## Introduction

COVID-19-related uncertainty and policy reactions caused major disruptions in almost every country, with the US and several Western European nations experiencing high levels of infections and societal fallouts1,2. With global eradication seen as unlikely, convalescent, and later vaccine induced immunity gained prominence as the path out of the crisis. In December 2020, the vaccine made by Pfizer and BioNTech became the first fully-tested immunization approved for emergency use3 and developers of several vaccines announced excellent results. Insights are being generated over time through careful longitudinal studies on safety, immunogenicity, and protection rate4. While the duration of vaccine-acquired immunity is yet uncertain, neutralizing antibodies after infection and induced immunological memory reactions to SARS-CoV-2 have been shown to persist for several months at least5,6. As the number of people with convalescent-induced (infection induced) and vaccine-induced immunity continues to increase, the utility of strict infection control measures decreases. The economic and social costs of such measures were high7.

Immunity certificates for convalescents were discussed by scientists early during the pandemic8,9,10,11 and they received policy attention in several countries. Such certificates were mostly intended as a transitory tool to reduce the total costs of the pandemic12; to allow individuals to resume their usual activities and interacting with others who were isolated or whose relationships were interrupted13; and to facilitate travel, taking into account that immunity passports were already used in public health and travel medicine14. While immunity certificates for convalescents were not introduced prior to the advent of vaccines, the introduction of Israel’s “Green Pass”—intended for those vaccinated or those who have already contracted the disease—prompted a surge of calls from policy makers for implementation in early 2021. The President of the European Commission, Ursula von der Leyen, presented a legislative proposal for a type of European “Green Pass” that would include information on whether there is proof that a person has been vaccinated, information on COVID19 recovery, or whether a person was tested for infection.

Acceptance of COVID-19 vaccines may differ among individuals. Existing evidence points to the usefulness of focusing on prosocial concerns when motivating vaccination uptake15. A global survey assessing potential acceptance of a COVID-19 vaccine indicates country-level differences in acceptance rates ranging from less than 55% in Russia to 90% in China16. Another study using European data reports that people’s willingness to vaccinate ranges from 30% (Hungary) to 80% (Denmark)17. Several individual-level differences are associated with the willingness to receive a vaccine18. Organized campaigns by vaccine-hesitant groups promote beliefs that vaccinations are unsafe via social media, leading to vaccine hesitancy; a large proportion of the content shared about vaccines on popular social media sites are anti-vaccination messages19. Thus, vaccine refusal is seen as a significant problem17. Individual acceptance of vaccination may also be negatively affected by increasing immunity levels in the population, as the incentive to free ride on others’ vaccinations grows20,21. This raises the question of whether immunity certificates would offer further motivation and positive incentive to vaccinate if they allow more individual freedom of movement and restoration of liberties, thereby empowering and motivating individuals to contribute to the common good8. Enforced measures have been shown to crowd out voluntary support for COVID-19 policies22.

These considerations make the question of immunity certificates an ongoing and important policy topic in addressing the impacts of the COVID-19 crisis. An even more critical aspect may be whether immunity certificates could serve more generally as a policy instrument to reduce the costs of future pandemics, even when vaccine-induced immunity does not become available as quickly as it has during this pandemic. Seeking more clarity on the role of immunity certificates as a tool to mitigate some of the health, social, and economic impacts of pandemics, we conducted a survey of 12,738 scientists between May 4 and June 3, 2020. Early uncertainty about the duration and strength of convalescent immunity and the doubts regarding the potential of vaccine development make our survey interesting for understanding the scope of immunity certification in the initial stages of such a crisis. While future crises may be of a different structure, the knowledge that some immunity may be conferred by infection could make immunity certificates a helpful policy tool in specific circumstances.

## Background and context

First and early evidence regarding the direct health effects of COVID-19 indicated that the high-risk group was largely comprised of elderly people and people with pre-existing medical conditions23,24. In March 2020, the World Health Organization noted that most patients (80%) experienced mild illness, based on data from China25. The number of asymptomatic cases was unknown but stated to be relevant for evaluation of policies and the ultimate severity of the pandemic26.

Although evidence cannot exist for a new type of virus, the probability of contracting the same illness from the virus a second time within a few months or even years was considered to be small compared to a first occurrence of the illness. Numerous experts such as Peter Doherty—recipient of the Nobel Prize—suggested in media reports that even if there was a reinfection, prior infection would give an individual a level of immunity, allowing them to recover quickly27. To some degree this was the—now seen as largely correct—prior belief.

Pandemics and the reactions to them increase the general problem of scarcity that always exists in society. The possible immunity of convalescents could make them each individually and as a group a valuable resource, and such resources multiply as the number of convalescents increases12. Moreover, from a point of view of freedom, it is not easy to justify wide-ranging restrictions on this group of immune people. To make individual immunity useful during the crisis and for the individual, a certification that a person has contracted and recovered from COVID-19 would be vital. However, certification policies are connected with questions about individual freedom, public health, economic benefits, fairness, and inequality; all issues that were controversially discussed by the scientific community9,10. In a statement on April 24, 2020 the World Health Orgazination suggested that “there is not enough evidence about the effectiveness of antibody-mediated immunity to guarantee the accuracy of an ‘immunity passport’ or ‘risk-free certificate’” and the organization stated that such certificates may increase the risks of transmission28. Immunity certificates might create incentives for self-infection if they are associated with large benefits. Evidence on robust immune responses after infection accumulated over time5,29,30, but with the advent of vaccines, widespread immunity came in reach at low risks. With vaccine availability in early 2021, Israel became the first country to introduce a “Green Pass” for vaccinated and previously infected individuals.

