Information Trust Falls: The Role of Social Networks and Information During the COVID-19 Pandemic Among Suburbanites

HOW NETWORKS INFLUENCE BEHAVIOR

Much research has underscored that networks influence our behavior: the choices and encouragement of people with whom we interact encourage us to alter our behavior in a process known as social contagion. A field experiment in Minneapolis and Denver showed that voters transmitted their motivation to vote from household member to household member (Nickerson 2008). Contact with advertisements and messages on social media platforms such as Facebook can influence network members to take on more activist behaviors (Kwon, Stefanone, and Barnett 2014). So too health-related behavior can be altered by network members talking about and encouraging activities such as quitting smoking or exercising (Latkin and Knowlton 2015).

These networks can include a pair of actors, or dyads, where information is directed from one actor, the source, to another, the target, but is not necessarily reciprocated (Yang, Keller, and Zheng 2017). Elements of these social networks, such as the information, influence, social credentials, or reinforcement provide individuals with resources that can be mobilized for action or gain (Coleman 1988; Lin 2001), which Nan Lin defines as social capital, or “the resources embedded in social networks accessed and used by actors for actions” (2001, 25). These networks can remain open and dispersed, allowing for bridges and weak ties to emerge and still facilitate information or influence, or dense and closed, giving greater leverage to realize the gains of social capital (Granovetter 1973; Lin 2001).

Research on social capital categorizes social ties into horizontal and vertical ones (Woolcock 2002); these function in different ways and lead to varied outcomes during shocks and crises. Close horizontal ties can facilitate physical, emotional, and financial support during an event (Hurlbert, Haines, and Beggs 2000). Tightly knit horizontal ties often between friends or family, for example, help would-be evacuees and affected individuals make critical, often life-saving choices about departure and return. A study of more than 1.5 million social media users before and during Hurricanes Harvey, Irma, and Maria demonstrated that broader and more diverse networks motivated people to evacuate from low-lying areas (Metaxa-Kakavouli, Maas, and Aldrich 2018). Communities with stronger horizontal social ties bring back more evacuees after shocks as well (Aldrich 2012).

Beyond horizontal ties, vertical ties to decision-makers also matter in shocks. Although more than fifty coastal towns and villages were devastated by Japan’s tsunami of March 2011, only one community—a city with a connection to well-resourced decision-makers in Tokyo—received central government funds for expensive reconstruction projects (Aldrich 2019). One-way vertical ties (individual trust in government officials), have also been shown to matter in disasters; in the 2014 Ebola outbreak in West Africa, for example, individuals expressing distrust in local officials were less likely to adopt preventive behaviors (Vinck et al. 2019). How these horizontal and vertical ties influenced health behaviors early in the pandemic remains an open question.

METHODS AND DATA

This study undertakes both quantitative and qualitative analyses to understand the associations between reliance on information from horizontal and vertical ties and behavioral change during the COVID-19 pandemic. We draw on an original dataset (N = 822) from a survey (Aldrich et al. 2020) conducted in May and June of 2020 in the Boston and New York City metro areas, including zip codes 02127, 02132, 02135, 10304, 10301, 11223, 11224. We recruited individuals by mail and social media, purchasing contact information from Valassis. Participants recruited by mail received a personalized letter of invitation with a paper survey and a prepaid return envelope. Those recruited by social media were recruited on Facebook with Facebook ads in our select zip codes and were invited to complete a survey identical to the paper survey via the Qualtrics survey platform.

Quantitative surveys captured respondents across a broad range of demographic characteristics—gender, age, income, education, and social capital—but the average respondent was a White, highly educated woman. Table 1 displays the demographic distribution of our sample. These findings are consistent with what we know of social science survey participants, especially in web-based surveys. Previous work on the racial, gender, and socioeconomic characteristics participation rates have found that non-Hispanic Whites and Asians are more likely than Black participants to participate in surveys, higher levels of education are associated with survey completion (Jang and Vorderstrasse 2019; Reinikainen et al. 2018), and women are more likely than men to participate in self-administered surveys (Mulder and de Bruijne 2019).

