Who Gets Vaccinated? Cognitive and Non‐Cognitive Predictors of Individual Behaviour in Pandemics

Andor, Mark A.; Bauer, Thomas K.; Eßer, Jana; Schmidt, Christoph M.; Tomberg, Lukas Article — Published Version Who Gets Vaccinated? Cognitive and Non‐Cognitive Predictors of Individual Behaviour in Pandemics Oxford Bulletin of Economics and Statistics Provided in Cooperation with: John Wiley & Sons Suggested Citation: Andor, Mark A.; Bauer, Thomas K.; Eßer, Jana; Schmidt, Christoph M.; Tomberg, Lukas (2024) : Who Gets Vaccinated? Cognitive and Non‐Cognitive Predictors of Individual Behaviour in Pandemics, Oxford Bulletin of Economics and Statistics, ISSN 1468-0084, Wiley, Hoboken, NJ, Vol. 87, Iss. 3, pp. 562-585, https://doi.org/10.1111/obes.12644 This Version is available at: https://hdl.handle.net/10419/323874 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. http://creativecommons.org/licenses/by/4.0/ OXFORD BULLETIN OF ECONOMICS AND STATISTICS, 87, 3 (2025) 0305-9049 doi: 10.1111/obes.12644 Who Gets Vaccinated? Cognitive and Non-Cognitive Predictors of Individual Behaviour in Pandemics MARK A. ANDOR,†,‡ THOMAS K. BAUER,†,‡,§ JANA EßER,† CHRISTOPH M. SCHMIDT†,‡,§ and LUKAS TOMBERG† †RWI - Leibniz Institute for Economic Research, Hohenzollernstraße 1-3 Essen, 45128, Germany ‡Ruhr University Bochum, Universit¨ atsstraße 150 Bochum, 44801, Germany §IZA - Institute of Labor Economics, Schaumburg-Lippe-Str. 5-9 Bonn, 53113, Germany Abstract This study investigates different cognitive and non-cognitive characteristics associated with individuals’ willingness to get vaccinated against Covid-19 and their actual vaccination status. Our empirical analysis is based on data obtained from three survey waves conducted in 2021 among about 2,000 individuals living in the German state of North Rhine-Westphalia. We find that individuals with a high level of trait reactance – a personality characteristic that entails the personal tendency to perceive persuasion attempts as restricting one’s freedom – display a significantly lower willingness to get vaccinated. They also tend to get inoculated later or never. Moreover, neuroticism, locus of control, and statistical numeracy appear to be associated with the willingness to get vaccinated, but these results are less pronounced and less robust. Our results indicate that vaccination campaigns and policies could be improved by specifically addressing those with a high level of trait reactance. I. Introduction Starting at the end of 2019, the Covid-19 pandemic spread out to the entire world quickly and forced many governments to implement public health measures. To decelerate infection and hospitalization dynamics, authorities mandated, for example, the use of face masks and compliance with regulations to keep distance. As vaccines became available in 2020, particular hope was placed on vaccination campaigns, since inoculation is shown to prevent infections with Covid-19 from turning into severe cases and may even prevent infections altogether (CDC, 2022). The effectiveness of public health measures, including JEL Classification numbers: D01, I12, I18. We gratefully acknowledge financial support by the Ministerium f¨ ur Wirtschaft, Innovation, Digitalisierung und Energie des Landes Nordrhein-Westfalen. We thank the Editor Climent Quintana-Domeque, the anonymous referees, Manuel Frondel, and Colin Vance for helpful comments and suggestions. Furthermore, we thank our project collaborators Philipp Breidenbach, Maximilian Dirks, Katja Fels, Nils Christian H¨ onow, Matthias Kaeding, Delia Niehues, Stefan Rumpf, and Torsten Schmidt. Before obtaining our research data, we preregistered our hypotheses at the American Economic Association’s RCT registry (AEARCTR0007710). 562 ©2024 The Author(s). Oxford Bulletin of Economics and Statistics published by Oxford University and John Wiley & Sons Ltd. This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited. Predictors of individual behaviour in pandemics 563 vaccination campaigns, depends critically on popular acceptance of and adherence to these measures. However, in many countries, there has been reluctance or refusal to follow the health guidelines and even strong protests against some of these measures, especially against vaccination requirements. The search for explanations for this reluctance and refusal has led to a large body of research since the outbreak of the pandemic. Behavioural research has identified important cultural and group-level characteristics including a