Pipes or Prisms? Personal Networks, Network Mechanisms, and Formal Support Receipt In The Wake Of Deepwater Horizon Oil Spill

Abstract

Social networks are commonly discussed in reference to processes of disaster recovery but rarely explicitly measured. We employ a mixed-methods approach drawing upon the personal-network data of 265 oysterworkers in the wake of the Deepwater Horizon oil spill and qualitative accounts of individual experiences during the recovery process. We find evidence of two potential mechanisms linking network structure with the receipt of formal support: a networks-as-pipes approach linking networks and access to relevant information in the wake of a disaster and a networks-as-prisms approach where networks signal their social identities, shaping post-disaster actions and behaviors.

How do people use social networks to recover from a disaster? Though disasters researchers frequently invoke networks as a mechanism in the recovery process (e.g., Adger et al. 2005; Aldrich 2012, 2019; Aldrich and Mayer 2014), there have been few network-analytic treatments of how individuals use network ties during the recovery process (for exceptions, see Aldrich 2019; Beggs, Haines, and Hurlbert 1996a, b; Hurlbert, Beggs, and Haines 2001; Hurlbert, Haines, and Beggs 2000). Personal network analysis enables researchers to study how individuals’ personal network features are associated with personal outcomes following a disaster. In this article, we explore how the composition of an individual’s personal network is related to which recovery strategies they pursue. We consider the association of personal-network variables with disaster victims’ use of two formal support programs established following the 2010 Deepwater Horizon oil spill: (1) an economic-loss claims program for damages and loss of income incurred and (2) a work program that paid available laborers to remove oil from the gulf.

We borrow and modify an existing set of metaphors for understanding how networks affect market outcomes — networks-as-pipes and networks-as-prisms (Podolny 2001) — and apply them to the study of disaster recovery. The networks-as-pipes metaphor stresses the capacity of network forms and patternings of network ties to retrieve important social resources, with some personal networks featuring properties that provide distinct advantages in terms of amassing resources. For example, weak ties are more effective than strong ties at providing novel sources of information (Granovetter 1973). The networks-as-pipes metaphor is well represented in the disasters literature by the work of Beggs, Haines and Hurlbert (1996a), who argue that structural and compositional properties of individuals’ personal networks confer relative informational advantages in the wake of disasters. In short, this metaphor highlights information retrieval as a mechanism explaining the relationship between network structure/composition and formal support receipt outcomes.

In contrast, the networks-as-prisms metaphor highlights how network variables communicate identity commitment. Network ties operate as “informational cues” that convey the identities and appropriate role behaviors of network-embedded actors (Podolny 2001). Here, individuals develop and use network relations to express membership within specific social groups, such that the quantity of social group members to whom one relates the individual’s attachment to that group (Stryker 1980; Walker and Lynn 2013). Following a disaster, individuals use their network connections to coordinate expectations for what post-disaster strategies are appropriate and convey commitment to particular lines of action (Chamlee-Wright 2008). The networks-as-prisms metaphor highlights social-identity commitment, observable via network relationships, as a mechanism explaining the relationship between network composition and formal support receipt outcomes.

Following disasters, formal recovery organizations are frequently activated or created to aid individual and community recovery (Drabek 1986). When organizations’ allocation of valued resources seemingly violates normative principles of fairness and equity, these organization-sponsored programs can be a source of feelings of confusion, powerlessness, and demoralization and can cause strife and conflict within the larger community (Mayer, Running, and Bergstrand 2015). Consequently, rates of participation in these formal programs can often vary by disaster and community context. The networks-as-pipes metaphor suggests that network variables are associated with differences in participation because individuals have unequal access to relevant information by virtue of the structure and composition of their networks. The networks-as-prisms metaphor suggests that this association exists because individuals, occupying different parts of the larger community network, interpret participation in support programs in divergent, potentially antagonistic ways. Our data on the recovery from the Deepwater Horizon oil spill allow us to compare participation rates in two different programs, creating an opportunity to assess how the qualities and organization of each program affect how different network mechanisms operate during the process of recovery.

We employ a mixed-method design, using quantitative methods to establish the existence of a quantitative pattern and qualitative data to elucidate which network mechanisms best account for the pattern. First, we use personal-network data to demonstrate an empirical association between personal-network metrics and the probability of successful participation in two formal support programs. Second, we use qualitative analyses using data collected through field observations and interviews to assess the plausibility of the posited mechanisms to explain the observed network associations. Taken together, these quantitative and qualitative analyses enable insight into how responses to institutionally provided post-disaster opportunities depend upon these programs’ meaning within local network constituencies.

Social Networks and Disasters

Researchers have long examined how dislocation from sources of social support (Erikson 1976) and how social isolation (Klinenberg 2003) inhibit both short-term survival strategies and long-term recovery efforts. As Elliott and colleagues (2010) observe, all disasters are inherently local because they, by definition, overwhelm local capacities for coping and managing that particular crisis. By this logic, disaster victims are forced to make due with resources that they can generate informally from friends, family, and other members of one’s own personal social networks until more sources of formal support might become available through state or nonprofit-based organizations such as the Red Cross (Beggs, Haines, and Hurlbert et al. 1996b). These vital social networks are most often treated as sets of interpersonal and organizational relations that are both individually shaped and utilized and socially structured (Granovetter 1985, Sewell 19992).

Examination of social networks’ role in disaster recovery has most explicitly occurred in disaster researchers’ application of the concept of social capital, which has been associated with individuals’ and communities’ preparedness, survival, resilience, and recovery (Aldrich 2012, 2019; Chamlee-Wright and Storr 2011). While conceptualizations of social capital vary based on analytical focus, most tend to view its function as socially embedded resources that provide privileged access to useful resources that can positively shape post-disaster outcomes (Elliott, Haney, and Sams-Abiodun 2010; Gill et al. 2012). Disaster recovery is therefore dependent on community’s capacity to re-establish fragmented networks and create new ones, often anchoring institutions central to the recovery process (Storr 2008). We are often limited to qualitative accounts of the effects of these networks, and their disruptions following disaster, as large-scale quantitative investigations of social network ties are limited to a few national panel datasets that are ill-suited to examining regional disaster responses.