Social scientists typically seek to capture large, representative samples and use statistical techniques such as weighting to reproportion their sample to look like the broader population should they fail (D’Exelle 2014). Yet epidemiologists and social scientists alike acknowledge the value of nonrepresentativeness in causal investigation (Rothman, Gallacher, and Hatch 2013). We build on the work of other articles in this issue that have looked at specific demographic groups such as vulnerable older adults (Pezzia and Leonard 2022, this issue), Native Americans (Evans et al. 2022, this issue), and residents of small towns (Burns and Albrecht 2022, this issue). We apply post-stratification weights only to gender, leaving our nationally nonrepresentative sample to match 235 counties in the United States with similar attributes, home to more than seventy-four million Americans. To calculate our matching counties, we selected U.S. counties that exceed the 2021 Census Bureau national averages of median household income ( > $64,994), percentage White alone, not Hispanic or Latino ( > 60.1 percent), and percentage bachelor’s degree or higher ( > 32.9 percent). The 235 counties matching these demographics include a number of large metro areas (for a complete list and map of matching counties, see figure A.1 and table A.4 in the online appendix at https://www.rsfjournal.org/content/8/8/32/tab-supplemental). We confine our conclusions about behaviors to individuals with higher levels of human capital and to residents in urban and suburban communities whose votes are coveted and whose behaviors are scrutinized during elections (Kurtzleben 2018). Finally, we used a nationally representative dataset of 5,275 respondents from the 2022 U.S. Census Pulse surveys to double-check our results.

We carried out follow-up semi-structured interviews with a subset of survey participants in June and July 2020 (n = 62) to add depth and causal process to our quantitative results. Interview participants were recruited via positive responses to a survey question about their willingness to participate in a follow-up interview. All interviews were conducted via Zoom by a team of two researchers, using an instrument designed to further probe responses to each segment of the survey instrument. Interviews were transcribed by Amazon Web Services’ Transcribe service for analysis. Table 2 displays the demographic distribution of our interviewees.

We analyzed the data from our qualitative interviews with a reflexive thematic analysis (Richards 2021), which includes the following five steps: collect the data, code the data, categorize the data, create themes, and interpret themes. We trained a team of three graduate students to code text segments in interview transcriptions that demonstrated both reliance on information from social ties and the subsequent behavioral change. With each segment, the graduate students coded the presence of reliance on information from social ties and the subsequent behavior change. We then categorized the data into the nature of the source of information. We then created themes generated deductively from the social capital literature, which signify horizontal and vertical ties.

Our study includes three dependent variables: avoidance of public spaces to minimize viral transmission, use of personal protective equipment, and increased frequency of handwashing (see table 3). These three behaviors were routinely promoted as effective and protective behaviors to curb the spread of COVID-19 throughout the pandemic. Individuals were asked, “Please indicate any changes in your behavior due to COVID-19 or related stay in place orders.” Responses to “Avoided going to public places,” “Started using personal protective equipment (such as masks and gloves),” and “Increased frequency of handwashing” were coded 1 if the individual reported the behavior and 0 if the individual did not (see table 3).

Emergent themes we observed from the interviews, such as that with Gabrielle, demonstrate respondents came to trust authorities even if their only connections came from one-way, media interactions: “Such as Dr Anthony Fauci, I learned the science behind the pandemic. I learned, um, I watched as the science evolved and we became more aware of what it was and what it meant.”

Meghan similarly talked about building trust through broadcasts from an elected figure: “And I would add that the other place I did go for information was Governor Cuomo. So I would listen to his daily conferences. . . . Well this is gonna sound really hokey but I felt like he was being honest because he was saying how bad it was.”

Further, when asked about vertical ties to local elected officials, informants responded as follows:

Yeah, so I probably once a week watch Mayor Walsh’s or Charlie Baker’s press conferences to the Boston.com link to get direct information from them. . . . Um, yeah, I think, and also just trying to make good decisions from my family about you know, what’s safe, what kind but parts of the reopening do I want to participate in what don’t I want to do? And so on. (Beth)

Yeah. So it was daily briefings with Cuomo. It was like the highlight of my day. I just felt they were very realistic and factual based. So I appreciated that without all the speculation and rhetoric from other sources. . . . His daily updates were informative. They were fact based and they were science based, so I appreciated that. The CDC and WHO I Googled and went on the CDC website a lot, especially concerning the baby and using masks and, you know, just general information like that. And then, um I know me and my friends were talking once a day and they said something that says what’s good. . . . Well, we stay in the house a lot. Um, wearing a mask—I’ve become much more conscious of washing my hands than I had been previously. I washed them, but not like I do now. Um, I carry hand sanitizer. Now, any time I go anywhere, I come from there, I put it on my hands before I get in my car. (Gigi)

These vertical ties match the definition of linking social capital, which involves “respectful and trusting ties to representatives of formal institutions” (Szreter and Woolcock 2004, 665). These connections capture aspects of political participation, including political efficacy and political trust (Poortinga 2012, 288).