To evaluate opinions regarding immunity certificates as a policy tool, we designed a survey in late March and April 2020. The aim of the survey was to gather information on the acceptance of immunity certificates among scientists during a phase of uncertainty about the disease and uncertainty about the precise immune reactions, such that our conclusions could be useful when related discussions emerge during future crises. Data were collected between May 4 and June 3, 2020; i.e., still during the first wave in most countries and where attention was heavily focused on COVID-19. Responses were gathered via the SurveyMonkey platform from scholars appearing in Scopus (see Method section). We also included scholars from the bibliographic database RePEc to increase the representation of social scientists (excluding these entries from the analysis does not influence our main conclusions, see Appendix SI). The design allowed us to collect the opinions of scientists across 37 subfields and 63 countries (see Supplementary Fig. 1). We gathered a sample of 213,648 email addresses from journal publications. The response rate was 13.9% based on emails opened (see Supplementary Table 16) and 12,738 scientists eventually concluded the survey. We aimed at a careful understanding and mapping of scholarly positions to establish the consensus of scientists’ opinions on different aspects of immunity certification. Survey respondents could skip any questions they did not want to answer. Ethical approval for the survey and the data collection was given on April 23, 2020 by the Ethics Commission of the Frankfurt School of Finance & Management (Frankfurt, Germany).

Humans are boundedly rational beings31 and subject to emotions32 such as fear. They react to the complexity of the environment33, which affects their level of trust in the government34. Limited information on contextual factors or dynamic changes may not allow for clear ideas about risks and facts in uncommon situations such as pandemics. The survey was addressed to scientists, and we explicitly address respondents in their function as a scientist. Being highly educated, being trained in rational thinking and dealing with aspects of uncertainty, scientists have the potential to represent an interesting group when exploring opinions regarding immunity certificates. Uncertainty is a core problem in pandemics, and historically pandemics have often caught governments and authorities unprepared and flat-footed, leading to confusion and improvisation35. Scientists are also aware of how surveys work and deal daily with hypothetical situations. They are trained to think about social responsibility and therefore beyond personal interest, with a natural concern for understanding how people make choices. While they are not immune to bias and do not possess all information, they are usually better aware of the state of affairs than the wider public. Moreover, during crises they often inform policymakers and the public regarding scenarios, policy options, or potential trade-offs. In particular, governments can be heavily dependent on experts and scientific advice during crises due to their need for rapid responses36. During pandemics such as COVID-19 that represent a massive global health crisis, scientists can help policy-makers, leaders, decision-makers and the public in general to better understand how to handle and manage potential threats, how to find solutions, how to navigate different contexts (e.g., social or cultural), or how to align individual or collective interests while also providing effective science communication37. Their attitudes towards policy tools can therefore shape how society copes with such crises. Policy-makers have a variety of tools available to deal with crises. It is therefore valuable to understand the acceptance of—and consensus on—available tools among scientists. Such an inquiry contributes to cognitive processes of learning to cope with new situations, feeding into the discussion process in which insights, attitudes, and preferences are exchanged. It provides a way of understanding not just existing knowledge but also values and priorities.

We informed the respondents in the survey by briefly introducing them to the idea of immunity certificates, stating: Suppose that there is enough evidence that suggests that people who were infected with COVID-19 and then recovered will likely be immune to re-infection for a certain period of time and are less likely to transmit the disease to others during this period. Authorities around the world may then consider policies for identifying people with immunity by using mass testing, assigning them “immunity certificates” (also called “immunity passports”) that officially declare them immune to COVID-19, and then employing them in critical positions on a voluntary basis (e.g., caring for the elderly). Much feedback on the questions suggested that the respondents were very well aware of the ongoing discussions on immunity certificates at the time of the survey.

## Results

### Overall attitudes towards immunity certificates

We asked scientists if they agree or disagree that immunity certificates are (1) good for public health; (2) good for the economy; (3) fair to others who do not have immunity; and (4) whether certification increases inequality (7-point scale from 1 (strongly disagree) to 7 (strongly agree)). The survey results are provided in Fig. 1A. About half of scientists agree that issuance of immunity certificates for the duration of immunity is good for public health (50.2%) and the economy (54.4%), while one-fifth (19.1%) and one-sixth (15.4%) disagree, respectively. In terms of fairness, about 36.5% of scholars think that issuing immunity certificates will not be fair to those who do not have immunity. 45.5% of the respondents think that immunity certificates will increase inequality in society.