lack of trust (Bargain and Aminjonov, 2020), political polarization (Allcott et al.,2020) and cultural individualism (Bazzi, Fiszbein, and Gebresilasse, 2021; Chen, Frey, and Presidente, 2021) as important determinants of reluctance to follow health guidelines. On the individual level, Alfaro et al. (2022) and Fang et al. (2022) identified prosociality and social preferences as factors fostering compliance. With this paper, we complement these lines of research by focusing on the role of cognitive and non-cognitive characteristics. Cognitive characteristics such as statistical numeracy can be expected to be relevant to vaccination decisions because they may contribute to an individual’s ability to categorize the disparate statistical information released during the pandemic and to make an informed decision based on that information. In addition, non-cognitive characteristics, such as personality traits, may play an important role in vaccination decisions, for example, by moderating recipients’ reactions to persuasive messages (Hirsh, Kang, and Bodenhausen, 2012), by shaping an individual’s proneness to experience certain emotions (Widiger and Oltmanns, 2017), or by determining an individual’s tendency to behave prosocially (Andor et al.,2022). Specifically, we examine the association of trait reactance, neuroticism, locus of control beliefs, as well as statistical numeracy with individuals’ vaccination status and willingness to get vaccinated against Covid-19.1Our analysis is based on three waves of survey data of more than 2,000 individuals living in the German state of North Rhine-Westphalia collected between May and November 2021. As our main outcome, we observe the selfreported vaccination status and, if the person is not yet vaccinated, the stated willingness to get vaccinated. For brevity, we refer to our main categorical outcome variable hereafter as the willingness to get vaccinated, which includes the revealed preference of already being vaccinated as one category. Our results indicate that the willingness to get vaccinated against Covid-19 decreases significantly with an increasing level of trait reactance, a personality characteristic that entails the personal tendency to perceive persuasion attempts as restricting one’s freedom. Neuroticism, locus of control and statistical numeracy also appear to be associated with the willingness to get vaccinated, but these results are less pronounced and less robust with regard to the econometric specification than the results regarding trait reactance. The estimation of an Accelerated Failure Time (AFT) model further shows that individuals with a high level of trait reactance get vaccinated significantly later than those with a low or medium level of trait reactance, while the remaining characteristics do not seem to be associated with the timing of vaccination. 1Before obtaining our research data, we preregistered our hypothesis that these characteristics may be associated with individual compliance with protective measures against Covid-19 (Andor, 2021). ©2024 The Author(s). Oxford Bulletin of Economics and Statistics published by Oxford University and John Wiley & Sons Ltd. 564 Bulletin The results regarding trait reactance contribute to prior research in medicine and psychology, which has shown that trait reactance is negatively associated with individual efforts to follow official health recommendations (D´ ıaz and Cova, 2022)aswellas with vaccination intentions (Dra˙zkowski and Trepanowski, 2022). We complement these findings by showing that trait reactance is strongly associated with actual vaccination behaviour and that this association is of high economic significance, as the likelihood of being vaccinated against Covid-19 is 9 percentage points lower in the group of respondents with high trait reactance. Our results further suggest that higher levels of neuroticism, which have been shown to be related to self-protective behaviour or adherence to public protection policies (e.g. Kroencke et al.,2020), do not seem to play a major role with regard to vaccination behaviour. The same holds with regard to the roles of locus of control (Devereux, Miller, and Kirshenbaum, 2021; Olagoke, Olagoke, and Hughes, 2021)and statistical numeracy (Lau et al.,2022). Overall, our results underpin the importance of considering reactance when designing health policies and health communication campaigns (Reynolds-Tylus, 2019; Ball and Wozniak, 2022). Specifically, we conclude that campaigns and policies intended to