|USFDC Home | USF Electronic Theses and Dissertations||| RSS|
This item is only available as the following downloads:
xml version 1.0 encoding UTF-8 standalone no
record xmlns http:www.loc.govMARC21slim
leader nam 22 Ka 4500
controlfield tag 007 cr-bnu---uuuuu
008 s2010 flu s 000 0 eng d
datafield ind1 8 ind2 024
subfield code a E14-SFE0004730
Lai, Jennie Wan Man.
Utilizing facebook application for disaster relief :
b social network analysis of american red cross cause joiners
h [electronic resource] /
by Jennie Wan Man Lai.
[Tampa, Fla] :
University of South Florida,
Title from PDF of title page.
Document formatted into pages; contains X pages.
Thesis (MA)--University of South Florida, 2010.
Includes bibliographical references.
Text (Electronic thesis) in PDF format.
Mode of access: World Wide Web.
System requirements: World Wide Web browser and PDF reader.
ABSTRACT: With the exponential growth of Facebook users worldwide, this platform for social network online has become a powerful tool to connect individuals and share information with each other. This study explores the phenomenal trend of utilizing a Facebook application called Causes to help users organize into online communities for a specific cause and mobilize their resources for disaster relief during the Haiti earthquake disaster. Two separate samples of 100 joiners each from the American Red Cross (ARC) Cause on Facebook were randomly selected before and after the Haiti earthquake disaster to examine the differences of the composition (i.e., attributes) and structure (i.e., relational ties) of each social network. The social network analysis performed for this thesis research intends to fill the gap of historical research literature on recruitment to activism and support provision following a disaster in the digital age of the 21st century. The results of this study show how understanding the membership size of online communities, salient identity for the cause through organizational affiliations, interpersonal ties among the joiners, density of the network as well as gender diversity can be crucial recruitment factors to leverage for disaster relief efforts. The findings reveal a beneficial partnership between disaster relief organizations and online social networks in mobilizing their resources for a speedy response to disasters.
Advisor: John Skvoretz, Ph.D.
American Red Cross
Social network analysis
t USF Electronic Theses and Dissertations.
Utiliz ing Facebook Application for Disaster Relief : Social Network Analysis of American Red Cross Cause Joiners by Jennie Wan Man Lai A thesis submitted in partial fulfillment of the requirements for the degree of Master of Arts Department of Sociology College of Arts and Sciences University of South Florida Major Professor: John Skvoretz, Ph.D. James C. Cavendish, Ph.D. Michael Kleiman, Ph.D. Date of Approval: August 26 2010 Keywords: American Red Cross, disaster relief Facebook Causes social movement, social network analysis Copyright 2010, Jennie W. Lai
D edication I dedicate my thesis to the memories of my beloved grandmother, Wong Mu Chun, and my dear friend, Wong Ching Man. I am utmost indebted to my grandmother for her selfless sacrifice to afford me the better opport unities in life my life has been infinitely enriched by her unconditional love and unwavering support Thank you, mah mah. I am also deeply grateful for my long standing friendship with Ching Man she will always be my anchor for resilience through the challenging times. I miss you, my dear friend. The completion of my thesis would not have been possible without the ongoing reminders of their strength and endurance
A cknowledgements First and foremost, I am deeply grateful to the chair of my thesis committee, Dr. John Skvoretz, for his guidance, patience and support throughout the entire research process (especially during the summer semesters) I greatly value each of our thesis discussion s to learn f rom his thinking and approaches for analysis. I am also very appreciative to Dr. James Cavendish and Dr. Michael Kleiman for their thought ful insights in shaping my thesis research. Additionally, I would like to thank my network of support though I cannot acknowledge them all individually but I want to express my sincere gratitude especially to Ana, Catherine, Chuck, Philia and Wendy. They have each played a very special role through this long journey in helping me reach the finish line. Last but not least, I am utmost thankful to Dr. Paul J. Lavrakas for inspiring me to pursue my gr aduate degree in Soc iology early in my career. His ongoing advice throughout this journey has been invaluable to my continual growth personally and professionally.
i Table of Contents List of Tables .................................................................................................................. iii List of Figures ................................................................................................................. iv Abstract .......................................................................................................................... vi Chapter 1: Introduction ................................................................................................... 1 Chapter 2: Literature Review and Background ............................................................... 4 Recruitment for Social Movements ..................................................................... 4 Disaster Community and Support Provision .......................................................13 Facebook ...........................................................................................................19 Causes ..............................................................................................................22 Social Network Analysis .....................................................................................25 OneMode Network ............................................................................................29 Two Mode Network ............................................................................................30 Chapter 3: Data and Methods ........................................................................................ 33 Research Questions ..........................................................................................33 Research Methods .............................................................................................33 Ethics for Online Social Research ......................................................................35 Chapter 4: ARC Network Composition Before and After Haiti Earthquake Disaster .......37 Trend of Joining the ARC Cause ........................................................................37 Or ganizational Affiliations of ARC Joiners ..........................................................38 Identifiable Recruitment Community ..................................................................42 Gender Diversity ................................................................................................44 Chapter 5: TwoMode Network of Sampled ARC Joiners and Their Facebook Causes 46 Chapter 6: One Mode Network of Sampled ARC Joiners and Their Facebook Causes 51 OneMode Network of Sampled ARC Cause Joiners .........................................53 One Mode Network of Facebook Causes (of Sampled ARC Joiners) .................59 Chapter 7: Conclusion ...................................................................................................65 Network Size Does Matter .................................................................................66 Linkage Between Identity and Action ................................................................. 66 Mobilizing Women Online for Support Provision.................................................67 Stronger Effect of Interpersonal Ties for Recruitment .........................................67 News Travels Faster in a Dense Network ..........................................................68
ii Limitations and Future Research .......................................................................69 References .................................................................................................................... 71 A bout the Author Page..END PAGE
iii L ist of Tables Table 1: TwoMode Network Matrix of ARC Joiners and Facebook Causes Before Disaster ..................................................................................................48 Table 2: OneMode Network Matrix of ARC Joiners Before Disaster .............................54 Table 3: Relational Ties of ARC Joiners Before and After Disaster ................................57 Table 4: Relational Density of ARC Joiners Before and After Disaster ...........................58 Table 5: OneMode Network Matrix of Facebook Causes Before Disaster ....................60 Table 6: Relational Ties of Facebook Causes Before and After Disaster .......................62 Table 7: Relational D ensity of Facebook Causes Before and After Disaster ..................64
iv L ist of Figures Figure 1: User Profile Page for ARC Cause on Facebook .............................................22 Figure 2: Homepage for ARC Cause on Facebook Cause Bulletin ..............................24 Figure 3: Homepage for ARC Cause on Facebook Fundraising Activity ......................24 Figure 4: Homepage for ARC Cause on Facebook Recent Cause Activity ...................25 Figure 5: Krackhardt's OneMode Network Matrix for Advice Relation ...........................3 0 Figure 6: DGG's Two Mode Network of Women by Event Matrix ..................................32 Figure 7: DGG's Two Mode Network of Southern Women and Social Events ...............32 Figure 8: Sample of Data Format for Actors and Events ................................................35 Figure 9: Number of Facebook Members Joining the ARC Cause ................................. 38 Figure 10: Average Number of Facebook Causes for ARC Joiners ...............................40 Figure 11: Distribution of Facebook C auses for ARC Joiners ........................................40 Figure 12: Average Membership Size of Facebook Causes ..........................................41 Figure 13: Proportion of Types of Facebook Causes for ARC Joiners Before Disaster .............................................................................................................43 Figure 14: Proportion of Types of Facebook Causes for ARC Joiners After Disaster .............................................................................................................44 Figure 15: Gender Diversity of ARC Joiners ..................................................................45 Figure 16: TwoMode Network Graph of ARC Joiners and Facebook Causes Before Disaster ..................................................................................................49 Figure 17: TwoMode Network Graph of ARC Joiners and Facebook Causes After Disaster .....................................................................................................49 Figure 18: One Mode Network Graph of Relational Ties of ARC Joiners Before Disaster .............................................................................................................55
v Figure 19: One Mode Network Graph of Relational Ties of ARC Joiners After Disaster .............................................................................................................55 Figure 20: One Mode Network Graph of Relational Ties of Facebook Causes Before Disaster ..................................................................................................61 Figure 21: One Mode Network Graph of Relational Ties of Facebook Causes After Disaster .....................................................................................................61
vi Utilizing Facebook Application for Disaster Relief: Social Network Analysis of American Red Cross Cause Joiners Jennie Lai A bstract With the exponential growth of Facebook users worldwide, this platform for social network online has become a powerful tool to connect individuals and share information with each other. This study explores the phenomenal trend of utilizing a Facebook application called Causes to help users organize into online communities for a specific cause and mobilize their resources for disaster relief during the Haiti earthquake disaster. Two separate samples of 100 joiners each from the American Red Cross (ARC) Cause on Facebook were randomly selected before and after the Haiti earthquake disaster to examine the differences of the composition (i.e., attributes) and structure (i.e., relational ties) of each social network. The social network analysis performed for this thesis research intends to fill the gap of historical research literature on recruitment to activism and support provision following a disaster in the digital age of the 21st century. The results of this study show how understanding the membership size of online communities salien t identity for the cause through organizational affiliations in terpersonal ties among the joiners, density of the network as well as gender diversity can be crucial recruitment factors to leverage for disaster relief effort s. The findings reveal a beneficial partnership between disaster relief organizations and online social networks in mobilizing their resources for a speedy response to disaster s
1 Chapter 1 : Introduction In recent years, the Internet has offered researchers with access to massive samples, instantaneous communication, expedient data collection, 24 hour accessibility, lower cost for research, automatic categorizing of survey data, etc. in studying an organic community outside of the traditional (physical) environment (Christians & Chen, 2004) These advantages of data collection and processing in the digital age afford social researchers with limited resources to study the growing online communities. It also presents a unique opportunity for social researchers to study how popular social networks on the Internet, such as Facebook can mobilize their users for disaster relief effort in an online environment According to a news report by CNN (2010), U.S. relief organizations raised a record breaking $ 1.3 billion in disaster relief funds within a span of six months following the January 12, 2010 earthquake in Haiti. In fact, the American Red Cross (ARC) alone raised $468 million through conventional fundraising events as well as innovative approach eses through the Internet and mobile text messages. It is undeniable the emergence of the new media or mass media on such digital platform s as the World Wide Web, electronic mail, chatrooms, Webbased publications, etc. introduces a new form of interaction and str ucture for online social networks Consequently, these online social networks can be leveraged to mobilize their members for disaster relief effort s and the formation of disaster communit ies made up of victims and nonvictims working towards a specific ca use
2 Even though online social research has continued to gain momentum in studying online communities only a very limited number of studies focus specifically on disaster communities on the Internet As a result, researchers are not fully aware of how online social network ing Web sites can be used to recruit members for social movements or to mobilize members for disaster relief. The emerging opportunities to study these online disaster communities should also generate interest i n the factors that predi ct recruitment to online disaster relief efforts (e. g. personal attitudinal affinity, referral from friends or organizations they belong to, and relational ties ). Moreover, other analytical techniques such as network analysis should also be considered in order to understand the structural properties of the network (i.e., how connected individuals are within the community they join through interpersonal or organizational ties) and the composition of the network ( i.e., characteristics of the us ers or the organizations they belong to ) The Haiti earthquake disaster on January 12, 2010 caught worldwide attention with the magnitude of earthquakes from 6.5 to 7.3 and a death toll of over 230,000 people as estimated by the Haitian government (Associated Press, 2010) The devastation of earthquake in Haiti generated an outpouring of support for the disaster victims through both conventional and innovative means of fundraising. The advancement of technology offered the general public instantaneous ways to make donations online or by text from their mobile phones which we re the two methods ARC used to collect their donations for the Haiti disaster relief effort. Online s ocial networks such as Facebook and Twitter also played a critical role in raising awareness of fundraising and disaster relief. In this thesis, I plan to draw on the literatures that have examined recruitment to social movements on the one hand, and social tie activation in a disaster context on the other, to guide my resear ch on the formation of online disaster communities. I select