### Sample description

Descriptive statistics of the sample are presented in Supplementary Table 17. We recorded a total of 12,738 responses from scientists across different disciplines. Females represent around 42% (n = 5335) of the sample, while 57% (n = 7218) of the participants are male. Most participants were in the age brackets 30 to 39 years old (32%, n = 4131) and 40 to 49 years old (29%, n = 3637). In addition, we recorded information of unique relevance to this demographic group, such as their field of study, how many of them have completed a PhD and whether they hold a professorship (28.5%, n = 3631). Most of the respondents are from Europe (42.3% n = 5408) and North America (37.22% n = 4759). The majority of participants held an assistant professorship (equivalent or below) (52.8% n = 6664).

Compared to economics scholars recruited from the top journals in the fields (n = 1440), the pool of survey participants drawn from the RePEc register (n = 1019) is composed of more males (5.3 percentage points, p = 0.004) and are from older age groups (p = 0.044 based on a two-tailed rank sum test). However, we did not find any significant discernible difference in other sample characteristics (e.g., professorship, political views, religiosity, marital status) between the two samples (all p > 0.1). Nevertheless, we control for this by including a dummy variable for participants from the RePEc sample in the regression analyses. While the sample from RePEc was more supportive of the immunity certificates in terms of inequality concerns (Supplementary Table 12) and have higher willingness to pay for the immunity certificates (Supplementary Table 13) than other economists and social scientists, removing the RePEc sample from the analysis does not change our qualitative and quantitative findings reported in the main text. The main results excluding the RePEc sample are reproduced in the Appendix SI (Supplementary Fig. 18 to 21 and Supplementary Table 18).

#### COVID-19 data

To control for contextual factors due to development of the COVID-19 pandemic, we collected the daily confirmed case and case fatality rate (CFR) statistics at the country level, as well as a measure designed to capture the stringency level of government policy responses. COVID-19 statistics were obtained from the Center for Systems Science and Engineering (CSSE) at Johns Hopkins University48 and the Stringency Index were obtained from the Oxford COVID-19 Government Response Tracker (OxCGRT))49. We add 1 to the daily number of COVID-19 confirmed case variable and implement a log transformation. The Stringency Index is the sum of eight containment and closure policy indicators together with the presence of public information campaigns (see49 for details). The three cross-country daily measures were merged with the self-reported country of residency and survey submission date variables. These variables were included as controls in the regression models, together with country and time fixed effects.

### Empirical approach

Appropriate statistical tests were chosen to perform the response comparisons between groups; both parametric and non-parametric tests were employed. Due to the ordinal nature of the response variables, we perform a non-parametric pairwise multiple comparison50 and adjust the false discovery rate using the Benjamini–Hochberg stepwise adjustments. We also report the results of the Kruskal–Wallis rank test of the hypothesis that responses from different fields are from the same population. We employed both mean comparison t-test and the non-parametric Wilcoxon rank-sum for comparison between US and non-US responses. For US and non-US difference within fields, we implemented the Bonferroni correction to account for multiple comparison by inflating the significance cut-off by five-fold. To calculate the effect size for these comparisons, we follow the transformation of Cohen’s d for ordinal data proposed by51, where d $${\text{ = ~2*z/}}\sqrt n$$. Exact p-values (two-tailed) are reported. Statistical analyses were performed using Stata MP 16.1.

### Consensus

To examine the degree of agreement among scientists, we closely follow the approach of52,53 when measuring consensus from variables with an ordinal scale. Consensus of the ordinal response variable X with i categories is defined as: $${\text{Consensus}}\left( X \right) = 1 + \mathop \sum \limits_{{i = 1}}^{n} p_{i} \log _{2} \left( {1 - \frac{{\left| {X_{i} - \mu _{X} } \right|}}{{max_{X} - min_{X} }}} \right)$$, where is the share of responses (excluding non-responses). A value of 0 indicates the participants’ responses are evenly split to the two extremes, while a value of 1 means that all responses are in the same category. The consensus score is around 0.45 (depending on the number of response categories) if responses are evenly split into each category. 95% confidence intervals (error bars) of the consensus measure are constructed by performing bootstrap resampling with 300 replications. We employ the two-sample t-test to test for statistically significant differences in the consensus scores between groups (across fields or US to non-US). Bonferroni adjustments were also used for multiple-field comparisons. Computing consensus using the Shannon Entropy equation, i.e., $$1 - \frac{{\sum p_{i} \times \ln p_{i} }}{{n \times 1/n \times \ln \left( {1/n} \right)}}$$, yields identical qualitative conclusions.

### Ordered logit regressions

We employed the ordered logit regression model to examine the effect of sample characteristics and other factors on the response outcome. The ordered logit model is a more suitable model than the commonly used ordinary least squared (OLS) model as it recognizes that the response data is ordinal rather than interval. The ordered logit coefficient indicates the expected increase in the log odds of being in a higher level of the response variable, given a 1-unit increase in the predictor variable, holding other variables in the model constant. For ease of interpretation, we report the estimated proportional odds ratios (by exponentiating the coefficients), which can be interpreted as the odds for being in a higher level of the response variable (i.e., proportional odds times larger).