encourage the unvaccinated to get inoculated should be specifically targeted at those with a high level of trait reactance. The paper proceeds as follows. In the next section, we provide a detailed description of our data set. In section III, we present empirical evidence on the relationship between cognitive and non-cognitive traits and individuals’ vaccination status and willingness to get vaccinated. Section IV presents the results of an AFT model to analyse the duration until first vaccination. Section Vputs the empirical results into perspective by reflecting on our key findings in comparison to the existing literature. In section VI, we discuss limitations of our study. Section VII concludes. II. Data Our empirical analysis is based on data obtained from three survey waves of about 2,000 individuals living in the German state of North Rhine-Westphalia conducted by RWI – Leibniz Institute for Economic Research in cooperation with the survey institute forsa throughout the year 2021. The first survey took place between 7 and 16 May (Wave 1), followed by two additional waves collected between 26 July and 21 August (Wave 2) and 20 October and 11 November (Wave 3), respectively. For Wave 2 and Wave 3, we re-sampled as many participants from Wave 1 as possible to establish a panel data structure. The survey population consisted of a sub-sample of forsa’s household panel. forsa maintains a large panel of potential survey participants that is representative for the population of German-speaking internet users. The data collection occurred as part of a research project that aimed at evaluating the epidemiological, economic and social effects of a controlled redemption of the severe lock-down measures imposed to fight the Covid-19 pandemic in certain municipalities of North Rhine-Westphalia.2These so-called Model Municipalities have been over-sampled during the collection of the data, i.e., half of the survey participants live in these municipalities. The surveys were conducted online 2These municipalities included the cities of Essen, Hamm, Cologne, Krefeld, M¨ onchengladbach, Lennestadt, Lippstadt, M¨ unster, and Soest, as well as the counties of Coesfeld, D¨ uren, Paderborn and Warendorf. ©2024 The Author(s). Oxford Bulletin of Economics and Statistics published by Oxford University and John Wiley & Sons Ltd. Predictors of individual behaviour in pandemics 565 using a tool that can be accessed on computers, tablets and smart phones. The target group were persons aged 18 and above. Approximately 3,000 respondents participated per survey wave. For our estimation sample, we consider only those individuals who participated in all three waves, which are 2,317 participants. Additionally, we exclude 100 participants who did not answer all questions relevant for our analysis, i.e., questions on their willingness to get vaccinated (elicited in all three survey waves) and their month of vaccination if they have already received their first vaccination dose (elicited in Wave 2), questions on cognitive and non-cognitive characteristics, as well as on age, gender and education level (elicited in Wave 1). To obtain information on individuals’ willingness to get vaccinated, we asked in May 2021: ‘Vaccinations against Corona started in Germany at the end of December. Have you already been vaccinated against Corona,will you get vaccinated as soon as you have the chance,or would you rather wait or not get vaccinated at all?’, allowing the following answers: •‘I have already been vaccinated against Corona’ (Vaccinated) •‘Will get vaccinated as soon as I have the chance’ (Willing) •‘Will rather wait’ (Waiting) •‘Will not be vaccinated at all’ (Unwilling) •‘Don’t know / not specified’ We count the 13 respondents who reported that they did not know or preferred not to specify their vaccination status as part of the 100 excluded participants who did not answer all questions relevant to our analysis. In addition, we delete 37 respondents from the sample as they provided inconsistent responses with regard to their vaccination dates.3 The final estimation sample consists of 2,180 individuals (see Figure 1for a visual representation of the exclusions).4 3These respondents stated in May that they had already been vaccinated but indicated in Wave 2 that they had been vaccinated in June or July. 4In the following, we examine whether there are differences between the participants who are included and participants who are not included, for example because they did not participate in