3 the ARC Joiners o f Facebook s Causes application for my research given its versatility to organize a community on the Facebook platform. The sample is limited to a random selection of joiners to the ARC Cause during a specific time period. Two separate samples of the ARC Joiners were drawn: one before the January 12th earthquake in Haiti, and the other after In each case, the joiners were randomly selected to study the structural and compositional differences of the two networks as defined by relational ties between the sampled joiners as well as the other causes to which these sampled joiners belong. Chapter 2 begin s by revie wing past research literatures on recruitment to social movement s and on support provision from the disaster community After summarizing these literatures, I provide a background overview of Facebook including the Causes application that is used to organize the online communities I n the last sections of Chapter 2, I describe the analytical technique of social network analysis for one mode and twomode networks as well as the sample selection process of the Facebook members who join the ARC Cause on Facebook Chapter 3 uses data collected from 1) members Facebook profiles to describe the characteristics of these samples and 2) their profiles on Causes to capture all the causes they belonged to. Finally, I present the results of the social network analysis of the two samples before and after disaster In Chapter 4, I focus on their network composition and in Chapter 5 I focus on the two mode network structure of ARC Joiners and their Facebook Causes In Chapter 6, I examine the one mode network s tructure of ARC Joiners and their Facebook Causes. Finally, I will discuss the implications of the findings on how they can potentially help ARC or other disaster relief organizations to tailor their recruitment effort s for disaster preparation or relief I also comment on the studys limitations and highlight areas in which further research is needed o n online disaster communities
4 Chapter 2 : Literature Review and Background What factors motivates individuals to join or participate in specific causes or social movements ? Is it their own ideological beliefs or their structural position for activism (e.g., education, gender income, etc. ) ? Perhaps it is due to influence s from other indivi duals or the types of organizations they belong to? If there are strong causal influences that lead to activism, i s there a model of recruitment factors that can predict participation in social movements ? The insights in to th ese questions can offer emergency response organizations, such as the American Red Cross assistance in targeting their effort in encourag ing activism for disaster preparation (in the event of natural disaster when advance warning is possible) and aiding disaster victi ms in their recovery process. Past research on social movement analysis as well as disaster research guide the direction on what historically influenced individuals to participate in social movement or disaster relief (specifically interpersonal or organizational ties) and how such findings may have evolved in the age of online social networks Recruitment for Social Movements One of the most examined research topics o n social movement s is differential recruitment, i.e., why do some individuals get involved while others are inactive? Some of the individual attributes cited as most correlated to activism are strong attitudinal affinity with the goal of the movement and/ or set of grievances consistent with the move ment ideology. However, McAdam (1986) indicated the study of social movement s in the recent decades pointed to structural availability as being more important than attitudinal affinity in encouraging activism. In fact, McAdam further examined the argumen ts for the structural perspective by exploring the types of prior contact that
5 could make a difference with activism ; f or example, he examined the role of informal friendship networks compared to formal organizational affiliations and the effect of weak ties on recruitment compared to strong ties in his research on the recruitment to high risk activism of the 1964 Mississippi Freedom Summer project. McAdam criticized movement scholars for their tend ency to oversimplify the recruit ment process and movement participation by not clearly distinguishing the differences between participants / nonparticipants activism / inactivism and members / non members of social movements. While it can be challenging to define the aforementioned differences he argued that it is possible to distinguish between low risk/cost and high risk/cost forms of activism at the minimum. The reference to cost is related to the expenditure of time, money and energy devoted to activism ; for example, signing a petition is considered a relatively low cost activity while volunteering in a movement for the homeless is a high cost activity in terms of time and energy. Also, the reference to risk refers to anticipated danger such as legal, social, physical, financial, and political risks. In such case, signing a petition could potentially be high risk during the McCarthyism era but volunteering in a movement for the homeless would be relatively risk free. McAdam also pointed out that prior contact with a recruiting agent can be a significant factor in low risk/cost activism the prior contact with a recruiting agent refers to the networks, relationships or communities that pull the individual into activism. There were three different ag ents he identified for the recruitment process: 1) organizations that serve as the associational network for a new movement, 2) merger of existing groups of established organizations known as bloc recruitment and 3) individual activists as the single act ivist who introduces the recruit to the movement. An example to illustrate how a prior contact with a recruiting agent pull s the individual in to
6 low risk/cost activism is the idea that a potential recruit is more likely to join a rally with a friends urg ing in order to not disappoint or lose the friends respect given the low cost/risk of the rally. However, the biographical availability which refers to the personal constraint s such as full time employment, marriage and family obligations may increas e the costs and risks of the movement participation using the same example of the aforementioned rally. On the other hand, the correlates of high risk/cost activism include the history for activism, deep commit ment to the ideology / goals of the movement, integrat ion into the activist networks and freedom from personal constraint s that would make participation risky. For the purpose of his research on recruitment to highrisk activism, McAdam used the 1964 Mississippi Freedom Summer project. This campaign was clearly a high risk/cost form of activism for hundreds of mostly White college students from the North to help with the civil rights movement throughout the South during a violent time when there were instances of kidnapping and murder of volunteers by renegade segregationists. Moreover, these students were asked to volunteer for two months of their summer without pay Prior to joining the movement, these prospective volunteers were asked to complete a detailed application on their organizational affi liations, college activities, reasons for volunteering and record of any previous arrests. These applicants were then categorized into three groups: 1) rejects, 2) participants (accepted to the project and participated) and 3) nonparticipants (accepted t o the project but later withdrew). In total, there were 1,068 applicants that included 55 rejects, 720 participants, 239 nonparticipants along with 54 with unknown status. McAdam used the data collected from these applications to compare the characteris tics of the participants and nonparticipants of the high risk/cost activism of the Freedom Summer project.
7 McAdam studied how the various types of prior contacts may differentiate the participants from the non participants. The participants belonged to m ore organizations than non participants : specifically, participants belonged to an average of 2.4 organizations compared to an average of 1.9 organizations for non participants. T he participants were also members of more explicitly political organizations than nonparticipants 50% of participants were affiliated with civil rights organizations as compared to 40% of non participants. McAdam also studied the interpersonal ties between applicants since they were asked to provide a list of at least ten people the applicants wished to keep informed of their summer activities (as part of the public relations effort s of the project). In this exercise, the participants provided many more names of other participants or known activists than nonparticipants the listing of names of other participants was twice as long and the names of the other known activists were three times as long from the participants compared to the number of names provided by the non participants. Finally, McAdam examined the biographical availability to measure the relative costs and risks associated with participation, and the findings did not support the hypothesis that those with less time or more personal responsibilities were less likely to participate. Three constraint s of marital status, employment and education were analyzed for their relationship to participation but being married and employed full time actually enhanced participation; and education was the only variable with negative effect on participation as antici pated the 1964 graduates were much more likely to participate in the project than the graduate students at the time. When McAdam assessed the combination of various factors on the likelihood for the applicants to participate in the project, there were f our important factors that drove participation: 1) attitudinal affinity, 2)
8 integration into activist networks, 3) prior history of activism and 4) absence of personal constraints on participation. In summary, McAdams study on the recruitment factors of high risk /cost activism showed participants scored higher on both organizational and interpersonal measures of integration into activist network than nonparticipants. It can be interpreted from the results that while attitudinal affinity for the movement and biographical constraint are important factors to consider but the extent of the structural pull from prior history of activism and integration into the supportive networks best accounted for participation in the Freedom Summer project. In other words, ideological identification of the movement can lead the individual in the direction of participation and prior history of activism along with integration into the supportive networks can pull the individual into participation. In 1988, Fernandez and McAdam teamed up and narrowed their research on the structural factors for recruitment to social movement s specifically on the network contact with recruitment agents that pull the individuals into participation. T he focus on network analysis became prominent for the study of recruitment to social movement as increasing evidence indicated the importance of structural factors such as the ties among the social movement organizations to mobilize resource s In fact, ties formed through overlapping membership among these organizations known as multiorganizational fields became the focus of the research for Fernandez and McAdam on the recruitment for the Freedom Summer project. Rather than studying the role of the number of social movement organizations in wh ich the recruits we re involved, they focused on the effects of the pattern of overlapping memberships in organizations on recruitment at two particular universities (i.e., University of California at Berkeley and the University of Wisconsin at Madison) tha t participated in the Freedom Summer project.
9 The activists at University of California at Berkeley (UCB) had a long history of civil rights activism, and the students recruited for the Freedom Summer project mostly came from the activist subculture. On the other hand, the University of Wisconsin at Madison (UWM) had no presence of any major civil rights organization on campus the activist community was small and not well organized. Indeed, Fernandez and McAdam had purposely selected Berkeley and Madis on because of t he sharp difference between the two colleges activist cultures The depende nt variable of the analysis wa s participation vs non participation Berkeley had 31 participants and 9 nonparticipants ; and Madison had 10 participants and 13 nonparticipants. The analysis for predicting activism included two steps: 1) the use of a prominence measure (which is explained further in the next page) to examine the potential effect of structural positions of individuals within multiorganizational fields; then 2) the study of independent effects of structural positions on individual activism such as parental income, years of education, gender, past level of civil rights movement and major area of study. The method of network analysis was also use d to study the multiorganizational fields for applicants to the Freedom Summer project from these two universities. The data were limited to the 40 applicants from Berkeley and 23 applicants from Madison to examine a small group network process in social movement. Fernandez and McAdam used the organizational affiliations listed on the application to create two peopleby organizational affiliations matrices with 23 people by 17 organizations for Madison and 40 people by 36 organizations for Berkeley. Also peopleby people matrices were created to analyze overlapping organizational affiliations the entries were the number of organizations each pair of individuals shared These interpersonal ties can likely affect an individual decision to participate online for social support. Among the Berkeley a pplicants, the density of the organizational overlap wa s .166 among participants and
10 .055 among non participants Among the Madison applicants, the density of the organizational overlap wa s .200 among participants and .026 among non participants. D ensity refers to the number of ties among the individuals as a proportion of the number of possible ties. The network among non participants from both schools was less dense than the network among participants which demonstrated that network factors were related to participation in the Freedom Summer project amongst applicants. In addition to measuring density, network prominence was also measured to describe each individuals position in relation to other individuals in the Berkeley or Madison network. P rominence refers to a measure of centrality which distinguishes among individuals who appear to be equally central (i.e., who are tied to the same number of others in the network) on the basis of t he centrality of individuals to whom they are tied. S haring the same number of organizational affiliations is not as relevant as being linked to more centrally located individuals in the network of overlap because more central individuals are more likely to experience social influence, and therefore, are more committed to recruit others. There was a tendency for more prominent individuals in the multiorganizational fields to be participants in the Freedom Summer project for both networks in Berkeley and M adison. Finally, t he results of the independent effects of structural positions on individual activism showed the pattern of higher family income, higher mean level of past activism and male applicants tend ed to yield higher overall rate of participation. The recruitment model on using structural positions within the multiorganizational field to predict participation in the Freedom Summer project has important implications : the effects of structural or independent variables on participation is dependent on the overall recruitment context, and the specifics involved between structural positions and
11 individual level background leading to activism can be complex in developing a precise recruitment model. McAdam and Paulsen followed up in 1993 with another research study to specify the relationship between social ties and activism given the increasing attention to social movement literature on the role of social ties in movement recruitment. They identified three sources of theoretical and empirical imprecision from the reported findings by past literatures: 1) the lack of theory in explaining the effects of social ties on activism; 2) the lack of understanding of how and why the social ties are important; and 3) the lack of acknowledgement to the extent of organizational or associational network or individual relationship that can lead social pressure in joining a movement. As a result, a recruitment model focused on interpersonal ties and membership in organizations was developed to determine the causal effects on the decision to participate in a movement. Interpersonal ties refer to knowing someone already involved in social movement activity; and membership in organizations r efer s to an extension of interpersonal ties through meeting people and being exposed to subsequent opportunit ies to be pulled into social movement activit ies McAdam and Paulsen examined these social ties in movement recruitment by studying the Freedom Sum mer project. In addition to using data collected from the detailed applications completed by the interested applicants about their organizational affiliations, past civil rights activities and reasons for volunteering, more information was also collected from a follow up survey on their alma mater, parents address, school major and other such information. A total of 212 participants and 118 non participants completed the follow up survey and another 40 participants and 40 nonparticipants took part in t he indepth interviews for the research study.