Our objective in this article is to remedy this oversight through the utilization of social network analysis involving patterned sets of personal social relations. Most commonly in the social sciences, network analysis involves the study of individuals’ relationships with one another. In this article, we employ personal network analyses, or the comparative analysis of networks of contacts (or alters) that form around particular subjects (or egos). As contemporary sociological research calls for greater articulation between the personal utilization of social networks and the cultural context in which they function (e.g. Breiger and Puetz 2015; Small 2017; Vaisey and Lizardo 2010), we also examine qualitative data on personal decision-making post-disaster and employ a mixed-methods approach to assess how people use their personal networks to recover. Specifically, we examine the influence of social networks utilizing two potential logics of network function: networks as pipes and networks as prisms.

Networks as Pipes

The networks-as-pipes metaphor regards network structure as “channels” or “conduits” through which contents, such as information, flow (Podolny 2001). Since Granovetter’s seminal work (1973), network researchers have demonstrated that network properties affect social ties’ capacity to retrieve information. Building upon this research, Lin (2001:41) defined social capital as “resources embedded in a social structure that are accessed and/or mobilized in purposive actions.” Focusing on agents’ capacities to mobilize resources embedded within network ties, Lin theorized personal network development as an “investment” that enables future returns in the form of social resources (Small 2011). These social resources are differentially accessible to individuals according to the organization and composition of their personal networks.

In an important series of articles (Beggs, Haines, and Hurlbert 1996a, b; Hurlbert, Beggs, and Haines 2001; Hurlbert, Haines, and Beggs 2000), Beggs, Haines, and Hurlbert drew upon social-capital theory precursors to explain differences in the receipt of resources, either institutionally provided (formal support) or informally offered (through social relationships) after Hurricane Andrew. Seeking to explain their respondents’ differential use of formal support programs, they focused upon informational advantages — “the transfers of information through and by interpersonal environments” (1996b:204) — as the primary mechanism of interest. Likewise, Elliott, Haney, and Sams-Abiodun (2010) examine access to both informal and formal sources of aid and find that the sequencing of a disaster matters, where inequalities in the availability of social resources increase over time.

We consider two personal-network characteristics associated with improved information retrieval. First, increased network heterogeneity contributes to increased novelty of information. Although weak ties are empirically associated with positive outcomes in terms of information retrieval, the individual’s bridging a “structural hole” actually provides the observed benefit by accessing “nonredundant” sources of information (Burt 1992). By virtue of having ties to individuals who are unconnected to one another, an individual is likely to enjoy informational advantages relative to others (Lin 2001). Because individuals tend to be connected to those who resemble themselves (McPherson, Smith-Lovin, and Cook 2001), measures of network heterogeneity indicate an individual’s opportunity to connect to dissimilar parts of the social structure and, accordingly, her capacity to retrieve nonredundant information.

Second, the presence of high-status network members contributes to improved quality of information. Lin (2001) theorizes that some network members are likelier than others to occupy information-rich positions within formal organizations. The ability to access such individuals via one’s personal network is helpful for the retrieval of high-quality information. For example, men more frequently occupy information-rich positions within the community in comparison with women. Accordingly, those whose personal networks are more likely to contain such individuals will enjoy relative advantage over those whose personal networks do not.

Networks as Prisms

In contrast, the networks-as-prisms metaphor regards social networks as sense-making tools that signal, reinforce, or otherwise communicate the identities of social entities. Personal networks are “projects” that people transform over the course of their lives with respect to their changing interests and identities and to express which of their social and categorical memberships are personally salient (Wellman, Wong, Tindall, and Nazer 1997). An individual’s personal network communicates an implicit “sense of one’s place” (Martin 2003). As prisms, networks “split out” and “induce differentiation among actors” (Podolny 2001) such that different parts of a network are associated with different social identities and associated strategies of action.

Disasters fundamentally cause disruptions to everyday strategies of action, requiring individuals to reconstruct the quotidian while simultaneously incorporating changes to practical expectations shaped by post-disaster realities (Erikson 1976). Faced with such uncertainty, people use networks to signal to one another their anticipated recovery strategies and develop a working consensus of appropriate post-disaster behavior (Aldrich 2012; Chamlee-Wright 2008). Identity theory provides a vocabulary for understanding how associations between personal-network structure and action develop through the mechanism of identity commitment. Identity theorists describe “immediate social networks as contexts” that facilitate hierarchies of salience among the multiple identities constituting the individual’s self (McFarland and Pals 2005; Stryker 1980). Grounding certain social identities as salient, personal networks can engender social and emotional commitments to particular behaviors and norms that render alternative lines of action untenable or dangerous (Coleman 1988; Portes 1998).

Previous disasters research has stressed the importance of identity for understanding how individuals adopt post-disaster strategies of recovery. Fothergill (2003) found that moral identifications as “self-reliant” are negatively associated with seeking formal social support. Similarly, Clarke and Mayer (2017) argue that identities based in local place, heritage, and moral self-understanding are symbolic resources that individuals use to select and justify particular patterns of action in the wake of disasters. Network-theoretic applications of identity theory suggest that individuals’ self-understandings are likely to have personal-network correlates such that an individual’s identity commitment can be “operationalized in concrete network terms” (Walker and Lynn 2013:170). We utilize a basic measure of identity salience through measurement of the quantity of alters within an individual’s network that matches their own — i.e., “the number of others to whom one relates through occupancy of a given position” (Stryker 1980:81). Through our mixed-methods approach, we aim to refine the understanding of how networks shape decision-making by activating particular identities as relevant and shape, redirect, and constrain lines of action consistent with these identities.

Methodology

On April 20, 2010, the Deepwater Horizon oilrig suffered a catastrophic explosion. Owned and operated by the offshore drilling company, Transocean, and leased by the multinational oil company, BP, the oilrig eventually sank 40 miles off the Louisiana coast and left an uncapped wellhead a mile underneath the surface of the Gulf of Mexico. By the time it was repaired 87 days later, the wellhead had released an estimated 185 to 205 million gallons of oil, resulting in the worst maritime oil spill in global history (National Commission 2011).