all three survey waves. First, participants who were excluded due to item non-response do not differ from the final estimation sample with respect to gender and age, the two variables that are available for all participants regardless of response behaviour (Table A2). Those excluded because of answering questions on their vaccination status or their psychological characteristics with ‘Don’t know / not specified’ tend to be slightly less educated than the final estimation sample (Table A3). Table A1 shows that those participants who did not participate in Waves 2 and 3 are significantly less likely to be vaccinated but more likely to be willing to get vaccinated in Wave 1. This may be due to their significantly lower age. Additionally, participants who did not participate in Waves 2 and 3 reported lower levels of neuroticism. As a robustness check, we therefore conduct an additional analysis (Table A14) with the larger sample of 2,893 individuals who participated in Wave 1, regardless of their participation in subsequent waves. Specifically, we estimate our main regression from Table A2 – in which we examine the association between the cognitive and non-cognitive characteristics and the willingness to get vaccinated using multinomial logit models – with the larger sample. The results remain qualitatively the same, in particular we find a robust negative association between trait reactance and being vaccinated. Furthermore, we find evidence of significant associations between the vaccination status or the willingness to get vaccinated and neuroticism, locus of control, and statistical numeracy. For our analyses in the main body of the paper, we focus on the sample of 2,180 respondents who participated in all three survey waves, as this allows a direct comparison of the analyses on vaccination status across all three survey waves and the analyses on vaccination timing due to the constant sample across all different specifications. ©2024 The Author(s). Oxford Bulletin of Economics and Statistics published by Oxford University and John Wiley & Sons Ltd. 566 Bulletin FIGURE 1. Exclusions from the final estimation sample FIGURE 2. Vaccination status and willingness to get vaccinated in May 2021 Figure 2shows the distribution of the willingness to get vaccinated in May 2021 as it appears in our final sample. A majority of 65% of the respondents was already vaccinated in May 2021, and most of the respondents who were not yet vaccinated were at least willing to get vaccinated (28% of the sample). Only a minority of about 4% and 3% of the individuals in our sample preferred to wait or reported to not get vaccinated at all, respectively. More detailed descriptive information on the distribution of the willingness to get vaccinated across socio-economic groups is presented in Table A4. In Wave 2, conducted in July and August 2021, we asked the respondents who were already vaccinated at this point in time (94.6%) about the month in which they received their first vaccination dose. These vaccination dates were influenced by a priority ©2024 The Author(s). Oxford Bulletin of Economics and Statistics published by Oxford University and John Wiley & Sons Ltd. Predictors of individual behaviour in pandemics 567 FIGURE 3. Kaplan–Meier plot of vaccination months as stated in July/August 2021 differentiated by the vaccination intentions stated in May 2021 scheme imposed by the German government to ensure that the most vulnerable groups for infections were vaccinated first. From the start of the vaccination campaign in December 2020 until 6 April 2021, persons aged 80 or older as well as persons who work in nursing, elderly care, and health care were prioritized. From 7 April 2021, to 6 May 2021, the prioritization was extended to persons aged 70 or older as well as to chronically ill persons and their close relatives or caregivers, contact persons of pregnant persons, as well as persons in public service, such as the police, fire department, and school service. From 7 May 2021, to 6 June 2021, the prioritization was further extended to persons aged 60 or older and to further professional groups that have a lot of contact with customers or work in system-relevant utilities. From 7 June 2021, there was no longer any special prioritization and everyone could get vaccinated. Figure 3contrasts the vaccination intentions elicited in Wave 1 with the vaccination timing elicited in Wave 2. It appears that the large majority of those who stated in May that they will get vaccinated as