12 The first area of analysis on the relationship between social ties and movement recruitment was the support applicants received for participation from parents, friends, civil rights organizations, other volunt eers (i.e., participants and non participants) and religious groups. The dependent variable was the participation of the Freedom Summer project. The independent variables were the multiple ties that may influence an individual to participate in movement as listed above. The persons reported directly on the application as individuals who positively influenced your decision to apply to the Freedom Summer project were coded as strong ties and those not reported by the individuals but linked to them by way of a strong tie were considered the weak ties. The results showed that the rate of participants reporting support from their parents were doubled that of nonparticipants, and support of strong ties from other volunteers was 75% greater than non participants. The logistic regression predicting participation of the Freedom Summer project also showed, in the aggregate, participants did receive greater support from their parents and peers. It was also concluded that the strength of a social tie play ed a significant role as a predictor of activism. The second area of analysis was the salience of the social tie to assess its impact on volunteers decision to participate in the project. The openended question of why they would like to work in Mississippi this summer were coded into five categories for organizational context the teachers, religious communities, socialists/leftists, liberal democrats and civil rights movements along with a no discernable group category. Over four fifths of those with an identifiable recruitment community participated while only threefifths of those without an identifiable recruitment community participated. The strong causal influence indicated a high degree of salience for the identification with the recruitment communities, and subsequently these communities provided strong support to link their identity with participation of the Freedom Summe r project.
13 McAdam and Paulsen concluded that the recruitment model for an individual to participate depended on four factors: 1) the recruitment attempt that led to their application to participate in the project; 2) the linkage between the movement and id entity, i.e., their willingness to apply due to some salient identity to which the recruitment community they belonged; 3) the support of the linkage by others who can sustain the identity and further set s participants apart from the nonparticipants; and 4) the absence of strong opposition for the identity. Their findings also showed a stronger effect of organizational ties rather than individual t ies to pull into collective action. The individual ties w ere nonetheless important but it was more influential in an organizational context for participation. Disaster Community and Support Provision Contrary to popular belief held by the general public that disasters often lead to looting, social disorganization and deviant behaviors, the communities impacted by the disaster actually often suspend previous conflicts and unite in providing mutual support to one another. Tierney (2007:510) best summarizes this insight in her extensive summary of the sociology of disaster literature: Research accounts emphasized that disasters generate broad consensus regarding the value of life, property, and community and that affected populations are invariably more generous and helpful than during nondisaster times. The mass media has played a significant role in promoting disaster myths of heightened social conflicts through reports of exaggerated severity of looting and lawlessness as was widely reported in the aftermath of the Hurricane Katrina disaster (Tierney et al., 2006; Voorhees et al., 2007). T he emerging platform of social networks on the Internet has offer ed an outlet for individuals impacted by the disastrous event to join in social cohesion to dispute the
14 inflated reports of social disorganization. In studying the wealth of literature on disaster research that focused on communities, Kirschenbaum (2004) incorporated the various elements of these communities and expanded the definition of a disaster community beyond the physical and geographic boundaries within which a disaster occur red Kirschenbaum emphasized th at concept of a disaster community is first and foremost a social community made up of networks that were directly or indirectly affected by the disaster rather than only a physical event Moreover, t he Internet has expanded the geographic boundaries of such disaster communities to beyond the actual physical site of the disaster The fundamental characteristic of a disaster community is the social cohesion formed through the feelings of dependency, trust and support of one another for physical and emotional safety which would intensify with the threat of disasters. More importantly, disaster can draw in people who are not directly or physically affected by the actual disaster into the dis aster community, i.e., uniting both victims and nonvictims. Erikson (1994:235) suggested that trauma can be a source that creates a community based on spiritual kinship and a sense of identity. A stage of euphoria can ensue after a disastrous event which binds unconnected persons together to develop a form of fellowship Erikson wrote : The energy with which rescuers work and the warmth with which neighbors respond act to reassure victims that there is still life among the wreckage, and they react with an outpouring of communal feeling, an urgent need to make contact with and even touch others by way of renewing old pledges of fellowship. Moreover, Nilsons (1985) description of such therapeutic or altruistic communities includes disaster victi ms providing emotional support to each other that can last from a few short days to a few years depending on the remnants of the experience. During the recovery period, a disaster community can offer emotional and physical support to one
15 another based on a network of victims or possibly nonvictims volunteering their time and resources for emergency relief and to help the victims return to pre disaster conditions quickly Haines et al. (1996) explored a model of support provision in the natural disaster context for 1) personal characteristics of the provider 2) characteristics of their personal networks and 3) characteristics of the community in which they live these factors are based on early findings found to affect provision and reception of social support. In fact, disaster victims tended to become resources in providing support to other victims rather than become helpless and dependent themselves The model was tested using data collected on the preparation and recovery phase of the Hurricane Andrew disaster in order to gain an understanding of the determinants of helping behavior during natural disasters. The data of this study were collected specifically from the two Louisiana parishes t hat were severely impacted when Hurricane Andrew struck the Gulf Coast in August of 1992. The substantial damage to housing, businesses and personal/economic injury in these two parishes totaled nearly $40 million. However, s ince there was minimal disruption to the telephone services, the data for this study were collected between October to December of 1992 by telephone with 594 respondents. The information collected include network data (i.e., asking the respondents to name up to five individuals with whom they discussed important matters in the past six months prior to Hurricane Andrew), personal characteristics of the providers (i.e., gender, education, etc.) and characteristics with the respondents (i.e., closeness of the tie). Additionally, Census data of the two parishes and the block groups of areas surrounding the towns were used as measures for community context.
16 The four measures of social support included: 1) whether respondents provided material aid during the preparation phase ; 2) the number of individuals they assisted; 3) whether respondents provided recovery assistance during the short term recovery phase ; and 4) the number of individuals they assisted. For the prediction of support provision during the preparation phase of Hurricane Andrew, only one provider characteristic i.e., age was significant the older individuals were less likely to have provided support than younger individuals. T here was no significance for any of the personal network variables The length of residence was a significant variable for community context characteristics the greater the length of residence, the greater the probability that support would be provided. For the prediction of the number of people supported during the preparation phase family income had a significant positive effect that financial resources allowed for support provision. When examining support provision during the short term recovery phase of Hurricane Andrew, age had the same effect as was noted in the preparation phase and house damage had a negative effect on providing support. There was one significant variable for personal network characteristics (i.e., residential proximity) and two significant variables for community context (i.e., trust in the local government and owner occupancy in the area). Consistent with past research on disaster community, gender had a significant positive effect with men providing more support than women in the recovery phase (though not found to be significant in the preparation phase) Furthermore, individuals in gender diverse network have greater access to support provision resources Also, other characteristics for community context such as membership in fraternal organizations, service organizations or other organizations had positive effect on the number of individuals helped.
17 Haines et al. found three sets of characteristics i.e., personal characteristics of providers, personal networks and community context affected support provided in the preparation and short term reco very phase s of the Hurricane Andrew disaster and especially on the timing when such social support was provided. Past disaster research found support was most prominent during short term recovery when presence of stress in terms of index of need for support is greatest. However, the determinants for longterm recovery phase may vary depending on the level of resources depleted during the short term recovery phase. Further understanding of these factors for support provision could be useful in identify ing and target ing potential providers for future disaster preparation and recovery effort s. In 2000, Hurlbert et al. followed up on the research for support provision by examining how social network structures allocate d resources for activation in the context of social support. This research study intended to bridge the gap by linking activated ties with the network structures rather just limiting attention to the dyadic relationship between the individuals and activated ties. Moreover, the results could provide a better understanding of how routine interpersonal environments can activate ties from pre existing social networks ( e.g., core networks) for social resources ( such as informal support) in non routine situation ( e.g. natural disaster s such as Hurricane Andrew). Social support researchers have argued that core networks that have strong and homophilous (or sharing similar characteristics) ties have better access to social support. According to Hur l bert et al. (2000) core n etworks constitute key sectors of routine interpersonal environments that serve as primary loci of interpersonal contacts. For the purpose of this study, the four aspects of network structure considered to be influential in the support context include 1) network density, 2) network size, 3) diversity dimension of network range ( e.g. geographic dispersion and gender diversity )
18 and 4) network composition. For network density and size, it was hypothesized that, the higher the density or the larger the core network, the greater the ties from the core network to activate for informal recovery support and the more likely informal support will be provided from within the core network. For the network diversity on geographic dispersion, it was hypothesized that greater geographic dispersion would reduce the proportion of core network ties to activate for informal recovery support and the less likely informal support will be provided from inside the network. For the network diversity on gender, it was hypothesized that greater the representation of men in the core network, the greater proportion of core network ties that will be activated for informal support al though with a caveat that the gender effect will be min imized as the proportion of men in the network increases. The results showed that a denser core network did increase the ties activated, and these individuals turned to core network ties for informal support in a nonroutine situation in both the preparat ion and recovery phase of Hurricane Andrew. For the effect of the network size, s maller core network ties activated for informal support within a larger core network. There was no effect observed for geographic dispersion. For gender diversity, the findings showed that core network with higher proportion of men increase d the degree of network ties activated; however if a core network is composed entirely of men then it would decrease the ties activated. Finally for network composition, having younger individuals and kin increase d the degree of core network ties activated. The proportion of support providers from within the core network was greater when the core network prior to the disaster had a higher density, was larger in size, and included men and women and kin. These analyses confirmed that core network structure affected both core network ties activated for informal support and the degree to which individuals activated these ties from within the core network. For social
19 resource s in non routine situation, the composition of the core network also served as a key determinant of core network ties activated. Based on these research findings on recruitment factors for social movement and network characteristics for support provision over the past several decades, the landscape for disaster communities in the digital age has undoubtedly evolved. I will draw on the research findings of these literatures to inform the changes of disaster communities in the age of online social media. Prior to social networking Websites, individual involvement in disaster community was mostly limited to physical participation. However, online social network s offer a new space for participation as the next section on Facebook will describe. This new se gment of online disaster community is not well researched at all and t his thesis research is intended to address that problem by examining this new segment of disaster community as an online social network. Facebook Facebook is arguably the most popular social networking Website today with more than 5 00 million active users worldwide as of July 2010 ( active users are defined as users who have returned to the site in the last 30 days). Founded in 2004m t his social networking platform helps people communicate more efficiently with their friends, family and coworkers. However, Petersen (2010) reported t he origin of Facebook has not always been so inclusive of different users Indeed, i n its early days Facebook was limited to persons affiliated with Harvard ( specifically users with a Harvard email address) who more than likely already kne w one another prior to joining Facebook (unlike the other popular social network Websites such as MySpace or Friendster which w ere designed f or users to connect with both friends and strangers online).