As part of its federally-mandated economic recovery program, BP created an economic-loss claims program and began accepting claims only weeks after the oil spill began. From April 25 through August 23, 2010, BP received 154,000 claims and wrote 127,000 checks covering $399 million in estimated damages (BP 2010a). The Gulf Coast Claims Facility (GCCF), an independent agency established three months after the spill, issued $2.6 billion in payments to 170,000 claimants (GCCF 2012). The GCCF operated for 18 months, distributing $6.2 billion to over 220,000 claimants (GCCF 2012). The psychological effects of the spill were significant, with increasing rates of mental health problems such as depression, anxiety, and stress reported across the region (Grattan et al. 2011; Morris et al. 2013).

BP also provided employment opportunities in the Gulf region. After the spill, BP created the Vessels of Opportunity (VoO) program to hire fishing and commercial crews to contain and remove the oil. If selected to participate in VoO, an individual could earn between $1,200 (for smaller boats) and $3,000 (for larger boats), with crewmembers being paid $200 per day (BP 2010b). Requirements for the program included passing a four-hour training course, a Coast Guard safety inspection, and crewing requirements based on vessel size. Over the course of 2010, the VoO program made approximately $594 million in payments to vessels and crew and employed approximately 3,500 commercial and charter fishing boats (Upton 2011). The two programs were not mutually exclusive.

That multiple recovery programs were established in the wake of Deepwater Horizon provides a unique opportunity to study differences in how networks contribute to participation in programs, depending upon the features of those programs. We contend the claims program and VoO differ in two important respects. First, the programs varied in their allocation procedures. All individuals who were financially affected by the disaster were eligible to participate in the claims program, whereas the VoO program was marked by a relative scarcity of positions. Only 3,500 commercial and charter fishing boats were selected for participation in the work program across the entire gulf. Second, the claims program and VoO vary in whether they require participants to engage in labor. Whereas the claims process represented “restitution” for losses suffered, the VoO program was amenable to framing as “remuneration” for services rendered. These differences led to divergences in how respondents evaluated the programs and shared relevant information.

Data Collection and Sampling

To understand how social capital shaped the trajectories of personal post-disaster recovery, an in-depth investigation of recovery from the DWHOS was conducted from 2011–2015 in Franklin County, Florida. Franklin County is located along Florida’s panhandle and is home to 11,000 individuals, twenty percent of which live below the federal poverty threshold. Apalachicola Bay has historically been the primary economic driver of the region, a draw for moderate levels of tourism and the natural estuarine habitat for an oystering industry that historically has supplied 90% of Florida’s oyster harvest and 10% in the U.S. (UF IFAS 2013). Oystering in the bay has consistently provided economic security in a region lacking other employment opportunities. Oystering in Franklin County is conducted by hand-tonging often in pairs, or individually, using small boats that require comparatively little investment and experience to operate. In 2012, approximately 800 individuals in Franklin County possessed licenses for the harvesting of oysters (UF IFAS 2013). To compensate for lost work opportunities, all licensed oysterworkers were eligible for economic assistance from BP and the GCCF, but to our surprise, not all eligible oysterworkers sought out this recovery aid.

To understand how this vulnerable population used social networks to manage the disaster, our research team partnered with Franklin’s Promise Coalition, a nonprofit organization dedicated to addressing local social needs, to conduct social network surveys in the spring of 2013, as well as stakeholder interviews and focus groups, with oysterworkers affected by the disaster. A total of 265 participants were recruited between January and May 2012 through two methods of direct advertising. First, oysterworkers who visited Franklin’s Promise Coalition’s offices to obtain various social services including food assistance, crisis counseling, and workforce development were personally invited to participate. Secondly, the local seafood worker association elicited participation within the broader oysterworker populations to recruit individuals potentially not participating in the recovery programs through emails, fliers, and word-of-mouth. If potential respondents were interested, fieldworkers at Franklin’s Promise verified their eligibility to participate in the study by reviewing their state-issued oyster-harvesting license, residence in Franklin County, and being over eighteen years of age. Once their eligibility was verified and the informed consent procedures approved by the University of Arizona Human Subjects Protection Program complete, respondents participated in a computer-assisted social network surveys utilizing the Egonet software program (McCarty 2003).

Our qualitative data come from nearly one hundred hours of participant observation and related field notes, six focus groups of 6–10 participants representing a particular economic sector or vulnerable social group, and 37 semi-structured interviews. Participant observations included attendance at regional meetings, community and activities such as county board of commissioner meetings, nonprofit activities, and cultural events, as well as everyday activities such as the morning launch and afternoon return of the oystering boats. Key stakeholder interviews were conducted with individuals representing a variety of business, political, and civic interests in Apalachicola, with an emphasis on workers in the seafood and tourism industries. The network and interview samples do not intentionally overlap. However, these interviews provide vital insights into both the individual strategies utilized by network survey participants and how others within the community viewed and influenced those strategic decisions. Interviews were digitally recorded at the time of collection and later professionally transcribed. We removed personal names, places of employment, and other identifying features to maintain respondent confidentiality.

Quantitative Variables and Analyses

Our survey instrument collected information on respondents’ personal networks via a name-generator and name-interpreter instrument (Marsden 2005). Personal-network name generators are effective tools for analyzing the embeddedness of the individual within a multiplicity of contexts because they set no a priori boundaries on which alters can be selected. Our survey instrument required each respondent to list 30 individuals, or alters, living outside the respondent’s household, who had had contact with the respondent in the last two years, and who might be called upon for some form of help. Respondents were given prompts to assist them in identifying the required 30 alters, if needed, by asking them to recall disaster-like events that required them to seek assistance. The survey instrument next asked respondents questions regarding whether they had participated in the claims process or the VoO program. Finally, the instrument asked respondent to provide information regarding each of alters, including information on which specific social resources the alter could provide during a disaster.

There are limitations to egocentric network data. First, researchers must decide whether to place artificial limits on the number of alters respondents select. Placing limits can compromise the accuracy of respondents’ personal networks in that they might otherwise be larger or smaller. However, limits can enable better comparisons among respondents, as there may be many potential reasons for individuals to stop listing alters and a set limit requires that all respondents expend the same effort (McCarty and Molina 2015). Second, previous researchers report substantial respondent burden resulting in inaccuracies (White and Watkins 2000). However, because we focused on a single site, we were able to carefully assess the accuracy of respondents’ answers through review and consultation with our local collaborators. If discrepancies occurred, particularly when describing alters, those records were removed them from the dataset.¹

Dependent variables.