Horizontal ties capture the high trust, frequently contacted connections involving friends, extended kin, and family (Adler and Kwon 2002). This form of social capital often results in homophily—high levels of demographic, attitudinal, racial, and linguistic similarity—in these networks (Mouw 2006). These thick connections inform life-saving decisions when neighbors knock on doors before a fire spreads out of control or rescue family and friends nearby who are trapped beneath the rubble of collapsed homes (Aldrich 2012). In our interviews, when asked to explain the information they relied on to get through the coronavirus outbreak, individuals routinely cited the importance of these ties: “Originally in the beginning, I was looking at all the news and it just got very overwhelming. And there were a lot of sources that didn’t feel very reliable. So eventually I just started asking my closest friends who were really following experts to sort of guide me on like, Okay, what are the new policy things? What are the changes? How much worse has it gotten? Because I just couldn’t take it. And I knew that my friends were really looking at reliable news sources. (Christina)”

Horizontal ties also include ties that connect people from different backgrounds through institutions, including workplaces, faith-based organizations, and horizontal associations. Early research on “thin” or “weak” ties indicated their tremendous power in helping young job searchers find work through their extended networks (Granovetter 1973). Even a small, neighborhood-based childcare center can provide users and their families with critical information and assistance (Small 2009). For example, in our interviews, individuals cited the importance of these ties to practicing their faith in the pandemic:

One thing I would say is that our rabbi was very strict about everything, and it made us feel very supported as, like the community that, like he took [COVID] seriously and also meant that like when he decided to start doing services again. So he made it very safe so that I felt comfortable that my husband could go back to a synagogue and pray there like, for example, like he had it outdoors. And everyone has to stand six feet apart. And everyone has to wear a mask and, you know, like otherwise like, I feel like he couldn’t have gone back in the community if he hadn’t taken it seriously. (Candice)

To understand the associations between reliance on information from horizontal and vertical ties and the practice of prescribed safe and healthy behavior during the pandemic, we used an iterated principal factor analysis to categorize the sources of information individuals relied upon based on the themes from interviews we conducted and on survey responses to the following question: “Which of the following sources have you relied on for information during the COVID-19 outbreak?” (for the full qualitative script and quantitative survey questions, see the online appendix).

The variables in our factor analysis loaded on two key factors: reliance on information from horizontal and reliance on information from vertical ties (see table 4). Information received and relied on from friends, neighbors, family, coworkers or colleagues, and community organizations and groups (such as church, temple, mosque, neighborhood groups) were categorized as reliance on information from horizontal ties.

Information from local elected officials, local government agencies, state elected officials, state agencies, federal government leaders, CDC, NIH, and the World Health Organization (WHO) were categorized as vertical ties, consisting of trust in or relationships between people separated by levels of power and authority (Szreter and Woolcock 2004). These ties provide information and resources that are often unavailable locally. In our interviews, individuals routinely cited reliance on the CDC, the NIH, Governors Baker and Cuomo, and Fauci:

Well, they’re trusted institutions, they’re institutions that I associate strongly with making decisions and issuing information based on science, not based on politics or the political moment that they would actually assess the risks and tell us. . . . There [was] a lot of uncertainty early on . . . the role of masks and how important masks were. So I relied heavily on those institutions to evaluate the science. . . . I trusted the CDC and NIH to actually evaluate [all of the claims made by researchers] publishing in not yet peer-reviewed pieces and tell us what we actually should and should not take away from them. (Christopher)

The CDC and the NIH I think provided many guidelines for what you should be doing? Whether it’s wearing mask or social distancing or, you know, touching surfaces or how to clean things on. I relied on them personally, but also for my job [with] Zipcar. . . . relied on a lot of both local and federal and state regulations to figure out how we could safely . . . support our employees and our members. (Emmie)

Work on vaccine hesitancy (Khubchandani et al. 2021) and responses to health recommendations during COVID-19 (Grossman et al. 2020) identifies factors correlated with accepting and acting on new information. Building on this literature, we controlled for demographic characteristics (see table 2), including age (O’Malley, Kerner, and Johnson 1999), income (O’Malley, Kerner, and Johnson 1999), education (Khubchandani et al. 2021), gender (Burke 2001; Carpenter et al. 2011; O’Malley, Kerner, and Johnson 1999), and political party affiliation (Malka, Krosnick, and Langer 2009; Grossman et al. 2020; Khubchandani et al. 2021; Barrios and Hochberg 2020; Gadarian, Goodman, and Pepinsky 2021; Allcott et al. 2020; Clinton et al. 2021).