20 Facebook extended its registration in December, 2004, by opening to Stanford, Columbia and Yale. I n 2005, membership was extended to high school and international school networks and in 2006, it was again extended to work networks. Finally, in September, 2006, Facebook lifted its elitist status and opened its site to anyone who wanted to join. From less than 1 million active users at the end of 2004, Facebook grew to over 12 m illion active users by the end of 2006 According to the Press Room Statistics on Facebook (2010), this online social networking site grew exponentially in the following years The membership base of over 50 million active users in October 2007 doubled to over 100 million active users in August 2008 and then tripled to over 350 million active users in 2009. The concept of this wildly popular online social network originated from the freshman facebook that colleges traditionally distribute to incoming students with a photo of each classmate and their general profile. Facebook has been able to leverage the versatility of the Web platform over the years to ma k e a number of key enhancements to the communication feature of the Website including ( but not limited to) : the Wall, which is essentially the space on each users profile page to share his/her status by responding to the question Whats on your mind? or to post messages on other users walls; Publisher a main feature of Facebook for users to post information and messages on their own W all or their friends W all s; News Feed, which resembles a personalized wire service that highlights activities of their Facebook friends such as profile changes, upcoming events, birthdays, and me ssages exchanged between the walls of the Facebook friends; and Photos which is a popular application that allows users to upload photos, tag friends (the photos would then be linked to that tagged users profile) and leave comment s on the photos
21 The user friendly navigation of the Website and versatility of staying in constant touch with other users in a variety of ways contribute to the drawing power of Facebook Moreover, Facebook is easily accessible by the users through the computer as well as mobile device s wherever they go. In fact, o ver 100 million active users access Facebook through their mobile devices and these users are typically more active than nonmobile users of Facebook. The Home page of Facebook allows easy access to the core functions such as Messages, Events (calendar), Photos, Friends (list), Applications, Games, etc. The users can also perform administrative maintenance on the homepage by managing the Friend request (confirming or ignoring incoming requests as Facebook fri end), adding a friend through a suggested list of new friends (linked through existing friends) or sending a message through a suggested list of existing friends that users may not have been in touch for awhile. The Profile page hosts the Wall, Info, Photos, etc. along with sponsored ads to the right of the navigation panel the Profile page is related specifically to the Facebook user while the Home page hosts information about the friends of the user (see Figur e 1 for an example of the Facebook Profile of the American Red Cross ) The Account page allows the users to manage their privacy setting to control the information they share and with who m they share it with.
22 Figure 1 : User Profile Page for ARC Cause on Facebook Causes Facebook offers users more than 500,000 active applications to play games, interact with friends, or to form individual groups with a specific purpose, e.g., for business, education, entertainment utilities, etc. Seven of every 10 Facebook users engage in at least one application every month. One of the most popular applications on Facebook is Causes which is ranked second in its popularity with over 125 million monthly active users as of August 2010 According to the homepage of Causes (Facebook, 2010): Causes strive to empower people from all walks of life to have a positive impact on the world in which they live. We allow Facebook users to organize into communities of action focused u pon specific issues or nonprofit organizations. Among the various organizations and groups utilizing this application, one of the most prominent disaster relief organizations to use Causes to reach the masses is the
23 American Red Cross (ARC) with over 64, 000 members as of March, 2010. According to the ARC homepage on Causes (Facebook, 2010): The American Red Cross is where people mobilize to help their neighbors across the street, across the country, and across the world in emergencies. Each year, in com munities large and small, victims of some 70,000 disasters turn to neighbors familiar and new the more than half a million volunteers and 35,000 employees of the Red Cross. ARC had raised over $35,000 through the Causes application, and has shown steady increase in membership over the years. With the increasing popularity of Facebook, individual activists or organizations have seized the opportunity to mobilize their network of Facebook friends toward a collective cause for disaster victims. In fact, the Causes application on Facebook was built in 2007 with the purpose to draw support for a specific cause and is been utilized by over 35 million active users to form online communities for disseminating important information, sharing experiences, signing petitions, making donations, etc. (see Figures 2 4 for ARC on Facebook Causes ). These online communities are comprised of groups of individuals interacting with each other using communication tools such as emails, online social networks, instant mess ages and online discussion boards for social, professional, educational, philanthropic or other purposes. Not surprisingly, such emerging online communities have become an important channel of communication and support for disaster victims in their recovery process.
24 Figure 2 : Homepage for ARC Cause on Facebook Cause Bulletin Figure 3 : Homepage for ARC Cause on Facebook Fundraising Activity
25 Figure 4 : Homepage for ARC Cause on Facebook Recent Cause Activity Social Network Analysis Given the popularity in recent years of online social networks in general and of Facebook in particular researchers have displayed growing interest in the analysis of online community network structures. Social network analysis can help researchers seek answer s to the fundamental question of how autonomous indiv iduals are connected to one another in a community structure In recent decades r esearchers from a variety of applied fields outside of social sciences such as management consulting, public health, crime and war/fighting have also used this analytical technique to examine relations within a group or population. S ome examples include: helping organizations leverage knowledge and
26 capabilities with its members for knowledge management stopping the spread of infectious diseases to provide better health care and social support for public health, and fighting organized crime or terrorist g roups for national security ( Borgatti et al., 2009). Social network analysis enables sociologists to examine the relations between individuals in the community and the varying properties of how these ties account for differences in the outcome for the indi viduals or the community. For a traditional social research data set, the two different sets of entities typically include persons (i.e., cases) and the attributes (i.e., variables) which is considered a 2 way 2 mode matrix. For a social network data set both sets of entities are typically persons or nodes (rather than persons and attributes) which are considered a 2way 1 mode matrix. Simply put, the traditional data focus on actors and attributes and network data primarily focus on actors and relations. The data allow social researchers to examine the choices that individuals make in join ing or affiliat ing with a particular group and how the groups themselves may affect the choices of the individual to join. According to Wasserman and Faust (1994), the actor (discrete individual, corporate, or collective social units ) the group (the collection of all actors on which ties are to be measured) and the relations (the collection of ties of a specific kind among members of a group) are the major concepts in network analysis : A social network consists of a finite set or sets of actors and the relation or relations defined on them. The presence of relational information is a critical and defining feature of a social network. Social network data include two types of variables : 1) structural variables which defined as meas urements on ties (linkage of actors to one another) among the
27 social units such as association or affiliation: for example, joining the same social club, or for the purpose of this thesis research, the same Facebook C ause; and 2) composition variables which are defined as measurements on characteristics at the level of individual actors (attributes of actors) such as age, gender, race, size of collective actors, etc or at the level of groups The centrality concept for actors is also important for social network analysis in illustrating social power by means of relations between actors or the entire population (Hanneman,and Riddle, 2005) Centrality highlights the actors who are extensively involved or have relational ties with other actors in the network I n other words, a central actor would be involved with many ties with no distinction between receiving and sending i.e., nondirectional relations (Wasserman and Faust, 1994). Furthermore, different measures of centrality are used to reflect the actors at the center of the set of actors which include degree (actors with the most ties to other actors in the network) closeness ( how close an actor is to other actors in the network) and betweenness (two nonadjacent actors may depend on other actors who are between the two actors for interaction) Given that network measurements focus on the use of structural or relational information of actors or groups for analysis, the traditional analytical methods of testing significance of relationship between two variables such as correlations, multiple regressions, t tests, etc. are not applicable for some aspects of the analysis (Wasserman & Faust, 1994) However, other standard descriptive statistical techniques such as univariate, bivariate, or multivariate analysis can be used to describe and summarize social network data (Hanneman, 2005). In fact, a separate set of network analytical methods was
28 developed to consider the relational tie among the actors or groups as well as their attributes. The key distinction between tradit ional social research analysis and social network analysis can best be summarized by Borgatti et al., (2009): Whereas traditional social research explained an individuals outcomes or characteristics as a function of other characteristics of the same individual (e.g., income as a function of education and gender), social network researchers look to the individuals social environment for explanations, whether through influence processes (e.g., individuals adopting their friends occupational choices) o r leveraging processes (e.g., an individual can get certain things done because of the connections she has to powerful others) Another key component of social network analysis is to visually illustrate the various network properties that characterize the structures, positions and dyadic properties (connectedness of the structure) and the overall distribution of ties (Borgatti et al., 2009). UCINET is a specialized software commonly used for social network analysis to visualize data. Social researchers can consider different types of social networks for analysis depending on the data structure and the number of sets of actors to analyze. Furthermore, the analysis can also include different types of actors (also known as social entities) which can be people, subgroups (usually made up of people), organizations (usually made up of subgroups of people), or collectives/aggregates such as communities or nationstates ( usually made up of many organizations and subgroups). Wasserman and Faust (1994) defined the mode of a network as the social entities (actors or events) w hich structural variables are measured and the number of modes derives from the number of distinct types of social entities within the network. The most common types are onemode networks which involve a single set of actors (such as Facebook members) followed by twomode networks which could be comprised of two
29 distinct sets of actors or one set of actors and one set of events (such as Facebook members and the causes to which they belong) OneMode Network Onemode network involves only one set of actors to measure one or more types of relations among them. Like actors, there are also different types of relations to study such as individual evaluation ( measuring sentiments such as liking, respect, etc. ), transfer of material resources (measuring transactions such as exchange of gifts, donations, etc.) transfer of non material resources ( measuring communication such as sending or receiving messages ), interaction (measuring their pres ence in the same place at the same time) kinship ( measuring relations such as marriage, descent, etc.). In addition to studying the relations of the set of actors, the attributes of the actors can also be studied such as age, gender, race, socioeconomic status etc. which is similar to traditional social research. Wasserman and Faust (1994) used Krackhardt s study of a single set of actors (i.e., 21 high tech manager s) in a small manufactur er that produces high tech machinery as a classic example of a one mode network. The objective of the analysis was to examine the managers perceptions on the structure of the network specifically on seeking informal advice and forming friendships as well as their reporting relationships The data also included actor attributes for age, tenure (length of time employed by the company), level (level in the corporate hierarchy) and department. These managers were given a questionnaire to self report who they would go to for advice at work and who was their friend f rom a roster of names of the other managers Figure 5 shows the relational data set of two way onemode for advice which includes the set of 21 high tech managers
30 (actors) with dichotomous data for example, manager ID #1 indicated that he/she would go to manager ID #2, #4, #8, #16, #18 and #21 for advice. Figure 5 : Krackhardt's OneMode Network Matrix for Advice Relation TwoMode Network Two mode network can involve two sets of actors or a single set of actors and a set of events. The types of actors, relations and attributes for one mode network can also be applied to a twomode network. For the twomode network with two sets of actors, there can be unique attributes for each set of actors and should measure at least one relation between the two sets of actors (relations can also be measured within the set of actors). The focus of this thesis research will be on another type of twomode network which involves one set of actors and one set of events. This type of network is also known as an affiliation network the second mode is a set of events with which the first mode of actors is
31 affi liated The affiliation is based on a subset of actors who are related to each other through one or more common events, and these events can be membership in clubs or voluntary organizations, attendance at social events, etc. I n addition to actor attributes, the events can also have attributes included in the data set such as size of the membership for the club/organization (or event), years of establishment, etc. A classic e xample of a twomode network analysis was conducted by Davis, Gardner, and Gardner (DGG) in studying the social activities of 18 women in a Southern cit y (Davis et al., 1941) The social events (e.g., social club meetings, church events, parties, etc.) that these women attended were recorded using newspaper records and interview s over a period of nine months. The objective of the analysis was to examine whether the women have social relations with other women within their own social classes. The first mode consist ed of the 18 women and the second mode consists of the 14 events attended by these women. Figure 6 shows the womanby event matrix which includes the set of 18 women (actors) affiliated with the social functions that these women attended in the binary measurement for example, Evelyn (CASE ID #1 ) attended Events #1, #2, #3, #4, #5, #6, #8 and #9. Figure 7 visualizes the 2 mode data using a graph with the actors and events as the nodes and the lines connecting the actors to the events (no lines connecting the actors or events directly).