For our analysis, we provide two measures for receipt of formal recovery support following the BP oil spill. We utilize two dichotomous dependent variables, one capturing respondents’ participation within BP’s VoO program and another capturing respondents’ participation in the claims process. For these variables, respondents are characterized as participating (1) or not participating (0) in each program. Descriptive statistics (mean, standard deviation, and range) can be found in Table 1.

Table 1.

Descriptive Statistics for Variables in Analysis

Variable	Mean	S.D.	Min	Max	N
Dependent Variables
Participation in VoO	0.219	0.414	0	1	265
Participation in Claims	0.687	0.465	0	1	265
Personal Characteristics
Male	0.653	0.477	0	1	265
Age	44.38	13.84	18	76	265
High School	0.468	0.500	0	1	265
Married	0.532	0.500	0	1	265
Children	2.223	1.658	0	8	265
Time in Community	37.65	16.64	1	76	265
Social Capital Variables
Proportion Alters in County	27.31	4.287	2	30	265
Proportion Alters Male	19.43	4.926	6	30	265
Proportion Industry	8.294	6.814	0	30	265
Age Heterogeneity	13.86	2.442	2.236	22.11	265
Religious Congregation	0.306	0.462	0	1	265
Business Association	0.0792	0.271	0	1	265
Social Support Index (+)	16.04	7.835	0	30	265
Small Loan (+)	16.38	9.480	0	30	265
Stay (+)	16.09	8.691	0	30	265
Confide (+)	16.43	8.677	0	30	265
Solve Problems (+)	14.60	8.702	0	30	265
Leader (+)	14.12	8.558	0	30	265
Information (+)	18.64	8.902	0	30	265
Social Support Index (−)	7.882	7.026	0	29.83	265
Small Loan (−)	8.328	8.224	0	30	265
Stay (−)	8.868	7.826	0	30	265
Confide (−)	7.887	7.747	0	30	265
Solve Problems (−)	7.909	8.042	0	30	265
Leader (−)	8.449	8.084	0	30	265
Information (−)	5.853	7.122	0	30	265
Alter Count Maybe	7.592	6.524	0	28	265
Alter Count Yes	5.509	5.857	0	30	265

Network variables.

We examine the effect of personal social networks on the receipt of formal recovery support across five distinct dimensions of network composition. Three of these dimensions are count measures, constructed by adding the number of alters characterized by the respondent as belonging to a particular category: (1) number of male alters, (2) number of alters living in Franklin County, and (3) number of alters whom ego met through an oystering work context. We hypothesize that as the number of male alters, local alters, and oystering alters increase, so does the quality, but also the redundancy, of information about the two recovery programs increase (see Table 2). We also include a measure of age heterogeneity within ego’s personal network, which is likely to increase the diversity of information available to the respondent as older network members are expected to have greater access to information due to more prestigious roles in Franklin County. More importantly, the description stresses the diversity of information, while this explanation suggests that older people would have better information. A better measure in this respect would be average age. Age heterogeneity is measured via the standard deviation of alter ages within ego’s personal network, per previous work (Marsden 1988). Finally, we asked respondents whether they believed that each network alter had participated in the claims process. Respondents could answer no, yes, or that they were not sure. We generated two count variables (0–30), corresponding to the “maybe” and “yes” answers, summing the number of alters ego characterized each way. We briefly discuss how each variable relates to the network-as-pipes versus networks-as-prisms mechanisms in Table 2.

Table 2.

Variables and Mechanisms

Variable	Networks-as-pipes mechanism	Networks-as-prisms mechanism
Number male alters	Indicator of higher-quality information, following expectation that men occupy more information-advantageous positions (+)	---
Age heterogeneity	Indicator of information diversity, following expectation that more heterogeneous personal networks will provide novel sources of information (+)	---
Number of alters living in Franklin County	Indicator of information redundancy, following expectation that more homogeneous personal networks will provide novel sources of information (−)	Indicator of strong self-identification as member of Franklin County community (−)
Number of alters met via work context	Indicator of information redundancy, following expectation that more homogeneous personal networks will provide novel sources of information (−)	Indicator of strong self-identification as member of the oystering industry (−)
Number of alters participating in claims process	---	Use of personal network contacts to assess whether other local individuals are participating in the claims process (+ for yes; − for maybe)

Social support variables.

While there are myriad measures of social support in the psychology literature, our focus on seeking assistance for formal support mechanisms required the creation of a more precise estimate of the potential sources of social support applicable in the context of disaster (Barrera Jr. 1986). We use measures of perceived social support by summing the counts of six resources embedded in personal networks. These resources were: (1) a small financial loan under $500, (2) a temporary shelter for one’s family, (3) a confidant with whom the respondent could share personal problems, (4) a problem-solving skillset, (5) leadership skills, and (6) provision of useful information. The initial survey questions required respondents to represent the likelihood that an alter could provide each resource on a ordinal scale, allowing us to frame social support in two ways: (1) the presence of resources and (2) the absence of resources available to respondents. We provide two indices that are aggregations of the aforementioned six forms of social support: one describing the number of network members who could provide each resource identified as present among his or her 30 alters (Cronbach’s α = 0.9521), divided by six, and another describing the number of network members the respondent identified as being unable to provide each resource (Cronbach’s α = 0.9579), divided by six. These indices feature exceptionally high values for Cronbach’s alpha, which suggest that the resources readily scale into a single measure. We call the first the Social Support Index and the second the Lack of Social Support Index.

Organizational membership variables.

We include two other measures related to social capital. We examine membership in two different types of organization, each type of which is conceptualized as a dichotomous variable whereby membership is coded (1) and non-membership is coded (0). The two types of organization in the analysis are business associations or churches.

Individual-level control variables.