soon as possible indeed got vaccinated in May, June or July. Only a small share of this group was still not vaccinated at the time of Wave 2 (3.4%). In contrast, 57.5% of those who stated that they would rather wait before getting vaccinated in Wave 1 and 82.3% of those who stated no intention to get vaccinated at all were still not vaccinated in Wave 2. Table 1presents summary statistics of the outcome and explanatory variables used in our empirical analysis. The latter can be categorized into two groups: Variables describing the cognitive and non-cognitive characteristics of the individuals in our sample, which are in the focus of this study, and socio-economic characteristics. We measured five different cognitive and non-cognitive characteristics that we considered relevant to vaccination decisions based on the scientific literature5as well as theoretical considerations: trait reactance, neuroticism, locus of control and statistical numeracy. 5On trait reactance, see D´ ıaz and Cova (2022) and Ball and Wozniak (2022). On neuroticism, see Kroencke et al. (2020). On locus of control, see Devereux et al. (2021) and Olagoke et al. (2021). Regarding risk and health literacy in the context of pandemics and vaccination, statistical numeracy is considered a key element (Riiser et al.,2020; Greer et al.,2021). ©2024 The Author(s). Oxford Bulletin of Economics and Statistics published by Oxford University and John Wiley & Sons Ltd. 568 Bulletin TABLE 1 Summary statistics based on the first survey wave (May 2021) Variable Explanation Mean (I) Outcome variables Vaccinated Dummy: 1 if respondent is already vaccinated 0.65 Willing Dummy: 1 if respondent is willing to get vaccinated as soon as possible 0.28 Waiting Dummy: 1 if respondent prefers to wait 0.04 Unwilling Dummy: 1 if respondent is unwilling to get vaccinated 0.03 (II) Cognitive and non-cognitive characteristics Trait reactance Low Dummy: 1 if respondent has a low level of trait reactance 0.24 Medium Dummy: 1 if respondent has a medium level of trait reactance 0.32 High Dummy: 1 if respondent has a high level of trait reactance 0.44 Neuroticism Low Dummy: 1 if respondent has a low level of Neuroticism 0.33 Medium Dummy: 1 if respondent has a medium level of Neuroticism 0.29 High Dummy: 1 if respondent has a high level of Neuroticism 0.38 Locus of Control Low Dummy: 1 if respondent has a low internal locus of control 0.30 Medium Dummy: 1 if respondent has a medium internal locus of control 0.31 High Dummy: 1 if respondent has a high internal locus of control 0.39 Lipkus et al. numeracy scale Low Dummy: 1 if respondent scores low on the selected items of the numeracy scale 0.30 High Dummy: 1 if respondent scores high on the selected items of the numeracy scale 0.70 Berlin numeracy test Dummy: 1 if respondent passed the Berlin Numeracy Test (short version) 0.25 (III) Socio-economic characteristics Female Dummy: 1 if respondent is a woman 0.47 Age Age of respondent 57.8 (standard deviation) (14.4) College degree Dummy: 1 if respondent has a college degree 0.34 Trait reactance is proxied using three selected items of the Hong Psychological Reactance Scale (Hong and Faedda, 1996). The personality construct that this scale is supposed to measure is psychological reactance, defined as ‘the motivational state that is hypothesized to occur when a freedom is eliminated or threatened with elimination’ (Brehm and Brehm, 1981, p. 37) and ‘reactance produces a desire to restore one’s attitudinal or behavioral freedom’ (Shen and Dillard, 2005, p. 74). While reactance can be viewed as a motivational state, it is argued that it can also be viewed as a personality trait that determines a person’s proneness to feel and act in a reactant manner (Shen and Dillard, 2005). Neuroticismwasretrievedusingtheneuroticism-relateditemsoftheBigFiveInventory (BFI-S) that is used in the German Socio-Economic Panel (Schupp and Gerlitz, 2008). This personality trait has two poles: neuroticism on one side and emotional stability on the other. Persons who score high on neuroticism have a stronger tendency to experience negative emotions such as anxiety, anger, guilt and depression (Widiger, 2009, p. 129). Locus of control was captured using the original items of the Psychological Coping Resources component of the Mastery Module by Pearlin and Schooler (1978), which is, for example, also used to measure locus of control in Cobb-Clark and Schurer (2013). ©2024 The Author(s). Oxford Bulletin of Economics and Statistics published by Oxford