32 Figure 6 : DGG's Two Mode Network of Women by Event Matrix Figure 7 : DGG's Two Mode Network of Souther n Women and Social Events
33 Chapter 3 : Data and Methods Research Questions The first set of research questions for this thesis research is to investigate the compositional differences of the social network s among the Facebook members sampled from the ARC Causes before the Haiti earthquake disaster as compared to after the disaster Specifically, I want w ould like to examine the differences between the two samples on the following: 1) the rate of Facebook members joining the ARC Cause; 2) the average number of other Facebook Causes these sampled ARC Joiners belonged to; 3) the average size of the Facebook Causes not related to ARC that ARC Joiners belonged to; 4) the types of Facebook Causes these joiners belonged to and 5) the gender diversity of two samples. The second set of research questions is to investigate the structural differences of the social network of the ARC joiners sampled before and after the disaster. Specifically, I want to examine the differences between the two samples on the 1) relational ties of the ARC Joiners and Facebook Causes ; and 2) the density of the networks for ARC Joiners a nd Facebook Causes. Research Methods The sample of the Facebook members was selected from the ARC Causes Members section which listed all the members who joined the cause in real time. In order to further assess how an actual disaster event may impact the
34 participation of these Facebook members on ARC Causes I randomly select ed1 1 The random selection entails using SPSS to randomly select 100 cases out of the all the Facebook members joined the ARC Cause during the specified period. I then copied and pasted the required data of the randomly selected sampled members from the Members section on the ARC Causes page to an Excel spreadsheet. 100 members out of the 553 total members who joined during a tenday period b efore the Haiti earthquake disaster (i.e., between January 2, 2010 and January 11, 2010 ) and another 100 members out of the 4,386 total members who joined during a tenday period after the disaster (i.e., between Janu ary 12, 2010, and January 21, 2010. In addition, I recorded the other causes that each sampled joiner belonged to and coded the data into 8 categories (i.e., administrative, animal protection, disaster relief, game, memorial, philanthropy, political religion and social services). This sampling method allows for comparison of any differences of the social networks of the sampled members before and after the Haiti earthquake. I collected all the causes each sampled member belonged to which ranged from as few as one cause (i.e., the ARC c ause) to as many as 118 causes (see Figure 8 for the format of the required data collected in the Excel spreadsheet ) as well as the membership size of the cause. The data on the causes were collected during the time frame of April 14, 2010 to April 18, 2010, which was approximately four months after the Haiti earthquake disaster The interpretation of results in Chapters 4 to 6 took the following assumptions into consideration: 1) the majority of the causes to which the sampled Facebook members belonged were joined before they joined the ARC Cause; 2) the rate of joining these other causes after they joined the ARC Cause is not significantly different from before they joined the ARC Cause. This is discussed further in Chapter 7 Conclusi on.
35 The two samples belong to a total of 1,281 causes, and the number of causes to which members belong ed before the disaster (625) is fairly equal to the number of causes to which they belonged after the disaster (664). Additionally, I also collected the demographic characteristics (i.e., gender) of the 200 sampled members to the extent allowed the privacy setting s allowed from their Facebook profile pages Figure 8 : Sample of Data Format for Actors and Events Ethics for Online Social Research Generally, individuals corresponding in public chat rooms or discussion groups could perceive the setting of their conversations to be private and research on these individuals would require the researcher to get informed consent from the subject under observation. H owever, public lists or discussion boards only require proper citation for materials used in the discussion (Barnes, 2004). The Facebook Website is considered to be a public forum, and the social networks on the Causes platform permits any Facebook members to join, so informed consent should not be required under these circumstances. However,
36 in an effort to protect the privacy of the Facebook members in this study, only first name will be referenced to ensure complete privacy of their identity.
37 Chapter 4 : ARC Network Composition Before and After Haiti Earthquake Disaster The fundamental data structure of social networks includes two types of variables: compositional and structural. Compositional variables refer to attributes of actors from sources other than the network itself (e.g., how many people joined an organization at a given time? W hat is the gender of the people in the organization?). Structural variables can be used to measure the ties between pairs of actors (i.e., connection between the actors or the organizations). Depending on the objective of the analysis, t hese variables are commonly used as measures for social network data. This chapter will focus on comparing the composition of the network before and after the Haiti earthquake disaster including the trend of join ing the ARC Cause, the average number of Facebook Causes not related to ARC that ARC Joiners belonged to, the average size of the Facebook Causes not related to ARC the types of Facebook Causes the ARC Joiners belonged to and the gender diversity of the network s Trend of Joining the ARC Cause Fig ure 9 clearly illustrates the drastic increase of Facebook members joining the ARC Cause in the period of ten days after the disaster compared to the period of ten days prior The average number of joiners of the ARC Cause wa s 55 per day within the ten days before the disaster, and the average nearly increased tenfold to 439 per day within the ten days after the disaster. Moreover, the difference between these periods is more than tenfold for Day 4 through Day 10 in compari ng the days bef ore and after the disaster. T he spike of this trend can be attributed to a variety of influences such as media coverage by the major
38 outlets (e.g., traditional media of television and radio as well as emerging media of t he Internet accessible on different platforms). However, the forthcoming findings from the other composition variables also lead to evidence of influences from the members and causes of Facebook mobilizing their resources to recruit others to join the ARC Cause after the disaster. Figure 9 : Number of Facebook Members Joining the ARC Cause Organizational Affiliations of ARC Joiners Figures 10 and 11 illustrate the differences of attributes of the two samples of ARC Joiners before and after the disaster In comparing the number of Facebook Causes the sampled ARC Joiners belonged to, Figure 10 shows a slightly higher average number of Facebook Causes after the disaster (12.3 causes) than before the disaster (11.6 causes). Note the ARC Cause was excluded fr om the calculation of the average number of causes It is also important to note the outliers of this average within the two different samples : there were as few as one cause (i.e., the Facebook members belonged to the ARC Cause only) and as many as 118 causes some of these joiners Day 1 Day 2 Day 3 Day 4 Day 5 Day 6 Day 7 Day 8 Day 9 Day 10 After Disaster 48 220 477 607 587 515 540 551 426 415 Before Disaster 85 96 91 68 38 40 33 44 25 33 0 100 200 300 400 500 600 700Number of Joiners
39 belonged to. However, Figure 11 illustrate s the differences with the distribution of Facebook Causes for the two samples of ARC Joiners. While the differences of the overall average number of Facebook Causes the sampled ARC Joiners belonged to can be considered small or perhaps not meaningful, t here are clear differences with the joiners before and after the disaster when analyzing the Facebook Causes at the incremental level. For Figure 11 14 j oiner s before the disaster belonged to ARC Cause only compared to 8 ARC Joiners after the disaster suggesting that the ARC Joiners sampled after the disaster were more active in terms of other causes they belonged to other than ARC Similarly, the se j oiners sampled after the disaster belonged to more Facebook Causes (i.e., 13 ARC Joiners belonged to 25 or more Facebook Causes) compared to the j oiners before the disaster (i.e., 10 ARC Joiners belonged to 25 or more Facebook Causes). The distribution of higher incremental numbers of Facebook Causes the ARC Joiners belonged to can indicate Facebook members sampled after the disaster were more active in disaster relief or philanthropic causes which will be further explored later in this chapter.