In order to concentrate on potential network effects, our models control for the potential effects of personal characteristics on seeking formal support. First, we control for the effect of gender, which has been demonstrated to significantly shape the personal experience of disaster (Peek and Fothergill 2008), create greater vulnerabilities for women (Fordham 2003), and influence general network use (Flaherty and Richman 1989). Second, we control for the effect of age and marital status, based on findings in the disaster literature that vulnerability increases over the lifecourse (Ngo 2001) and that marital status provides substantial protective effects in the event of crisis (Klinenberg 2001). We also control for the effect of parental status, which also shapes the experience of disaster, with an interval measure assessing an individual’s number of children living at home (Peek and Fothergill 2008). While we did not include an explicit measure of income, as wealth and financial impact was a particularly sensitive topic during the time of the oil spill, we do rely on a measure of educational attainment dummy coded for completing a high school degree or equivalent. With less than two-thirds of Franklin County’s adult population possessing a high school degree in 2010, our dichotomous measure of education is a fairly reliable proxy for socioeconomic status for this region where employment opportunities are largely divided between those requiring a high school degree such as education and public service and those that do not, such as seafood harvesting and processing and service sector jobs (U.S. Census Bureau 2019).

Qualitative Analyses

To determine whether networks-as-pipes and networks-as-prisms operate as mechanisms facilitating the use of social networks in the wake of the Deepwater Horizon oil spill, we complement our quantitative analyses with qualitative data concerning individuals’ perceptions of and experiences with support programs. By reconstructing the cultural and relational context in which people make evaluations of support programs and decisions about whether to provide information about such programs to others, we can better assess which mechanisms pertain in our site.

We draw upon a mixed-methods framework to achieve complementarity (Small 2011). Spillman (2014) argues for a mixed-methods approach in which description provides a “description of sociological patterns” which must be contextualized and explained through reference to qualitative data. Our quantitative data point to observable regularities in the association of network patterns and formal support receipt but do not lend analytic access as to why these patterns exist or what mechanisms are operative. Using these qualitative data, we seek to determine the relevance of two proposed mechanisms potentially affecting formal support receipt: the network-as pipes mechanism of informational redundancy and the networks-as-prisms mechanism of identity commitment. Identification of which network-relevant mechanisms are contributing to the observed differences is important because they implicate different potential solutions to encourage participation and recovery.

We utilize interpretive data regarding a particular problem situation (Gross 2009): Confronted with the prospect of economic vulnerability, oysterworkers made decisions regarding whether to seek formal support receipt in a restitution-based claims program and/or a remuneration-based work program. These programs, as phenomenologically distinct cultural objects, possess unique values and meanings. Individuals could value the VoO work program and claims process differently, and these meanings potentially affected the pursuit of support, use of information, and sharing of information within their networks. Using the QSR NVivo 10 software (QSR 2012), we used open coding to identify emergent themes and topics within the data, followed by a closed-coding scheme to structure the qualitative findings. Both authors conducted multiple rounds of coding, first applying open coding to identify topics based on the disaster impact and recovery literatures (e.g. socioeconomic impacts, physical and mental health, help seeking, frustrations, aspirations) for recovery and a second round of axial coding to apply theoretically informed themes connecting those topics into more structured concepts (Strauss and Corbin 1998). Both authors equally contributed to these processes and regularly conferred to discuss and synthesize findings. The thematic results of the axial coding are presented in the discussion section below.

Results

To discern the effects of different dimensions of social capital on receipt of formal recovery support, we performed two logistic regression analyses to capture the distinct effects of social-network variables on participation in each support program (the claims process and the VoO work program). Ultimately, we decided upon three models for each support program. First, we provide a model containing personal characteristics but no network- or organization-related variables. The next two models contain network variables; in the first we include the social-support index and in the second the lack-of-social-support index.[²]

We begin our analysis of the network influences on participation in the claims program with an empirical puzzle: our sample consists entirely of individuals in the oystering industry who were negatively affected by the oil spill, yet our empirical results show that many (31.3%, n = 83) did not participate in the claims process. In the immediate aftermath of the spill, the barrier to entry was marginal. Appearing at the local claims office, centrally located within Franklin County, could result in a check as large as $5,000. If everybody in our sample was negatively affected by the oil spill and therefore possessed a legitimate entitlement to compensation, then why did some not participate? Given the low barrier to entry for the claims process, which required only basic paperwork including oystering licenses, evidence of economic loss including receipts for expenditures, and past income tax returns, we should expect a smaller percentage of oysterworkers to be administratively ineligible. And while it is possible that a small fraction (<5%) of oysterworkers operated without a license and would thus be ineligible for financial compensation (Hartsfield 2020), this deviant behavior should have also reduced the likelihood of participating in our study and is therefore not a primary concern in interpreting the results.

Table 3 features the models for participation in the claims program.[³] Model 1 features the coefficients only for personal characteristics. We find that age and amount of time the respondent has lived in the community provide competing pressures on the likelihood that he or she participated, with age decreasing the likelihood of participation (b = −0.04, p < .01) and time in community increasing the likelihood (b = 0.03, p < .05). Each one-year increase in age is associated with the respondent being 0.96 times (exp[−0.04]) as likely to participate, and each one-year increase in time in community is associated with the respondent being 1.03 times (exp[0.03]) as likely to participate. Surprisingly, other personal characteristics commonly associated with greater social vulnerability due to the need to provide for dependents, being married and having children, are not significant in any of the models.

Table 3.

Logistic Regression Coefficients for Participation in Claims Program

Independent Variables	Model 1		Model 2		Model 3
Personal Characteristics
Male	0.41	(0.29)	0.52	(0.36)	0.53	(0.36)
Age	−0.04	(0.01)**	−0.03	(0.02)†	−0.03	(0.02)*
High School	0.10	(0.28)	0.14	(0.30)	0.15	(0.30)
Married	0.05	(0.29)	0.01	(0.31)	−0.01	(0.31)
Children	0.02	(0.09)	0.04	(0.09)	0.04	(0.09)
Time in Community	0.03	(0.01)*	0.03	(0.01)**	0.03	(0.01)**
Social Capital Variables
Proportion Alters in County	—		−0.09	(0.04)*	−0.09	(0.04)*
Proportion Alters Male	—		−0.00	(0.04)	−0.00	(0.04)
Proportion Industry	—		−0.02	(0.02)	−0.02	(0.02)
Age Heterogeneity	—		0.01	(0.06)	0.03	(0.06)
Religious Congregation	—		−0.69	(0.33)*	−0.77	(0.33)*
Business Association	—		0.01	(0.54)	0.02	(0.54)
Social Support Index (+)	—		−0.01	(0.02)	—
Social Support Index (−)	—		—		−0.02	(0.02)
Alter Count Maybe	—		−0.05	(0.02)*	−0.05	(0.02)*
Alter Count Yes	—		0.04	(0.03)	0.03	(0.03)
Constant	1.17	(0.55)*	3.60	(1.74)*	3.61	(1.72)*
N = 265
Log-likelihood	−159.41		−152.00		−151.81

Notes: Unweighted data.