ANALYTICAL STRATEGY

For our quantitative analysis, we use logistic regression models to test the reliance individuals had on vertical and horizontal ties and our three dichotomous dependent variables: avoidance of public spaces, use of personal protective equipment, and increased frequency of handwashing. We structure additional regression models, including ordinary least squares, Poisson, and negative binomial regression models to ensure the results we uncovered in our logistic regression were not artifacts of model choice. All three of the models have heteroskedasticity–robust standard errors, and multicollinearity tests indicated low levels of collinearity between the variables in our models. The variance inflation factor for the models was less than 3.79, which is well below the accepted threshold (Salmerón Gómez et al. 2016). Further, to correct for oversampling of women in our study, we applied a post-stratification weight to our models based on the gender distribution reported by the Census Bureau.

RESULTS

The results of our statistical analyses are organized by three behaviors designed to curb the spread of the virus promoted by the CDC: avoiding public places, using personal protective equipment, and practicing frequent handwashing.

Staying Home and Avoiding Public Places

In our first model (table A.1), our results indicate that individuals in our study who relied on information from their horizontal and vertical ties were more likely to indicate avoiding public places in May and June of 2020. A one-unit increase in reliance on information from horizontal ties results in a 1.407 (p < .05) increase in the odds of an individual avoiding public places. A one-unit increase in reliance on information from vertical ties results in a 1.736 (p < .01) increase in the odds of an individual avoiding public places. Further, holding all other variables at their mean, post-estimation predictive margins reveal that (figure 1a, figure 1b) the probability of avoiding going to public places goes from 82 percent to 93 percent from the lowest to highest measures of relying upon horizontal ties, and from 67 percent to 91 percent from the lowest to highest measures of relying on vertical ties. Gender, race, and employment status were also significant variables in this model. Women were more likely than men in the reference group to report avoiding public places (1.671, p < .1). Non-White individuals were less likely than White individuals in the reference group to report avoiding public places (–0.565, p < .1). Further, individuals unable to work were more likely than individuals employed full time in the reference group to indicate avoiding public places (9.443, p < .01).

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Margin Plots

Source: Authors’ tabulation, analysis, and visualization in STATA from original data.

Note: Adjusted predictions with 95 percent Cls.

DISCUSSION

Much has been made of the role of political party identification in determining behavior during the COVID-19 pandemic (Grossman et al. 2020; Khubchandani et al. 2021; Barrios and Hochberg 2020; Gadarian, Goodman, and Pepinsky 2021; Allcott et al. 2020; Clinton et al. 2021). For example, newspaper articles and blogs alike have emphasized that Republicans seem less likely to take on mask-wearing or engage in physical distancing and the other preventive measures endorsed by health agencies. However, in our models consisting of primarily Independents and Democrats to accurately characterize the partisanship of the geographies we sampled, we did not find party affiliation to be a significant predictor. Instead, our research pushes us to look beyond simple party affiliation to the role of social networks—horizontal and vertical ties—that influenced healthy and safe behavior during the COVID-19 pandemic.

Our findings indicate vertical ties with entities of power, or the reliance individuals had on mayors, governors, and government leaders such as Dr. Fauci, was a statistically significant variable in all three of our models. These findings confirm that the information from these sources of power and prestige was associated with a higher probability that survey participants would report staying home during the pandemic, using PPE, and increasing the frequency of handwashing. In the case of PPE use, vertical ties were associated with the near-universal use of facemasks or gloves during the early months of the pandemic. In terms of staying home, horizontal ties were also important.

Our qualitative findings (discussed in the following section) from sixty-two one-on-one interviews suggest that reliance on horizontal ties, in particular, played a role, for example, in getting to and from work in a way that avoided the public. In our follow-up interviews of survey participants, individuals further confirm these findings. Direct quotes are reported in the participants’ own language; respondent names have been changed to pseudonyms to protect their identity.

THEORETICAL AND POLICY IMPLICATIONS

Our findings resonate strongly with those of other investigations that have used nuanced categorization of social capital to illuminate the role of horizontal and vertical social capital (Fraser and Aldrich 2021; Hawkins and Maurer 2010; Kyne and Aldrich 2020; Page-Tan 2020, 2021; Poortinga 2012). Research envisioning social capital in a single dimension fails to advance our understanding of how vertical, horizontal, inclusive, and exclusive ties can operate differently from one another. Furthermore, our inclusion of one-way ties expands on previous definitions of vertical social capital, facilitating valuable overlap with previous work on the role of trust in outbreak response (Vinck et al. 2019).