40 Figure 10 : Average Number of Facebook Causes for ARC Joiners Note: The ARC Cause was excluded from the calculation of the average number of causes Figure 11 : Distribution of Facebook Causes for ARC Joiners Figure 12 shows the average size of the causes collected from the sampled ARC Joiners. Note t he ARC Cause was excluded from the calculation of the average size of causes There is a sizeable difference of 162,944 Facebook members when comparing the average membership size of causes 11.6 12.3 0 2 4 6 8 10 12 14 16 18 20 Before Disaster After Disaster Average Number of Causes 14 18 9 4 15 13 2 5 0 2 3 1 4 10 8 12 13 11 10 5 10 2 5 3 2 4 2 13 0 2 4 6 8 10 12 14 16 18 20 Number of ARC Joiners Before Disaster After Disaster
41 before and after the disaster (i.e., 500,700 Facebook members after the disaster and 337,756 before the disaster). For the causes that sampled ARC Joiners belonged to before the disaster, the membership size ranged from as few as 6 F acebook members to as many as 6,176,464 Facebook members. Similarly, for the causes that sampled ARC Joiners belonged to after the disaster, the membership size ranged from as few as 1 Facebook members to as many as 6,158,144. While the range of the membership size of the two samples is similar, the drastic difference between the average number of Facebook members can indicate the causes with larger membership size or causes that joined in forces (i.e., bloc recruitment) were more successful in recruiting new members after the disaster by mobilizing their active members to recruit others Figure 12 : Average Membership Size of Facebook Causes Note: The ARC Cause was excluded from the calculation of the average number of Facebook members 337,756 500,700 100,000 200,000 300,000 400,000 500,000 600,000 Before Disaster After DisasterAvg Number of Members
42 Identifiable Recruitment Community Following up from the previous finding that sampled ARC Joiners after the disaster belonged to more Facebook causes than the joiners before the disaster, Figures 13 and 14 show the types of causes to which they belong. The ARC Joiners sampled after the disaster belonged to almost twice as many disaster relief causes compared to the joiners sampled before the disaster. Figure 13 shows that 4% of the Facebook Causes the sampled ARC Joiners belonged to after the disaster are related to disaster relief (i.e., 24 out of 658 total causes), and Figure 14 shows that 2% of the causes the joiners belonged to before the disaster are related to disaster relief (i.e., 14 out of 614 total causes). Note the number of Fac ebook Causes coded differs from the original sample size (i.e., 614 vs. 625 before the disaster and 658 vs. 664 after the disaster) due to the fact that a fraction of these causes were disabled since the sample was drawn, so the disabled causes could not be coded and subsequently included in the calculation. Over half of these disaster relief causes are related to the Haiti earthquake such as Feed Hungry Children in Haiti, Help Earthquake Survivors in Haiti, Help Haiti Now !! Hope for Haiti Now: A Global Benefit for Earthquake Relief, etc. In addition to the increase of Facebook Causes related to disaster relief, the causes related to philanthropy also increased from 45% before the disaster (i.e., 273 out of 614 total causes ) to 50% after the disaster (i.e., 330 out 658 total causes). This indicates that ARC Joiners sampled after the disaster were more likely to come from philanthropic causes such as child abuse, diseases/illnesses, domestic violence, environment, military, etc. the more prominent examples in terms of membership size include A World Without Breast Cancer, Amber Alert on Facebook, Free Postage For All Families of Deployed Military, Society
43 Against Child Abuse, Stop Global Warming, etc. On the other hand, th ere is a decrease of Facebook Causes related to games from 18% before the disaster (i.e., 108 out of 614 total causes) to 12% after the disaster (i.e., 82 out of 658 total causes). The most popular causes for games by far were related to Farmville with o ver 70% of all causes related to games have Farmville in the name of the cause both before and after the disaster The diminished share of these causes for games could result from the greater proportion of causes related to disaster relief and philanthr opy as these causes were more likely to recruit new ARC members when the salience of disaster relief is heightened by the Haitian episodes Figure 13 : Proportion of Types of Facebook Causes for ARC Joiners Before Disaster 1% 13% 2% 18% 4% 45% 8% 9% Administrative Animal Protection Disaster Relief Game Memorial Philanthropy Political/Religious Pop Culture
44 Figure 14 : Proportion of Types of Facebook Causes for ARC Joiners After Disaster Gender Diversity Interestingly, when comparing the gender composition of the network s before and after the disaster in Figure 15, female joiners of the ARC Cause dominated the sample after the disaster with 79% compared to 21% of male joiners. Moreover, there is an increase of 11% of these female joiners sampled after the disaster when compared with the proportion of female joiners sampled before the disaster (i.e., 68%). The proportion of female joiners of the ARC Cause sampled before and after the disaster also exceeded the proportion of female members of the Facebook population in the U.S. (i.e., 56%) by 12% before the disaster and 23% a fter the disaster Historically, the disaster research literature consistently reported men providing more support than women in the event of a disaster. However, this finding may indicate a new trend in the digital age that women are more likely than men to participate in low cost/low risk form of activism on the Internet in providing emotional support rather than physical support. 1% 8% 4% 12% 5% 50% 5% 15% Administrative Animal Protection Disaster Relief Game Memorial Philanthropy Political/Religious Pop Culture
45 Figure 15 : Gender Diversity of ARC Joiners 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% U.S. Facebook ARC Cause (Before) ARC Cause (After) 56% 68% 79% 44% 32% 21%Gender Proportion Female Male
46 Chapter 5 : Two Mode Network of Sampled ARC Joiners and Their F acebook Causes Following Fernandez and McAdams (1988) research study on the role of multiorganizational fields in recruitment to social movements, this chapter will examine the structural factors that may pull the sampled ARC Joiners into a specific cause by evaluating the Facebook causes they belong to. The multiorganizational fields are defined as ties formed by overlapping memberships among organizations (Fernandez and McAdam, 1988: 358). The ties amo ng these Facebook causes can potentially be leveraged to mobilize their resources (i.e., existing members) in recruiting others to their cause. The focus of this chapter is on the description of multiorganizational fields of the Facebook Causes the sample d ARC Joiners belonged to before and after the Haiti earthquake disaster and any clusters of joiners or causes apparent from the network illustrations. Further analysis of network density will be discussed in Chapter 6 The method of network analysis is used to examine the multiorganizational fields of the sampled ARC Joiners, and it is limited to the interpretation on the pattern of overlapping organizational membership given no other data are available to verify any par ticular pattern of relational ties. Social network analysts typically use matrices or graphs to represent network data to illustrate patterns of ties among the actors. When a network data set has one set of actors and a set of events or organizations to which the actors belong, it is a type of twomode network known as the affiliation network (the two modes are actors and events). The primary interest of this data set is to understand the
47 macro micro relationship, i.e., the Facebook Causes the sampled AR C Joiners belong to (microstructure) and how the Facebook Causes may influence the decision to join their cause (macro structure) The Facebook Causes of each sampled ARC Joiner belonged to were collected before and after the disaster t wo peopleby org anizational affiliation matrices were created for the two two mode networks: 100 sampled ARC Joiners by 625 Facebook Causes before the disaster, and 100 sampled ARC Joiners by 664 Facebook Causes after the disaster. Table 1 is an example a n ARC Joiner by F acebook Cause matrix with 100 sampled Facebook members joined the ARC Cause ten days before the disaster by the pool of Facebook Causes they all belonged to. The twomode network data are arranged in a rectangular data matrix of actors (rows) by events (c olumns). The first mode in an affiliation network (or sometimes known as membership network) is a set of actors and the second set is a set of events (or organizations) they affiliate with Each Facebook Cause is a variable, and a binary measurement is u sed to indicate whether a specific ARC Joiner belonged to the cause. The only requirement is the actors must be affiliated with one or more organizations. The entry of 1 indicates the sampled ARC Joiner belongs to the Facebook Cause adjacent to it, otherwise it will be coded as 0 for not being affiliated with that particular cause. For example, ARC Joiner #1 (Tangela) belonged to 25 Facebook Causes (including the ARC Cause) which is coded as a 1 for Cause #1 25. The horizontal elements highlighted in red show which cause s Tangela belonged to Also, the vertical elements highlighted in red show all the ARC Joiners that belonged to the specific cause. For example, all the sampled ARC Joiners belonged to Cause # 4, the ARC Cause which is coded a for ARC Joiner #1100.
48 Table 1 : Two Mode Network Matrix of ARC Joiners and Facebook Causes Before Disaster Note: The data on the causes were collected from April 14, 2010 to April 18, 2010. Figures 16 and 17 present the graphs used to visualize the twomode data in which both actors (sampled ARC Joiners) and events (Facebook Causes) are treated as nodes The lines are used to show the ties between the joiners and the causes. As Figure 16 shows, t he graph before the disaster has 72 4 nodes (100 nodes of joiners illustrated in red circles and 624 nodes of causes illustrated in blue squares) As Figure 17 shows, the graph after the disaster has 763 nodes (100 nodes of joiners and 663 nodes of causes ). Note t he ARC Cau se was excluded from the individual sample of these graphs The isolates on the left of the graphs are the sampled ARC Joiners that belonged only to the ARC Cause and not affiliated with any other causes or joiners (a total of 14 isolates from the sample before the disaster and 8 from after the disaster)
49 Figure 16 : Two Mode Network Graph of ARC Joiners and Facebook Causes Before Disaster Figure 17 : Two Mode Network Graph of ARC Joiners and Facebook Causes After Disaster In evaluating the graphs of the twomode networks of ARC Joiners and Facebook Causes before and after the disaster (Figures 16 and 17), there does not seem to be any apparent pat tern of clusters of causes or joiners in common
50 though it is difficult to observe the pattern of connections or social significance with over 700 nodes given how closely they are clustered together. In fact, the causes and joiners are mostly clustered in a single network rather than broken out into clusters that can be visually apparent. As a result no further analysis is conducted for the twomode networks H owever, these networks can be converted into two onemode data sets and examine relations within each mode separately. In Chapter 6 a n actor by actor data set was created to measure the ties between each pair of actors by the number of Facebook causes these sampled ARC Joiners belonged to and connection between these causes. A n event by event data set was also created to measure the ties between each pair of causes by the number of sampled ARC Joiners they have and connection between these joiners. Th e analyses for these onemode networks include the relational ties of the sampled ARC Joiners /Facebook Causes at various levels of relations before and after the disaster as well as the density of the overall network and each level of relations within each network
51 Chapter 6 : OneMode Network of Sampled ARC Joiners and Their Facebook Causes To further examine the role of multiorganizational fields from the last chapter, the twomode network of ARC Joiners by Facebook Causes is converted into two onemode networks of ARC Joiners by ARCJoiners and Facebook Causes by Facebook Causes in order to study each network separately. Furthermore, t his chapter also takes into consideration McAdams (1986) argument for the need to distinguish the varying forms of activism (i.e., cost and risk) in relation to the responsibility of the prior contact with a recruiting agent (i.e., networks, relationships or communities) that pull the individual into activism. The act of joining a Facebook cause can justifiably be considered a form of low cost/low risk activism following the same definition as McAdam for cost in terms of time, money and energy required for activism; and risk in terms of anticipated danger or behavioral consequences (of being perceived negatively) The membership for Facebook and joining the ARC cause is free (donation to the ARC is at th e discretion of the joiner), and the level of participation is also dependent on the joiner (e.g., posting wall messages, recruiting other members, making donations, fundraising, etc.). It is also relatively risk free to participate given all activities are completely Webbased so there should be minimal danger posed for financial, physical or social harm (unlike the 1964 Freedom Summer project) Moreover, the ARC has the reputation of being the nations premier emergency response organization so joining its cause should not have bear any negative perceptions.
52 Historically, prior contact is a strong predictor of participation in low cost /low risk activism. To study this further with the two ARC samples in the digital age of online social movement the relational ties (i.e., linkage between a pair of actors ) of a one mode network of ARC Joiners before and after the disaster are used to examine how connected the y are The analysis of the network can start with the number of ties (or connections) among each sample of ARC Joiners. Any differences of how connected they are before and after the disaster can be a potential indicator of heightened recruitment by prior contact of the sampled ARC Joiners towards a common cause ( based on liberal interpretation of data available) The adjacency matrix (or sociomatrix) is the most common form of matrix for onemode network the rows and columns of the actors in the network are in identical order. The matrix to illustrate the onemode network of ARC Joiners before and after the disaster consist s of rows and columns of the names of these ARC Joiners. The entries in the matrix represent the number of Facebook Causes each ARC Joiner has in common wi th the other ARC Joiners in each sample. The graphic display of network analysis uses points to represent the actors in the network and lines to represent ties or relations among the actors. T he ties illustrated in the graph between the ARC Joiners in each sample show the circle labeled with their names and connection to each other through one or more Facebook Cause s the collection of ties can also be illustrated of a specific relation such as the set of ARC Joiners with more than one Facebook Cause s in common (other than the ARC Cause).
53 OneMode Network of Sampled ARC Cause Joiners Table 2 is the ARCJoiner by ARCJoiner one mode network matrix with 100 sampled Facebook members who joined the ARC Cause ten days before the disaster these ARC Joiners are the set of actors for this data set There are as many rows and columns as the number of actors in the data set (which is not all shown in the table illustrated ) The primary interest of this data set focuses on the relational ties among these sampled ARC Joiners to examine how many other causes (in addition to the ARC Cause) they co joined. The elements (score in the cells) of the matrix indicate the tie between each pair of actors, i.e., number of Facebook Causes co joined. Note these ARC Joiners belonged to at least one cause in common (i.e., the ARC Cause). The diagonal elements highlighted in red show the number of Facebook Causes each ARC Joiner belongs to. For example, ARC Joiner #1 (Tangela) belonged to a total of 25 Facebook Causes (including the ARC Cause) and connected to ARC Joiners #2 4 (Anthony, Masa and Barbara) through the ARC Cause only. However, Tangela shared membership of two Facebook Causes with ARC Jo iner #5 (Claire) the ARC Cause and one other
54 Table 2 : OneMode Network Matrix of ARC Joiners Before Disaster Note: The data on the membership were collected from April 14, 2010 to April 18, 2010. Figure 18 and 19 are the graphical presentation of ties at each level of relations between pairs of ARC Joiners for the samples before and after disaster. The four graphs for each sample show the collection of ties for a specific relation among the ARC Joiners, i.e., more than one Facebook Cause in common with other ARC Joiners; more than two Facebook Causes in common with other ARC Joiners; and so on. The standalone points (or nodes) on the left of the graph separated from the rest of the network are isolates which have less than the cut off number of causes in common with any other joiner. In particular, the first graph with a cut off value of two causes shows that a number of joiners have only the ARC Cause in common with the others in the sample
55 Figure 18 : One Mode Network Graph of Relational Ties of ARC Joiners Before Disaster Note: Topleft shows the ARC Joiners with greater than one relation, topright shows greater than two relations, bottom left shows greater than three r elations and bottom right shows greater than four relations. Figure 19 : One Mode Network Graph of Relational Ties of ARC Joiners After Disaster Note: Topleft shows the ARC Joiners with greater than one relation, topright shows greater than two relations, bottom left shows greater than three relations and bottom right shows greater than four relations.