^†p < .10

^*p < .05

^**p < .01

^***p < .001 (two-tailed tests)

Models 2 and 3 feature the coefficients for personal characteristic and network- and organization-related variables, with Model 2 including the social-support index and Model 3 including the lack-of-social-support index. Interestingly, neither the provision of social support (Model 2) nor lack of social support (Model 3) for the provision of key resources are significant in our claims program models. Examining the other coefficients from the two models, we see that having a support network consisting of local residents (b = −0.09, p < .05) and belonging to a religious congregation (b = −0.69, p < .05) are negatively associated with participation in the claims process. For each additional alter who is local, an individual is 0.91 times (exp[−0.09]) times as likely to participate in the claims process. Controlling for other factors, an individual who belongs to a religious congregation is less than half as likely to participate in the claims process (exp [−0.69]). There is one more statistically significant variable: individuals who do not know whether others in their support network are pursuing formal aid through the claims process are themselves less likely to participate (b = −0.05, p < .05). That is, for each network member whose participation in the claims process the respondent is unsure of, the respondent is 0.95 times (exp [−0.05]) less likely to participate. Interestingly, there is no statistically significant effect, positive or negative, for actually knowing others who participate.

Table 4 features the models for participation in the VoO work program. Model 4 features the coefficients only for personal characteristics. According to this model, we find that being male (b = 1.05, p < .01) and graduating high school (b = 1.31, p < .001) had strong, positive effects for recruitment. In particular, controlling for other factors, men are 2.86 times (exp [1.05]) as likely as women to participate in the work program, and high school graduates are 3.71 times (exp [1.31]) as likely to participate as non-graduates. The models also consistently demonstrate that being married has a negative effect (b = −0.78, p < .05), but this is accompanied by a positive effect for each child claimed (b = 0.21, p < .05). Unlike Model 3, featuring the claims process, we see here mildly significant effects for the more traditional indicators of social vulnerability, being married and having children, but the effects are ultimately washed out in their oppositional directions to each other.

Table 4.

Logistic Regression Coefficients for Participation in VoO Work Program

Independent Variables	Model 4		Model 5		Model 6
Personal Characteristics
Male	1.05	(0.38)**	1.25	(0.47)**	1.20	(0.47)*
Age	−0.00	(0.02)	−0.00	(0.02)	0.00	(0.02)
High School	1.31	(0.35)***	1.38	(0.37)***	1.43	(0.38)***
Married	−0.78	(0.35)*	−0.80	(0.36)*	−0.94	(0.38)*
Children	0.21	(0.10)*	0.27	(0.10)*	0.29	(0.11)**
Time in Community	0.02	(0.02)	0.02	(0.02)	0.02	(0.02)
Social Capital Variables
Proportion Alters in County		—	−0.01	(0.04)	−0.02	(0.04)
Proportion Alters Male		—	0.01	(0.04)	0.01	(0.04)
Proportion Industry		—	−0.08	(0.03)*	−0.08	(0.03)**
Age Heterogeneity		—	0.02	(0.07)	0.03	(0.07)
Religious Congregation		—	−0.52	(0.41)	−0.58	(0.41)
Business Association		—	1.17	(0.54)*	1.32	(0.56)*
Social Support Index (+)		—	0.03	(0.02)		—
Social Support Index (−)				—	−0.07	(0.03)**
Constant	−3.45	(0.74)***	−3.67	(1.84)*	−2.83	(1.85)
N = 265
Log-likelihood	−123.45		−115.25		−111.99

Notes: Unweighted data.

^†p < .10

^*p < .05

^**p < .01

^***p < .001 (two-tailed tests)

Models 5 and 6 feature the coefficients for personal characteristics and social capital variables. Model 5 includes the social-support index and Model 6 includes the lack-of-social-support index. Using the coefficients from Model 6, we find three social capital variables are associated with participation in the work program: proportion of support relationships originating in a work context (b = −0.08, p < .01), participation in a local business association (b = 1.32, p < .05), and the Lack of Social Support Index (b = −0.07, p < .01). For each support-network relationship that originated in a work context, the respondent is 0.92 times (exp[−0.08]) as likely to participate as somebody whose social network contains no within-industry alters. An individual who belongs to the local union is 3.74 times (exp[1.32]) as likely to participate in VoO as somebody who does not belong. Finally, for each interval increase in the Lack of Social Support Index, an individual is 0.93 times (exp[−0.07]) as likely to participate as someone whose entire support network can provide valuable resources.

Discussion

Unlike previous research projects investigating the impact of formal recovery programs in post-disaster contexts, we are able to examine two distinct programs attempted post-oil spill to compensate for economic losses across the Gulf Coast. The results from our logistic regression models suggest two distinct patterns of potential mechanisms explaining participation in the two programs that fit with our networks-as-pipes and networks-as-prisms perspectives. First, the network-as-pipes metaphor suggests that network metrics associated with access to higher-quality information should predict higher rates of service utilization. For Franklin County, a higher proportion of network members are male, as men are more likely be directly employed as oysterworkers overall and thus should be more immediately familiar with any support programs.

Network metrics associated with informational redundancy, where less diverse networks with more members living in the same county, or higher proportion of network members also being oysterworkers were also hypothesized to be negatively associated with support receipt. Finally, network metrics associated with informational diversity were proposed to be positively associated with formal support receipt. Furthermore, because the networks-as-pipes metaphor posits a social mechanism with a high degree of generality, we might expect that these metrics have uniform effects across different formal support programs.

Our quantitative results lead us to regard informational advantage as a plausible mechanism. We find one indicator of informational redundancy (proportion of alters originating from work contexts) to be negatively associated with participation in VoO at the .05 level. Likewise, we find one indicator of informational redundancy (proportion of alters located in-county) to be significantly associated with participation in the claims process at the .05 level. However, indicators of informational diversity (age heterogeneity) and higher-quality information (proportion alters male) are not significant for either program, nor is any network variable associated with a uniform outcome across support programs.