We find that information from trusted entities with political and administrative power shared a strong association with recommended behavior to protect individuals from contracting and spreading the novel coronavirus. Further, staying home or avoiding public places was correlated with information—and often assistance—from close, tightly knit horizontal ties. For example, one interviewee drove her roommate to and from a construction job every day to protect him from commuting via public transit. By measuring and testing these various ties, we can better understand and leverage these social ties in crises and emergencies. In sum, this research speaks to the importance of avoiding monolithic measurements of networks and trust, which would miss the differential effects of the different horizontal and vertical ties.

Further, our findings hold implications for previous work on outbreaks showing negative effects of misinformation on institutional trust and adopting preventive behaviors (Daszak et al. 2021; Vinck et al. 2019). Although we find information from trusted entities of power (vertical social capital) strongly associated with the adoption of preventive behaviors, information from federally elected officials (namely, former President Trump) known to be frequent sources of misinformation was not a significant factor in our analysis (see Kessler 2021). Some interview participants discuss the loss of trust in federally elected officials and federal institutions due to misinformation and mixed messages, resonating with findings from research during previous outbreaks (Vinck et al. 2019). Yet precedent is limited for large-scale health communication campaigns during outbreaks in the United States today (Fisher et al. 2020). Our findings complement initial studies focused on vulnerable communities (Pezzia, Rogg, and Leonard 2022, this issue; Evans et al. 2022, this issue; Burns and Albrecht 2022, this issue), reinforcing the importance of horizontal and vertical ties among a large, constituent group constituting one-fifth of the country. This underscores the need for further research to understand the dynamics of health communication, misinformation, and institutional trust in the U.S. context during the COVID-19 pandemic.

These findings have implications for policymakers both preparing for and responding to future outbreaks and other prolonged crises. Our research suggests that policies focused on disseminating information through community-level influencers and leaders—whether in faith-based organizations, civil society organizations, or schools—may have beneficial effects because these individuals’ beliefs and behavior can serve as a bellwether for others in their networks. Our findings furthermore reinforce the importance of horizontal ties in facilitating decisions regarding movement and travel during crises. When it comes to decisions about whether and how to avoid public spaces during an outbreak, the presence of close horizontal ties may mean the difference between getting a ride or taking public transit and between wearing a mask or going out unprotected. Therefore, policies facilitating the development of close ties within community and neighborhood networks—such as the development of community centers and funding for neighborhood-based civic groups—may lay the groundwork for the uptake of safe behaviors during future outbreaks by providing individuals with necessary social support structures.

CONCLUSION

The COVID-19 pandemic called for public health communication efforts of a scope and scale unprecedented in the United States today (Fisher et al. 2020), in that Americans were inundated with information about the importance of nonpharmaceutical interventions—such as staying home and avoiding public places, wearing PPE, and handwashing—to curb the spread of disease. Our research demonstrates the importance of social networks—information embedded in horizontal and vertical social capital—to individual adoption of safe behaviors. Associations between vertical social capital and adoption of all three preventive behaviors we study, alongside an association between close horizontal ties and avoiding public places during the early months of the pandemic, further underscore the importance of using nuanced categorizations of social capital, as opposed to treating social capital as a monolith (see also Fraser and Aldrich 2021; Hawkins and Maurer 2010; Kyne and Aldrich 2020; Page-Tan 2020, 2021; Poortinga 2012).

These findings represent an early step in unpacking the complex dynamics of social and civic networks as information sources for health communication in the United States during COVID-19. Interpretation of our findings is limited by the nature of our sample, which is limited to two northeastern U.S. metro areas, namely, Boston and New York. Survey participants were disproportionately White, upper-middle-class, and politically liberal, mirroring a suburban population subset of seventy-four million people to add critical context to studies of health communication typically (and rightfully) focused on vulnerable groups. In highlighting weaknesses in the U.S. health-care system—including inexperience with mass health communication efforts—the pandemic underscores the need to understand how such efforts are received across broad swaths of the population, including privileged groups historically facing fewer obstacles in access to health care. Future studies should examine the extent to which these findings hold among samples more demographically representative of the U.S. population. Further, panel data are needed to capture variation and trends over the course of a prolonged crisis. To address this, our research group conducted follow-up studies in February 2021 and February 2022.

Developing a more nuanced understanding of the role civic networks play in information-sharing during prolonged crises has important implications for policymakers and communities. This applies to both preventive initiatives—such as incentivizing the development of horizontal ties within neighborhoods and communities—as well as crisis response efforts—including targeting health communication to reach key stakeholders within community networks. Further, these findings contribute to integrating social science approaches with public health research to better inform policy in future crises (see Subbaraman 2021). Given that our societies are certain to face shocks in the future, sophisticated, social capital, and social-infrastructure-based responses will better help us build resilience to these shocks.