University and John Wiley & Sons Ltd. Predictors of individual behaviour in pandemics 569 Locus of control captures an individual’s belief about how strongly events in her life are shaped by own behaviour (Rotter, 1966;GatzandKarel,1993). Individuals who believe that events strongly depend on own behaviour have a high internal locus of control, while those believing that events are mainly determined by fate or luck, for instance, have an external locus of control. Statistical numeracy was measured in two ways. First, we utilized three selected items of Lipkus, Samsa, and Rimer’s (2001) extended numeracy scale, which were framed in a health context. These items were relatively simple percentage calculation tasks. Therefore, we used a second indicator of statistical numeracy by employing the short version of the Berlin Numeracy Test proposed by Cokely et al. (2012), which is a more demanding test of statistical numeracy.6 For our empirical analysis, we first calculate the individual mean responses on the Likert scales for all items of each characteristic. Second, we divide the sample for each characteristic into three approximately equally sized categories: Low (the mean of the responses is below the lower tercile), medium (the mean of the responses is between the lower and the upper tercile), high (the mean of the responses is above the upper tercile).7 Exceptions are the measures of statistical numeracy. The measure based on Lipkus et al. (2001) yields only one binary response (correct or incorrect/’don’t know’) per item. To aggregate the items, we calculate the sum of the correct answers per respondent, which reveals that almost 70% of the respondents answered all items correctly. Since this 70% includes both the lower and the upper tercile, there are only two categories for this measure: Low statistical numeracy (at least one incorrect or ‘don’t know’ response) and high statistical numeracy (all responses are correct). Similarly, the short form of the Berlin Numeracy test produces only one binary result (correct or incorrect/’don’t know’). Therefore, we use a dummy variable that is 1 if the correct answer to the test was given, which is true for 25% of the individuals in our sample, and 0 if the answer was wrong or the answer option ‘don’t know’ was chosen. Table A5 compares the study sample with the population of the German state of North Rhine-Westphalia. The results show that individuals aged 19–45 are underrepresented in our sample, while the 45–65 and 65–75 age groups are overrepresented. The proportion of women in our sample is slightly lower, while the proportion of college graduates is higher than in the general population. Consistent with the higher average age of the study population, there are more married and fewer employed persons in the sample and fewer persons with children living in the household than in the population. The average income category in the sample is slightly lower than the average net household income in the population. Finally, the proportion of vaccinated individuals in our sample is quite high 6See Cokely et al. (2012) for a detailed discussion on the differences between the two measurement instruments. 7Since the mean responses tend to consist of a limited number of discrete values due to the small number of items per trait, there is a high probability for ties in which one of these discrete values lies exactly on the tercile boundary. In this case, we assign all respondents with this value to the higher group. For this reason, the respective groups are not exactly of the same size. In Appendix E, we conduct robustness checks and form the groups in an alternative way: The low category consists of those individuals whose rounded mean response on the Likert scale of all items of a characteristic is either 1 or 2. The medium category consists of those whose mean response is 3 (on 5-point Likert scales) or 3, 4 or 5 (on 7-point Likert scales). The high category consists of the remainder. This approach results in more unequally sized groups, where the low and high categories tend to consist of those with a rather extreme realization of the respective characteristic. ©2024 The Author(s). Oxford Bulletin of Economics and Statistics published by Oxford University and John Wiley & Sons Ltd. 576 Bulletin TABLE 3 Coefficients of Accelerated Failure Time models (Weibull distribution) on the time until vaccination (1) (2) (3) (4) (5) (6) Medium trait reactance 0.020 0.020 (0.013) (0.012) High trait reactance 0.088** 