In contrast, the networks-as-prisms metaphor suggests that network metrics could be negatively associated with participation in formal support programs if, by virtue of having a strong commitment to a context-related identity, the individual faces social pressures to not participate. Franklin County is a majority Evangelical Protestant community, with more than 60% of religious adherents in the county associating with the thirty predominately Southern Baptist congregations located there (Grammich et al. 2012). Certain segments of network structure can discourage participation in formal support programs if network members articulate anti-welfare attitudes, which is true of many Southern Baptist communities (c.f. Fothergill 2013; Hochschild 2018). Anti-BP attitudes included not only blamefor causing the spill, but also for providing “handouts” — or more cynically as “hush money,” which threatened respondents with status loss for participating. We find one indicator of identity commitment (proportion of alters originating from shared work contexts) to be negatively associated with participation in VoO at the .05 level. Likewise, we find one indicator of identity commitment (proportion of alters located in-county) to be significantly associated with participation in the claims process.

The network-as-prisms metaphor also suggests that people use their personal contacts to develop an understanding of what is happening in a larger context. Here we find evidence that individuals who did not know how their personal-network members responded to the opportunity to participate in the claims process were themselves less likely to participate. The claims program was an important post-disaster source of formal support, as it was immediately accessible to assist struggling oysterworkers and promised funds so long as support-seekers could convincingly demonstrate that the disaster had caused them to incur financial losses. With respect to the claims process, our quantitative models found that the proportion of a respondent’s alters who resided in-county was negatively associated with participation. However, it is possible to interpret this result as supporting both networks-as-pipes (information redundancy) and network-as-prisms (identity commitment) mechanisms. Consequently, we rely upon our qualitative data to adjudicate these competing explanations.

Many respondents in the qualitative component of our research project claimed to have abstained from the claims process for moral reasons. Discussing the claim process, respondents frequently provided moral assessments of program participants. For example, one respondent, a charter operator, told us:

One thing that I noticed was that when BP handed out money, everybody had their hand out for BP money. I don’t go along with that… There are always stories, but there were people dumping motor oil in the Gulf to keep the BP money coming. I think that the whole experience shows me how greedy people are and how they would rather have somebody take care of them. We are more self-sufficient; my husband didn’t cancel [work] one day.

In our interviews, respondents commonly associated participants in the claims program with negative attributes, such as greed, irresponsibility, community disloyalty, and lack of self-reliance. This is keeping with previous research in which individuals disavow formal support because participation threatens their self-identifications and reputations as “self-reliant” and “independent” (Fothergill 2003). Gulf coast residents affected by the spill regularly interpreted the claims program as an entitlement program or handout (Mayer, Running, and Bergstrand 2015). Among our respondents, many viewed participation as an act of complicity with the corporation responsible for creating the crisis and as compromising community autonomy and self-determination. Respondents commonly referred to claims compensation as “dirty money” — as a payoff rather than legitimate form of restitution.

Such accounts suggest that the primary mechanism shaping participation in the claims programs in our data is identity-based rather than information-based. The networks-as-prisms metaphor highlights the capacity of social relations to confer social identities (Podolny and Baron 1997) and a perspective from which agents reflexively consider the appropriateness of pursuing lines of action. Together, our quantitative and qualitative analyses suggest that particular network foci (religious organizations) and segments of network structure local to Franklin County became sites for the construction of a semiotic code (Swidler 2001) establishing an association between claims participation and community disloyalty, self-interestedness, and dependence upon morally questionable sources of assistance. By rejecting participation in the claims process, individuals avoided complicity with the institution responsible for the crisis and association with morally unworthy individuals exploiting the disaster for their own private benefit.

Generally speaking, individuals learn about, come to consensual understandings about, and find solutions to problems through network interaction (Pescosolido 1992). From the extant disaster literature, we know that socioeconomic status is an important factor in decision-making and capacity for accessing post-disaster assistance (Bertrand et al. 2004; Peek and Fothergill 2008). Lacking a high school degree can be a significant structural barrier in accessing resources where applications require complex forms and calculations (Grube et al. 2018). Although our quantitative findings identify having a high school degree as significantly correlated with utilizing the VoO program and not the claims process, this pattern conforms with our qualitative findings that the claims processes were initially advertised as being available to everyone impacted by the spill while the VoO program had explicit requirements such as owning a vessel, licensing, and participation in training. We interpret the significance of possessing a high school degree as an indicator of economic class and therefore greater financial impact and thus motivation for participation in the two recovery programs.

Our quantitative results also suggest that, for each alter about whom ego does not have knowledge regarding claims process participation, ego is 0.95 times (exp[−0.05]) as likely to participate. Correspondingly, we find that lack of knowledge of others’ post-disaster activity dampens claims participation. Negatively charged emotions about the claims program made it difficult to achieve accurate understandings of how others were acting. If some individuals experienced difficulties identifying who had and had not participated, we suspect this lack of knowledge is rooted in community members’ concerted attempts to avoid the negative connotations some assigned to claims participation. Some interviewees confirmed that, due to widespread negative perceptions of the claims process, they avoided it as a topic of conversation:

I worked really hard and was compensated pretty well for it — but it was very isolating because there was such animosity in the community that you just didn’t want to go out because you didn’t know who was angry about what. I’m not much of a socializing person, but we really did not go out in public at all during this whole thing. It was just difficult to get into a social conversation about BP. You just didn’t want to talk about it anymore. I had a job to do and I was tired of apologizing for taking BP compensation and it was just really isolating.

The claims program’s air of controversy inhibited oysterworkers’ ability to construct a working consensus about which routes to recovery were appropriate to pursue, with some people hesitant to open themselves up to criticism by admitting participation and some avoiding claims as a topic of conversation to avoid exacerbating already inflamed tensions within the community. Among some, this contributed to a general uncertainty about how to proceed within the recovery process and resulted in decisions to forgo participation in the claims program despite clear economic need.

