The two subsamples (France and Italy) are similar in terms of age, income, and education (see supplementary table 1). To acquire information about individual personality traits, such as risk attitude, the DOSPERT psychometric test 16 in the health domain was administered. For the results, see the methods section.
For all subsamples, the revealed preferences for receiving the vaccine increased as the time of inoculation decreased. Answers are consistent from a temporal perspective: from now to 3 months, those intending to be vaccinated increase, and vaccine hesitancy decreases. In absolute terms, vaccine refusal is much higher in France than in Italy, as is vaccine hesitancy.
Globally, 39.7% of Italians refuse the vaccine if administered today, 25.6% in one month, and 19.4% in three months. In France, 63.6% refuse to be vaccinated today, 44.2% in one month, and 35.1% in three months. The higher vaccine hesitancy in France compared to Italy is consistent with previous studies17.
Instability of individual preferences over time.
Looking deeper into the dynamics of vaccine time preferences at the individual level, one can observe some interesting phenomena. Combining answers about the three times, one can see individuals as polarized between stable and unstable preferences. The three time perspectives that were elicited in the questionnaires are different from actual COVID-19 vaccine inoculation timeframes (now prescribed every five or six months). The questionnaire was designed in this way to better clarify the spectrum of human behavior hidden behind the usual “anti-vaxer” label. Respondents giving the same answer over the three different times (all answers yes vs no vs maybe; Tables 3, 4, 5) are 66.18% of the total sample. Respondents expressing two or three different preferences, namely, the other 24 combinations, represent 33.82%. A few examples are reported in Tables 6, 7, and 8.
Reading newspaper impacts the intended behaviour.
Out of the 27 possible combinations of answers (yes vs. no vs. maybe; three months vs one month vs today), 7 categories were abstracted according to the scheme shown in Table 9 to identify human types among respondents: 3 stable decision-makers (stable pro-vaxers, stable anti-vaxers, stable procrastinators) and 4 unstable.
We then generate the hypothesis that belonging to one of the seven categories is the dependent variable and that the perceived risks and expected benefits the independent risks. The categorization was validated by applying neural networks that showed, for all the treatment groups, a high predictive power, allowing a between-subjects comparison. For details of the neural networks, see the methods section.
The distributions of human behaviour categories in each group are shown in Tables 10 and 11.
In France, in the control group, stable pro-vaxers (category 1) account for 17.3%, and stable anti-vaxers (category 7) account for 29.1%.
In Italy, in the control group, category 1 represents 39.8% of the subsample, while category 7 represents 14.6%.
Categories 1 and 7 rapidly change after reading texts.
After reading the abstract text, anti-vaxers shift to 36% in France and 21% in Italy. Category 1 decreases to 12% in France and to 30% in Italy.
After reading concrete text, category 7 arrives at 35% in France and 17% in Italy, while category 1 arrives at 11% in France and 36% in Italy.
Both texts increase vaccine hesitancy, but the abstract text surprisingly affects readers more than the concrete text. It is worth noting that the increase in the anti-vaxers category is coupled with a decrease in the pro-vaxers category (category 1) and a slight decrease in the unstable categories.
Grouping together the unstable categories (2-3 and 5-6), one can observe a similar trend in both nations. In Italy they are 36.9% in the control; 33.5% in the abstract text; 30.7% in the concrete text; in France respectively 40%, 39.9%; and 37.2%.
Category 4, respondents choosing maybe for three times, in France, represents 14% of the control group, 12% in the abstract text group, and 16% in the concrete text group; in Italy respectively 9%, 15% and 17%.
No significant differences in vaccine-revealed preferences were detected among the control group and the two groups performing tasks.
Ordinary least square (OLS) multivariate analysis was applied to find the functional relationship among the 8 independent variables of the questionnaires and the category (dependent variable). The results are shown in Tables 12 and 13. The linear model, for all groups, is significant (based on an F test, a=0.05). After treatments, a change in the relative importance of the independent variables in all treatment groups is observable. The relevance of demographic variables recedes after reading the texts or performing tasks.
In regard to texts and tasks, vaccine expected benefit and perceived risk become the most relevant independent variables. Vaccine hesitancy increases after reading newspaper articles, and vaccine acceptance decreases. This phenomenon in our model based on human behavior categories appears to be driven by vaccine risk-return individual assessment.
Trade-off between male and female vaccine hesitancy.
In both nations, gender matters to the control group as an independent variable. The distribution per gender of stable pro-vaxers and stable anti-vaxers is shown in Tables 14 and 15. In the control group, in Italy, women accept vaccination more than men and are more likely to be in category 1 – pro-vaxers: 48.2% of women total sample (TS); 29.8% of men TS. In category 7 - anti-vaxers - 7.1% of women and 23.4% of men occur. In France, in category 1, women accounted for 21% of TS, and males accounted for 13.2%. In category 7, women accounted for 22.8%, and men accounted for 35.8%.
In contrast, in the abstract text group, in Italy, women accept the vaccine less than men: category 1 is made up of 23.1% of TS of women and 36.4% of men. In category 7, 38.5% of the TSs were women, and 3.9% were men. In Italy, in concrete text, 39.5% of females belong to category 1. Males in the same category were 29.3%. In category 7 (anti-vaxers), females are 15.1% of TS and 20% males.
Symmetrically, in the French abstract text group, in category 1, females account for 10.4% and males for 13.2%. Anti-vaxers of category 7 were 40.3% females and 31.6% males. In France, in the concrete text in category 1, the females were 13.8%, and the males were 7.8%. The prevalence of anti-vaxers was 7.9% in cats and 32.5% in males.