56 According to Table 3 there was greater number of relational ties, i.e., linkage between pairs of ARC Joiners after the disaster compared to before the disaster. For ARC Joiners with more than one Facebook Cause in common, there were 2,614 ties (i.e., 1,307 dyads or pairs of actors with ties between them) after the disaster compared to 2,114 ties (i.e., 1,057 dyads) before the disaster There were also consistently more linkages after the disaster for ARC Joiners with more than four Facebook Causes in common after the disaster. Moreover, 22% (22/100 = 0.22) of the ARC Joiners sampled before the disaster had no relational tie to other ARC Joiners outside of the ARC Cause itself (i.e., isolates) compared to the 12% (12/100 = 0.12) isolates sampled after the disaster. Note that 8 out of the 22 isolates in the sample before the disaster belonged to more than one cause but onl y had a relational tie to other sampled joiners through the ARC Cause only T he remainder of the isolates did not belong to any other causes other than the ARC Cause as illustrated in Figure 18 Similarly, note that 4 out of the 12 isolates in the sample after the disaster belonged to more than one cause but had a relational tie to other sampled joiners through the ARC Cause only T he remainder of the isolates did not belong to any other causes other than the ARC Cause as illustrated in Figure 19. These results indicate the ARC Joiners were more connected to each other after the disaster compared to before the disaster which may be attributed to heightened recruitment effort s by prior contacts for these ARC Joiners It should be acknowledged, however, that other factors such as attitudinal affinity for disaster recovery, constant media coverage of the disaster, etc. may also increase the connection among the ARC Joiners.
57 Table 3 : Relational Ti es of ARC Joiners Before and After Disaster Ties Relations Before After Isolates 22 12 >1 relation 2114 2614 >2 relations 882 1062 >3 relations 430 476 >4 relations 202 218 Network density measured by the number of ties among the actors as a proportion o f the number of possible ties can provide an index of the degree of dyadic connection in a population. For example, information about network density can be leveraged to assess the speed at which information can be shared within the network based on how connected the actors are. For the samples of the ARC Joiners before and after the disaster, the density of the network can contribute to how quickly existing ARC members recruit others to join the ARC Cause. The density of a graph can be computed by dividing the sum of all ties by the number of possible ties and the maximum possible ties depend on the number of nodes (or actors in the network) For the ARC Joiners in the graphs illustrated in F igure s 18 and 19, there is a total of 4,950 possible ties [100(1001 )/2 = 4,950]. The most dense relation in the two samples is among the ARC Joiners after the disaster The density for the network with ARC Joiners after the disaster is consistently higher for the four levels of relations detailed in Table 4 The density for ARC Joiners with more than one Facebook Causes in common after the disaster is 0.26 (1,307/4,950 = 0.26) compared to 0.21 (1,057/4950 = 0.21) for ARC Joiners before the disaster. This indicates that 26% of all possible ties are present for the sample after the disaster compared to 21% for the sample
58 before the disaster. The gap of the d ensity between the two samples becomes closer for ARC Joiners with more than four Facebook Causes in common after the disaster (109/4,950 = 0.022) compared to before the disaster (101/4,950 = 0.020). The density for the entire network of ARC Joiners after the disaster is 1.48 compared to the network of ARC Joiners before the disaster of 1.41 (including the ARC Cause in the calculation of density) Unlike the density referenced previously on the proportion of ties present in a dichotomized matrix at a spec ific cut point (of relational ties ), t he density referenced here indicates the average number of common organizational affiliations in each 100 X 100 matrix of sampled ARC Joiners. Table 4 : Relational Density of ARC Joiners Before and After Disaster Ties Relations Before After >1 relation 0.21 0.26 >2 relations 0.09 0.11 >3 relations 0.04 0.05 >4 relations 0.02 0.02 In addition to examining the onemode network of the sampled ARC Joiners, the Facebook Causes each of these joiners belonged to can also be analyzed on how connected these organizations may be to one another before and after the disaster For the one mode network of the Facebook Causes, the rows and columns of the actors in the network are the pool of organizations (or Facebook Causes) the sampled ARC Joiners belonged to the sampled joiners belonged to a pool of 62 5 Facebook causes before the disaster and the joiners sampled after the disaster belonged to a pool of 664 Facebook causes T hese sampled joiners can belong to as few as one cause (i.e., ARC Cause) or over a
59 hundred other causes from the pool of aforementioned Facebook Causes. The matrix to illustrate the onemode network of Facebook Causes the ARC Joiners belonged to before and after the disaster is consisted of rows and columns of the name of these causes The entries in the matrix represent the number of sampled ARC Joiners these causes have in common with each other The graphic display illustrate s the ties between these Facebook Causes in each sample the collection of ties can also be illustrated of a specific relation such as the set of Facebook Causes with more than one sampled ARC Joiner in common. One Mode Network of Facebook Causes (of Sampled ARC Joiners) Table 5 is a n example of a one mode network matrix with 625 Facebook Causes the sampled ARC Joiners belonged to ten days before the disaster these Facebook Causes are the set of events for this data set. There are 625 rows and 625 columns in the data set (which is not all shown in the table illustrated). The primary interest of this data set focuses on the relational ties among these Facebook Causes to examine how many other ARC Joiners they shared. The diagonal elements highlight ed in red show the number of ARC Joiner each cause had. The entry of 0 indicates the two causes adjacent to each other do not have any members in common. H owever, if the entry is greater than or equal to one, then they share at least one ARC Joiner in common For example, Facebook Cause # 4 ( American Red Cross ) has a hundred sampled ARC Joiners in the entry given the entire sample was selected from the ARC Cause. Moreover, the ARC Cause also shared four ARC Joiners with the cause called Mandatory Li fe Sentences for Pedophiles and Child
60 Molesters, Death for Child Killers, five ARC Joiners with A Real Man Never Hits A Woman and nine ARC Joiners with Allow God in School Table 5 : OneMode Network Matrix of Facebook Causes Before Disaster Note: The data on the causes were collected from April 14, 2010 to April 18, 2010. Figure 20 and 21 are the graphical presentation of ties at each level of relations between pairs of Facebook Causes the sampled ARC Joiners belong ed to before and after disaster. The four graphs for each sample show the collection of ties for a specific relation among the Facebook Causes (excluding the ARC Cause) i.e., no ARC Joiner in common with other cause ( no relation at all ); more than one ARC Joiners in common with other causes (more than one relation); and so on. The isolates on the left of the graph separated from the rest of the network are the Facebook Causes that do not have any sampled ARC Joiners in common with another cause
61 Figure 20 : One Mode Network Graph of Relational Ties of Facebook Causes Before Disaster Note: Topleft shows the Facebook Causes with zero relation, topright shows greater than one relation, bottom left shows greater than two relations and bottom right shows greater than three relations. Figure 21 : One Mode Network Graph of Relational Ties of Facebook Causes After Disaster Note: Topleft shows the Facebook Causes with zero relation, topright shows greater than one relation, bottom left shows greater than two relations and bottom right shows greater than three relations.
62 Contrary to the networks of the ARC Joiners, there w ere more relational ties, i.e., linkage between pairs of Facebook Causes before the disaster compared to after the disaster according to Table 6 For the Facebook Causes with more than one ARC Joiner in common, there were 3,348 ties (i.e., 1,674 dyads or pairs of causes with ties between them) before the disaster compared to 2,344 t ies (i.e., 1, 172 dyads). Given that the network of Facebook Causes (excluding the ARC Cause) before the disaster was slightly smaller than the network after the disaster ( 625 vs. 664 ), this could increase the chances of connection between pairs of causes for the network before the disaster since the sampled ARC Joiners had a smaller pool of causes to share. There were also consistently more linkages before the disaster for Facebook Caus es with zero to three ARC Joiners in common after the disaster. Moreover, only two out of all the Facebook Causes before the disaster had no relational tie to other causes compared to the three isolates after the disaster. The minimal number of isolates for Facebook Causes can indicate most of these causes are connected to each other one way or another Table 6 : Relational Ties of Facebook Causes Before and After Disaster Ties Relations Before After Isolates 2 3 > 0 relation 38,420 31,964 > 1 relations 3,348 2,344 > 2 relations 682 594 > 3 relations 228 206 Note: The ARC Cause was excluded from the calculation of the relational ties of Facebook Causes
63 For the Facebook Causes in the graphs illustrated in Figures 20 and 21, there is a total of 195,000 possible ties [625( 6251 )/2 = 195, 000] before the disaster and a total of 220,116 possible ties [664(6641)/2 = 220 ,116] after the disaster The most dense relation in the two samples is among the Facebook Causes before the dis aster. The re is smaller difference for the density of the network s with Facebook Causes before the disaster at each of the four levels of relations detailed in Table 7 compared to the causes after the disaster The density of the Facebook Causes with more than one sampled ARC Joiner in common before the disaster is 0. 0085 ( 1, 665 / 195,000 = 0. 0085) compared to 0. 0053 (1, 172 / 220,116 = 0. 0053 ) for ARC Joiners sampled after the disaster. This indicates that 1 % of all possible ties are present for the network of causes before the disaster compared to less than 1% for the network of causes after the disaster. The gap of the density between the two samples becomes insignificant for the networks of Facebook Causes with more than two or three Facebook Caus es in common. The density for the entire network of Facebook Causes before the disaster is 0.1159 compared to the network of Facebook Causes after the disaster is 0.0858 this indicates the overall network of Facebook Causes before the disaster is more connected which can potentially be more influential in recruitment dependent on how active their members may be. Again, the density referenced here indicates the average number of common ARC Joiners in each 625 X 625 or 664 X 664 matrixes of Facebook Caus es that ARC Joiners belonged to.