The VoO program provided work opportunities for commercial crews and remained relatively without stigma due to the program’s connotations of remuneration for labor. Suggesting a networks-as-pipes mechanism, interviewees complained about perceived disparities in the distribution of information regarding how to participate. This information was initially disseminated into the community by the Franklin County Seafood Workers Association (FCSWA), which accounts for the positive association between membership and participation in VoO (b = 1.32, p < .05). Founded in 1982, the FCSWA is a dues-collecting organization and maintains sufficient credibility to be partners with other local organizations, such as the Riverkeeper Alliance, Franklin’s Promise Coalition, and the county Workforce program. Through these ties, the FCSWA receives resources that it distributes only among its active, dues-paying members. To participate in activities such as reshelling projects, in which oysterworkers redistribute oyster shells into the bay for pay, one must maintain organizational membership. Contemporarily, its membership consists of approximately an eighth of the active oysterworker population, though this fluctuates seasonally.

Although FCSWA did not attempt to restrict information exclusively to its members, members nonetheless received an advantage. This is due, partially, to some oysterworkers’ antagonisms toward FCSWA, many of which predate the oil spill and ensuing environmental crisis. Information about VoO was occasionally communicated electronically through web sites and email; people sympathetic to the organization (even if they were not members) were likelier to encounter this information. Dillon, a middle-aged oysterman, expressed his frustration with FCSWA’s control of information:

The one that used to be the president of the oyster association? Her and her friends all got hired… Then they’re saying well, if the oil comes in here, you can sign up with us. And you’re gonna work for us… Then you hear a couple months later everybody goin’, ‘no, that ain’t the way it is…’ The leader of the oyster association was always gettin’ hired on to help do the groups and stuff. They learned a lot of it, and then they only benefited so many people in telling the right information. Instead of helping all the oystermen out, they only give the information to a certain few.

Turnover of the FCSWA leadership has historically been high, with accusations of nepotism frequently directed at past regimes. At the time of DWHOS, the leadership of FCSWA was not immune from similar criticisms. Excluded oysterworkers criticized the perceived unequal distribution of information benefits as unfair and as FCSWA as engaged in self-interested activity at the expense of the general oysterworker community.

This problem was exacerbated by the relative scarcity of positions in the VoO program. Unlike the claims process, the work program did not offer assistance to every individual in need. Although no cap on how many individuals could seek compensation through the claims program existed, only 3,500 commercial and charter fishing boats were selected for participation in the work program across the entire gulf. Viable candidates tended to be fewer than for the claims compensation program, and competition from nonresidents, often perceived as benefiting from nepotistic mismanagement, significantly reduced available opportunities. Of our sample, only a fifth (21.89%, n = 58) were chosen to participate.

This relative scarcity created a competitive environment in which information did not travel via the network conduits through which news and chatter about work-related activities conventionally flowed. Respondents suspected that, to improve their own probabilities of participation, their acquaintances, friends, and even family members intentionally withheld important information about work opportunities. Respondents frequently reported that they felt others became less community-oriented following the oil spill and called their motivations “self-centered” and “greedy.” Our quantitative analyses lend support to the idea that, faced with conditions of interpersonal competition for scarce resources, many oysterworkers restricted the flow of information. Oysterworkers whose personal networks largely consisted of other oysterworkers were substantially disadvantaged; for each support-network relationship that originated in a work context, a respondent is 0.92 times (exp[−0.08]) as likely to participate as somebody whose social network contains alters from other settings such as family, friends, and neighbors. Accordingly, oysterworkers who drew their associations from non-work contexts possessed an informational advantage in accessing new information about VoO.

Conclusions

Disaster recovery of individuals and communities has been an important topic of sociological research since its inception. Yet studies of how personal network contexts affect individuals’ responses to formal disaster recovery programs remain rare. Through our mixed methods approach, we find that a networks-as-prisms mechanism best explains network-related effects on formal receipt outcomes in the claims process. Qualitative accounts of the process show that information about the claims process was readily available even among people who decided not to participate, and individuals who declined participation framed their non-participation as a moral choice. Due to the claims program’s association with BP, local pockets of the community interpreted participation in the claims process in terms of “complicity” with the party responsible for the disaster. This stressed relationships and deterred participation among individuals with strong local identity commitment.

We also find that a networks-as-pipes mechanism best explains network-related effects on formal receipt outcomes for the VoO work program. Analyzing participation in VoO, we find evidence that individuals with personal networks consisting of large numbers of oysterworkers were less likely to pursue support. In the wake of the oil spill and fishery collapse, oysterworkers reported feeling greater levels of competition for resources. Because the VoO program featured a limited number of work opportunities, people with information were incentivized not to share in hopes of reducing competition. Although previous research theorized a mechanism of informational redundancy (Beggs, Haines, and Hurlbert 1996a), our qualitative data suggest that VoO’s restricted number of open positions created an environment of competition that transformed existing social relations into competitors, diminished trust, and reduced the flow of relevant information (Small 2011). Individuals whose personal networks were most deeply integrated into the oysterworker community suddenly became the most disadvantaged, and they experienced this as something of a betrayal.

Our results suggest opportunities for expanding the reach of support efforts in future environmental disasters and preventing perceptions of second-order recreancy from inflicting unintentional damage to community relationships. First, we argue that social networks provide key mechanisms for facilitating disaster recovery through distinct mechanisms. Often in the extant literature, assumptions are made that denser social networks universally improve recovery. Our findings suggest that future research can be improved by carefully specifying potential mechanisms linking networks to recovery. Identification of which network-relevant mechanisms is important because they implicate different potential solutions to encourage participation and recovery. First, compensation programs could take efforts to avoid connotations of “entitlement” by explicitly emphasizing alternative framings, such as “compensation.” Concerns about complicity with the responsible party inhibited participation; this could be avoided by transferring the responsibility for compensation provision to a neutral third party. Second, the VoO work program was characterized by disparities in opportunity resulting from some individuals’ informational advantage, which contributed to tensions within the community. While the selection of the FCSWA to diffuse relevant information was a logical choice, the organization had a polarizing presence in the oysterworker community even before the crisis. Future support efforts should ensure that information about potential participation is widely available and that criteria for inclusion are articulated clearly. Support providers should also consider expanding work opportunities, as they are less likely to suffer the negative connotations of “entitlement” programs.