How to reduce vaccination hesitancy? The relevance of evidence and its communicator

Abstract

Even though the immediate urgency of the COVID-19 pandemic seems to have passed, many countries did not reach the vaccination rates they initially aimed for. The stagnation in vaccine uptake during the height of the pandemic presented policy makers with a challenge that remains unresolved and is paramount for future pandemics and other crises: How to convince the (often not insubstantial) unvaccinated proportion of the population of the benefits of a vaccination? Designing more successful communication strategies, both in retrospect and looking ahead, requires a differentiated understanding of the concerns of those that remain unvaccinated. Guided by the elaboration likelihood model, this paper has two objectives: First, it explores by means of a latent class analysis how unvaccinated individuals might be characterized by their attitudes towards COVID-19 vaccination. Second, we investigate to what extent (i) varying types of evidence (none/anecdotal/statistical) can be employed by (ii) different types of communicators (scientists/politicians) to improve vaccination intentions across these subgroups. To address these questions, we conducted an original online survey experiment among 2145 unvaccinated respondents from Germany where a substantial population share remains unvaccinated. The results suggest three different subgroups, which differ regarding their openness towards a COVID-19 vaccination: Vaccination opponents (N = 1184), sceptics (N = 572) and those in principle receptive (N = 389) to be vaccinated. On average, neither the provision of statistical nor anecdotal evidence increased the persuasiveness of information regarding the efficacy of a COVID-19 vaccine. However, scientists were, on average, more persuasive than politicians (relatively increase vaccination intentions by 0.184 standard deviations). With respect to heterogeneous treatment effects among the three subgroups, vaccination opponents seem largely unreachable, while sceptics value information by scientists, particularly if supported by anecdotal evidence (relatively increases intentions by 0.45 standard deviations). Receptives seem much more responsive to statistical evidence from politicians (relatively increases intentions by 0.38 standard deviations).

Keywords: COVID-19; Elaboration likelihood model; Evidence provision; Latent class analysis; Persuasive messaging; Vaccination hesitancy.

Fig. 1

LCA model: Identification of classes of vaccination hesitancy.

Fig. 2

Classes of vaccination hesitancy: Identification via the 5C scale. Notes: Results refer to mean values of the 5C scale for vaccination attitudes, separately for each class as identified by the LCA. The mean values shown here are those of the initial 5C scale, measured on a 7-point Likert scale. See Tables A3-A5 in Appendix B for the condensed scale. Values of the 15 5C items (three items for each aspect) were averaged for each of the five aspects of vaccination hesitancy that we aim to capture. See Figure A1 in Appendix B for the same graphic with all 15 items. In order to better illustrate class differences regarding the 5C vaccination hesitancy scale, the graph employs a definite class assignment where respondents were assigned to the class with the maximum predicted probability.

Fig. 3

Heterogeneity by Class: Separate Treatment Effects of Evidence Type and Communicator on Vaccination Intentions. Notes: The figure shows the mean estimation coefficient and 95 % confidence intervals of 1,000 simulations for each of the three classes of vaccination hesitancy. Class assignment for each respondent is based on the class membership probability, which is derived from the LCA. The left column shows the treatment effects of the communicator, with the reference category ”Politician”. The right column shows treatment effects for evidence type with the reference category ”no evidence”. Estimations include controls for age, gender, education level, state of residency, and income level. Detailed estimation results are available in Table A11, A13, and A14 in Appendix B.

Fig. 4

Heterogeneity by Class: Interacted Treatment Effects of Communicator-Evidence Combinations on Vaccination Intentions. Notes: The figure shows mean treatment effects and 95 % confidence intervals of 1,000 simulations for each of the three classes of vaccination hesitancy. The reference category is “No Evidence Politicians”. Class assignment for each respondent is based on the class membership probability, which is derived from the LCA. Estimations include controls for age, gender, education level, state of residency, and income level. Detailed estimation results are available in Table A11, A13, and A14 in Appendix B.