64 Table 7 : Relational Density of Facebook Causes Before and After Disaster Ties Relations Before After > 0 relation 0. 0 985 0. 0726 > 1 relations 0. 0085 0. 0053 > 2 relations 0. 0017 0. 0014 > 3 relations 0. 0005 0. 0005 Note: The ARC Cause was excluded from the calculation of the density of Facebook Causes
65 Chapter 7 : Conclusion There are two undeniable realities for emergency response organizations to take into consideration going forward with their disaster relief effort 1) the magnitude and impact of disasters are often unpredictable and require speedy response to fulfill basic humanitarian needs for disaster victims; and 2) the users of online s ocial networks (whether it is Facebook or another platform) will likely continue to grow exponentially in size and use the platform of online social networking sites as a real time communication tool with other online users .Given the unpredictable nature of natural and manmade disast ers, the capability to expedite the mobilization of resources for disaster relief is critical in saving lives and helping the disaster victims with the response/recovery process. Therefore it is crucial for disaster relief organizations to target their recruitment effort of these resources by leveraging the online social networks to activate interpersonal or organizational ties from the disaster communities globally. In order to fill the gap of research on the relational ti es of the disaster communities organized through the social networks online, the analyses from Chapters 4 to 6 presented empirical evidence of unique differences of the Facebook members joining the ARC Cause (an online platform for disaster community) befo re and after the Haiti earthquake disaster. The results from this thesis research are not directly comparable to the research findings from McAdam et al on social movement analysis or Hurlbert et al on utilizing networks for social resource provision in the context of a disaster, but they do provide insights into the predictors of successful recruitment to social movement s specifically disaster relief efforts based on network composition and structure
66 Network S ize D oes M atter C omparing the sampled joiners of ARC Cause before and after the Haiti earthquake disaster reveals that the average membership size of the Facebook Causes that sampled joiners belonged to after the disaster is 48% greater than before the disaster (500,700 members after the disaster compared to 337,756 members before the disaster). This finding indicates the Facebook Causes with larger membership sizes can potentially be more successful in mobilizing their resources to recruit other Facebook members to their cause. Additionally, analyzing the distribution at the incremental level reveals that the sampled joiners after the disaster also belonged to more Facebook Causes than they did before the disaster As illustrated in Figure 11, 41% of the ARC Joiners sampled after the disaster belonged to 11 or more Facebook Causes compared to only 27% of the ARC Joiners sampled before the disaster. This is consistent with the finding s of McAdam (1986) who reported that participant s of the Freedom Summer project (high cost/high risk form of activism) had a greater number of organizational affiliations than the non participants (i.e., those accepted to the project but later withdrew) As my analyses reveal, Facebook members who joined the ARC Cause (low cost/low risk form of activism) after the disaster may share similar characteristics in support of a specific cause. Linkage B etween I dentity and A ction The sampled joiners of ARC Cause after the disaster belonged to twice as many Facebook Causes related to disaster relief than sampled joiners before the disaster (4% of Facebook Causes related to disaster relief after the disaster compared to 2% before the disaster). Similarly, the ARC Joiners sampled after the disaster also belonged to more Facebook Causes related to philanthropy than sampled joiners before the disaster (50% of Facebook Causes related to philanthropy after the disaster compared to 45% before the disaster). For the Freedom Summer project, McAdam and Paulsen (1993)
67 al so reported the participants not only belonged to more organizations but ones such as civil rights organizations, teacher associations, etc. the Facebook members who joined the ARC Cause after the disaster can also be compared in the same vein in linking their identity with action through salience of organizational affiliations Mobilizing W omen O nline for S upport P rovision Among the sampled joiners of the ARC Cause women dominated the network after the disaster by a ratio of four to one, compared to a ratio of three to one before the disaster The Haiti earthquake disaster pulled substantially more women to the social networks online for support provision, a finding that is contrary to the results of past research on differential participation in d isaster communities and the results of Haines and Hurlbert (1996) who reported that higher proportions of men provide support in disaster relief compared to women. It is likely that women can now leverage online social networks to provide informal su pport of disaster relief effort s, something that was not possible in the physical sense prior to the popularity of social networks online. In fact, Figure 15 shows the differential participation of women over men in the general population of Facebook in t he U.S. (56% women compared to 44% men) T he disaster pulled an even greater proportion of women into disaster relief networks, with 68% of women joining the ARC Cause before the disaster and 79% of women joining after the disaster. Stronger E ffect of I nt erpersonal T ies for R ecruitment The sampled joiners of ARC Cause are more connected to each other after the disaster than the sampled joiners before the disaster. When analyzing the relational ties of the ARC Joiners after the disaster at the increment al level, they consistently have more relational ties from greater than one relation to greater than four relations as detailed in Table 3. However, when analyzing the relational ties of the Facebook
68 Causes before and after the disaster, the causes before the disaster consistently have more relational ties from greater than one relation to greater than four relations as detailed in Table 6. This indicates a heightened (or more effective) recruitment through activating interpersonal ties among ARC Joiners sampled after the disaster rather than organizational ties among Facebook Causes. Since there is a slightly higher number of Facebook Causes that sampled joiners belonged to after the disaster compared to before the disaster (664 Facebook Causes after the disaster compared to 625 before the disaster), this could potentially impact the number of relational ties to consider in the calculation. News T ravels F aster in a D ense N etwork The sample of ARC joiners after the disaster has a greater density of relational ties at each incremental level compared to the joiners sampled before the disaster. According to Fernandez and McAdam (1988), the density of a network can predict participation, and moreover, the density of a network can offer insights in to the speed of information diffus ion among its members. This suggests that ARC Joiners may be recruited faster after the disaster as a result of their participation in a denser network with more joiners connected to each other. On the other hand, the Facebook Causes of the ARC Joiners before the disaster has greater density of relational ties at each incremental level (though the differences become more minimal with the higher level of increments). This should also take into consideration the higher number o f Facebook Causes in the sample after the disaster as discussed in the previous section. This study was conducted in hopes of offering insights to emergency response organizations such as the American Red Cross on tailoring their recruitment effort of act ivists and expediting the mobilization of readily available resources (in particular online) for emergency assistance in a moments notice. While the findings from this
69 thesis research may not be directly comparable to the historical findings cited in ear lier chapters, the differences between the two samples selected before and after the Haiti earthquake disaster are mostly consistent on recruitment factors based on number of organizational affiliations, linkage to identifiable recruitment communities, int erpersonal ties and density of network. There is also unique distinction for highly differential participation of female Facebook members joining the ARC Cause compared to historical support for stronger male presence in the disaster community. Limitation s and Future Research T he data collected for this study can be considered more observational than self reported (except for the gender of the ARC Joiners was self reported from their profile). A follow up study to further explore the motivation to partic ipate (such as attitudinal affinity, personal relations or organizational affiliations) and other attributes (such as age, education, income, etc.) can confirm the conclusions based on the observational data. Another limitation is the data collection period of the attributes and organizational affiliations occurred after four months (April 14, 2010 to April 18, 2010) when the Facebook members first joined the ARC Cause in January, 2010. Their organizational af filiations for other Facebook Causes they may have joined after they joined the ARC Cause may differ from when they first joined the ARC Cause. Unfortunately, t here was no way to differentiate the causes joined after becoming an ARC Joiner compared to bef ore they joined the ARC since the date on which they joined is not indicated by the Causes application on Facebook Given the wealth of administrative data available through Facebook future research can explore how the larger Facebook Causes may join for ces in mobilizing their resources in recruiting others for disaster relief effort (i.e., bloc recruitment) Another area to explore further is how Cause profiling (i.e., targeting the Facebook
70 members belonging to greater proportion of disaster relief or philanthropic causes) may be leveraged as the stronger ties in recruiting others to join their cause. This can potentially maximize the effort for disaster relief organization to mobilize their resources for recruitment in the event of a disaster. Furt hermore, knowing that information can spread quicker in a more dense network, fundraising effort can also be targeted to these Facebook Causes in order to raise donations needed for disaster victims in a speedy manner. Although this thesis research only s cratched the surface of the data mining opportunities possible with observational data alone, it does reveal encouraging findings of a beneficial partnership between emergency response organizations and online social networks working towards a common cause of helping to rebuild the lives of disaster victims.
71 References American Red Cross. (2010, April 17). Facebook Homepage. Retrieved from http://www.facebook.com/press/info.php?statistics#!/redcross?ref=mf American Red Cross. (2010, April 17). Facebook Causes Homepage. Retrieved from http://www.causes.com/causes/397 Associated Press (2010, February 10). Haiti Raises Earthquake Toll to 230,000. The Washington Post. Retrieved from http://www.washingtonpost.com/wp dyn/content/article/2010/02/09/AR2010020904447.html Barnes, S. (2004). Issues of Attribution and Identification in Online Social Research. In Johns, M. D., Chen, S. S., & Hall, G. J. (Eds.), Online Social Rese arch: Methods, Issues, & Ethics (pp. 203 222). New York: Lang. Borgatti, S., & Everett, M. (1997). Network Analysis of 2 Mode Data. Social Networks 19, 243 269. Borgatti, S., Everett, M., & Freeman, L. (2002). Ucinet for Windows: Software for Social Network Analysis. Harvard, MA: Analytic Technologies. Borgatti S ., Mehra, A ., Brass, D., & Labianca, G. (2009 ). Network Analysis in the Social Sciences Science, 323 892 895. Christians, C. & Chen, S. (2004). Introduction: Technological Environments and the Evolution of Social Research Methods. In Johns, M., Chen, S., & Hall,
72 G. J. (Eds.), Online Social Research: Methods, Issues, & Ethics (pp. 15 23). New York: Lang. Davis, A., Gardner, B., & Gardner, M. (1941) Deep Sout h. Chicago: The University of Chicago Press. Erikson, K. (1994). A New Species of Trouble: The Human Experience of Modern Disasters. New York: Norton. Facebook. (2010 March 5 ). Press Room Statistics. Retrieved from http://www.facebook.com/press/info.php?statistics Facebook. (2010 March 5). Application Directory. Retrieved from http://www.facebook.c om/apps/directory.php Faust, K., Willert, K., Rowlee, D., & Skvoretz, J. (2002). Scaling and Statistical Models for Affiliation Networks: Patterns of Participation Among Soviet Politicians During the Brezhnev Era. Social Networks 24, 231259. Fernandez, R. & McAdam, D. (1988). Social Networks and Social Movements: Multiorganizational Fields and Recruitment to Mississippi Freedom Summer. Sociological Forum 3, 357 382. Haines, V., Hurlbert, J., & Beggs, J. (1996). Exploring the Determinants of Support Prov ision: Provider Characteristics, Personal Networks, Community Contexts, and Support Following Life Events. Journal of Health and Social Behavior, 37, 252 264. Hanneman, R. & Riddle, M. (2005). Introduction to Social Network Methods. Retrieved from http://faculty.ucr.edu/~hanneman/ )
73 Hurlbert, J., Haines, V., & Beggs, J. (2000). Core Networks and Tie Activation: What Kinds of Routine Networks Allocate Resources in Nonroutine Situations? American Sociologi cal Review 65, 598618. Kirschenbaum, A. (2004) Chaos Organization and Disaster Management. New York: Marcel Dekker, Inc. McAdam, D. (1986). Recruitment to HighRisk Activism: The Case of Freedom Summer. The American Journal of Sociology 92, 6490. McAda m, D. & Paulsen, R. (1993). Specifying Relationship Between Social Ties and Activism. The American Journal of Sociology 99, 640 667. Nilson, L. B. (1985). The Therapeutic Community: The Typical Victim Response and Its Variations. Policy Studies Review 4( 4), 689 697. Petersen, C. (20 10, February 25 ). Facebooks Ivy Origins Matter. The New York Review of Books Retrieved from http://www.nybooks.com Smith, A. (2010, July 9 ). Haiti Donations: 1.3 Billion. CNN Money. Retrieved from http://money.cnn.com/2010/07/09/news/international/haiti_donation/index. htm Tierney, K. (2007). From the Margins to the Mainstream? Disaster Research at the Crossroads. Annual Review of Sociology, 33, 503525. Tierney, K., Bevc, C., & Kuligowski E. (2006). Metaphors Matter: Disaster Myths, Media Frames, and Their Consequences in Hurricane Katrina. The Annals of the American Academy 604, 57 81.
74 Voorhees, C., Vick, J., & Perkins, D. (2007). Came Hell and High Water: The Intersection of H urricane Katrina, the News Media, Race and Poverty. Journal of Community & Applied Social Psychology 17, 415 429. Wasserman, S., & Faust, C. (1994). Social Network Analysis: Methods and Applications. New York, NY: Cambridge University Press.
About the Author Jennie Wan Man Lai was born in Hong Kong and earned a B.S. degree in Multinational Business Operation, Marketing with specialization in Marketing Research, and Finance from the College of Business at Florida State University. She has been working in the field of survey research methodology for The Nielsen Company since 2001; and experienced in the areas of media and consumer research. She is also active in participating and presenting at the Amer ican Association of Public Opinion Research (AAPOR)