of narrative identities to predict goal reached. Because another possible indication of the success of a text persuasion appeal in the present context might also be the number of funders who responded to the appeal, we decided to also explore any correlation between the number of funders and each specific narrative identity and cluster of identities.
  • 59. 50 Chapter 4: Study 1 (Inductive Coding) In study 1, I conduct in-depth inductive analyses to identify the types of self- identity narratives nonprofit entrepreneurs employ in their crowdfunding requests. The method of data assignment into categories as used in Study 1 is a qualitative approach. According to Thomas (2006, p. 238), the method of inductive coding refers to “approaches that primarily use detailed readings of raw data to derive concepts, themes, or a model through interpretations made from the raw data by an evaluator or researcher.” The particular method of coding used in this study is an inductive assessment of specific occurrences in the data set, leading to assignment of instances in separate categories of identity narratives. To accomplish this, the researcher conducted a detailed analysis of text in the sampled funding requests on IGG, deriving concepts, themes, or terminology which indicated a thematic direction to the identity claimed, either overtly or subtly, allowing it to be assigned a binary code as present or not. The inductive method is consistent with the grounded theory of Strauss and Corbin (1998) who describe this method as beginning with a study area and coaxing the theory out of the data, rather than beginning with a theory and looking for data which fit it (Thomas, 2006). Replicated Codes Given the goal of this study to replicate some aspects of Herzenstein et al.’s (2011) research findings that specific narrative identities predicted funding success in for- profit online fundraising, I first developed coding schema for each of these previous variables. Since the analysis by Herzenstein and colleagues did not offer a detailed explanation of how instances of text were assigned to specific coding categories, I
  • 60. 51 developed a coding scheme to identify these previous codes, as well as new ones that were conceived during the textual analysis. To do this, the five basic procedures of inductive coding described by Thomas (2006) were followed: 1) Raw Data Preparation; 2) Text Reading; 3) Category Creation; 4) Text Overlapping; and 5) Category Revision and Refinement. Raw Data Preparation The raw data files were first collected and saved to a relational database using Python programming language script. Thus, all raw data script was formatted in a standard font and size and saved in table form, with rows comprising each separate webpage and columns comprising each category of webpage elements. The latter elements include not only textual items, but also images, sound files, URLs and other background items tangential to the current study. With an eye toward future analysis, backups of all original files were saved in various locations. Text Reading Once the textual elements were collected in the database, a close reading of the material was performed to gain familiarity with the content and understand the themes and events it contained. Category Creation Once an understanding of the basic elements was obtained, this researcher began to identify and create categories and themes. Initially, the evaluation group of six categories from Herzenstein et al. (2011) was identified as existing in at least some of the posts. Additional categories were derived after repeated readings of the text. The
  • 61. 52 categories were generally created from instances of phrasing or apparent meanings in actual segments of text. No analytical computer software was used for the actual coding or creating of categories. Instead, the primary coder individually determined all new categories. Uncoded or Overlapping Text Given the qualitative nature of the inductive coding process, some text may contain multiple themes, while other text may contain none of relevance to the analysis at hand. This was the case in the present dataset. Thus, some text contained overlapping themes, while other text contained no coded themes. Category Revision and Refinement Throughout the coding process, category names, descriptions, and content were changed, revised, and refined until the final coding categories were finalized. For example, the category titled hip at one point was conceived as youthful but ultimately renamed as hip to help distinguish it from other categories that contain language focusing on children and youth. In addition, the kinds of hip language elements that were found are not necessarily the exclusive domain of young people. This author can think of several late-night talk show hosts who use several of the phrases, even though they are well past middle-aged. Such phrases include “make some noise for,” and iterations of “awesome,” and “cool,” for example. For such reasons, it was determined to rename the category hip instead. In this way, category names were progressively analyzed, re- considered, and changed when needed.