By: Marcella Alsan and Sarah Eichmeyer. Marcella is a Professor of Public Policy at Harvard Kennedy School, marcella_alsan@hks.harvard.edu. Sarah is an Assistant Professor of Economics at the University of Munich, sarah.eichmeyer@econ.lmu.de

In order to curb, and ultimately overcome the coronavirus pandemic, the development, production, and distribution of vaccines is key. Now that the former two steps are well underway, distribution of the vaccine to as many people as possible has taken center stage. Equitable distribution within the United States requires lowering barriers to vaccination as much as possible, including addressing system level issues such as reducing time and travel costs to vaccination sites and maintaining a robust reliable supply (Chevalier et al., 2021). It also involves addressing individual-level “behavioral” factors that reduce take-up, including inertia, inattention and misperceptions of the net benefits. Recent studies have demonstrated the importance of cues and reminders for helping those who are willing to be vaccinated execute that intention (Milkman et al., 2021; Dai et al., 2021). However, for those who are not convinced that the technology is of net benefit to them or their communities, what messaging strategies might persuade them to change their mind?

These issues are of crucial importance for communities that have been historically disadvantaged in the United States – where trust in the healthcare system has been frayed due to historical and contemporaneous racism, mistreatment and neglect, and where rates of infectious disease are generally higher. This has been exemplified with COVID-19 where mortality rates are 1.9 and 2.3 times higher among Black and Latinx communities, respectively (CDC, 2021).  Although misperceptions of net benefits often stem from inaccurate information on the safety and effectiveness of the technology, a growing theoretical and empirical literature on persuasion finds the provision of accurate information is often insufficient to change peoples’ views; instead, other features, such as message framing, as well as the receivers’ beliefs over the sender’s preference, may matter greatly for how persuasive the information provided is (Coffman and Niehaus, 2021; Haghtalab, 2021; Bertrand et al., 2010).

In a recent study, we evaluated the effectiveness of messaging interventions designed to shift knowledge, beliefs and take-up behavior regarding vaccines (Alsan and Eichmeyer, 2021). Our sample consisted of White and Black men ages 25-51 without a college education who had not received their seasonal influenza vaccine at the time of recruitment, which occurred during the last two flu seasons. We developed and randomly distributed standardized video messages on the safety and effectiveness of the influenza vaccine, narrated by ten separate senders. The videos varied along three policy-relevant dimensions: (1) the perceived medical expertise of the sender (“expertise”), (2) the race of the sender (“concordance”), and (3) the admission/omission of acknowledgement of past injustice committed by the medical community by discordant senders (“acknowledgement”). Variations (1) and (3) were distributed to Black respondents only, because uunderstanding the potential of concordant community members to substitute for medical experts, as well as the role acknowledgements of past injustice by discordant physicians may play in bridging trust gaps, holds particular relevance amidst challenges in rapidly diversifying the physician workforce and persistent racial health inequalities.

Respondents completed an online survey (“baseline”) in which they watched the randomly assigned video message, followed by a survey module that elicited the following outcomes: evaluations of the sender and the signal, content recall, beliefs about the safety of the flu vaccine, incentive-compatible measures of interest in a coupon for a free flu shot, as well as self-reported intent (“willingness”) to receive the vaccine. At least two weeks after baseline, participants were invited to complete a follow-up survey, in which we elicited self-reported flu vaccination status. In the most recent flu season, we also assessed whether our message on flu shots had spillover effects on intent (“willingness”) to receive a COVID-19 vaccine.

We establish three main results. First, when comparing layperson to expert senders, we find that lay senders are rated by respondents as substantially less qualified (0.54 standard deviation units) to give medical advice. However, individuals in the non-expert condition exhibit greater recall of factual message content and increase their willingness to receive the COVID-19 vaccine by 8.8 percentage points (20%). Furthermore, respondents assigned to lay senders were 15 percentage points (39%) more likely to report that a household member had received the flu vaccine in the weeks between the baseline and follow-up surveys.

Second, we find that Black respondents rate the sender and the signal significantly more positively when assigned to a concordant expert sender (relative to a discordant expert sender), while we detect no concordance effects at all among White respondents. We further find that acknowledgement of past breaches of trust by a race-discordant expert sender increases ratings of the signal by approximately the same magnitude as a race concordant expert sender providing the standard message without acknowledgement. Neither intervention, however, significantly affects self-reported influenza vaccine take-up as measured in the follow-up survey, although treatment effect estimates on self-reported intent to vaccinate against influenza and against COVID-19 elicited at baseline are weakly positive in both arms.

Third, we find striking heterogeneity by treatment arm across respondents with different levels of prior experience with flu vaccination. Viewing previous flu vaccination experience as a proxy for distance from a take-up “threshold,” we divided the sample into never-takers, ever-takers, and recent-takers based on the date of a respondent’s last influenza vaccine. We find that both the concordance and acknowledgement interventions demonstrated significant effects on flu and COVID-19 vaccination intent among recent takers, those who had received seasonal flu vaccines within the past two years (about a quarter of the sample). In sharp contrast, the effectiveness of non-experts was strongest among those who had never previously received a flu vaccine (another quarter of the sample), with individuals in this group rating the non-expert message significantly higher than respondents who had previously taken up the flu vaccine, and exhibiting substantial increases in flu and COVID-19 vaccination intent (by 47% and 49%, respectively).

Taken together, these findings represent a step towards identifying effective ways to influence immunization views and behaviors. While messages from concordant and empathetic experts may resonate most among individuals familiar with vaccination, our study suggests that peer figures, such as community health workers or citizen ambassadors, could play an important role in communicating benefits and dispelling myths about vaccines among those least inclined to receive one. In the context of the COVID-19 pandemic, vaccination information campaigns involving community health workers, such as Massachusetts’ COVID-19 Vaccine Equity Initiative or the Department of Health and Human Services “We Can Do This” Campaign, may be important complements to campaigns involving physicians (Alsan et al. 2020).