0.087** (0.012) (0.012) Medium neuroticism 0.008 0.001 (0.013) (0.013) High neuroticism −0.008 −0.015 (0.012) (0.012) Medium locus of control 0.004 0.010 (0.013) (0.013) High locus of control −0.015 −0.004 (0.012) (0.012) Low on Lipkus et al. numeracy scale −0.002 0.002 (0.012) (0.012) Berlin numeracy test passed 0.011 0.013 (0.014) (0.013) College degree −0.046** −0.049** −0.046** −0.048** −0.050** −0.049** (0.011) (0.011) (0.011) (0.011) (0.011) (0.011) Female −0.002 −0.002 −0.004 −0.003 −0.003 0.000 (0.010) (0.011) (0.010) (0.011) (0.011) (0.011) Age≥80 −0.305** −0.315** −0.310** −0.312** −0.311** −0.304** (0.031) (0.032) (0.033) (0.033) (0.033) (0.030) 80>Age≥70 −0.191** −0.203** −0.202** −0.202** −0.200** −0.190** (0.012) (0.012) (0.012) (0.012) (0.013) (0.013) 70>Age≥60 −0.111** −0.118** −0.114** −0.116** −0.115** −0.111** (0.014) (0.014) (0.014) (0.014) (0.014) (0.014) Constant 1.844** 1.897** 1.899** 1.896** 1.892** 1.845** (0.012) (0.012) (0.012) (0.010) (0.011) (0.017) ln(p) 1.569** 1.551** 1.552** 1.551** 1.551** 1.571** (0.021) (0.022) (0.022) (0.022) (0.022) (0.021) Notes: The sample is restricted to those respondents who participated in all three waves. Vaccination months were elicited in Wave 2 and the covariates were elicited in Wave 1. Here, we control for age categories rather than continuous age, which is due to the nature of the vaccination prioritization scheme that made the vaccine accessible in a staggered way according to the exact age cutoffs used here to form the age groups. Robust standard errors are reported in parentheses. Number of observations: 2,180. ** and * indicate statistical significance at the 1% and 5% level, respectively. Again, we compare the effect of high trait reactance to the effect of the education level and find that the 9% increase in the duration to first vaccination in the high trait reactance group is about twice as large in absolute terms as the 5% decrease in the duration to first vaccination in the college degree group. The results assuming a Log-Logistic distribution corroborate the main results from the Weibull model (Table A20). The exception is that in this model, a high internal locus of control seems to be significantly related to a shorter time until vaccination. Yet, this effect becomes non-significant in the combined model (Column (6) of Table A20). This suggests that the results concerning locus of control are less robust than those concerning reactance. An additional robustness check is presented in Table A21. This robustness check examines the extent of possible time aggregation bias, which arises because the observed ©2024 The Author(s). Oxford Bulletin of Economics and Statistics published by Oxford University and John Wiley & Sons Ltd. Predictors of individual behaviour in pandemics 577 vaccination dates are discrete in nature due to the monthly observation rhythm, even though the model assumes continuous data (Petersen, 1991). To assess the sensitivity of our results in this regard, we follow Allison (2010) and recode our data to be interval censored, i.e., rather than assigning respondents a specific month-indicator as the outcome variable, we assign them an interval where the upper interval bound is the month-indicator used before and the lower interval bound is the previous month-indicator plus an epsilon of 0.0001. The latter has the purpose of avoiding zeros for respondents who were vaccinated in the first month observed. Respondents who were not vaccinated at the end of the observation period are assigned an interval that has an open upper interval bound, while the lower interval bound takes the value of the latest observed month. This interval censored data can also be analysed by a Weibull AFT model. The results obtained by using this interval censored model are qualitatively similar to the ones in Table 3(Table A21). In tendency, the coefficients of the interval censored model are somewhat larger, but there are no changes in the signs or significance of the results. As a final robustness check, we examine the extent to which the results are driven by individuals who were not yet vaccinated at the time the vaccination dates were elicited and who therefore enter the AFT model as right-censored observations. To do this, we omit these individuals from the estimation sample (Table A22). The signs and significance of the resulting coefficients are similar to the results in Table 3. A high level of trait reactance is still significantly associated with a later vaccination date, although the magnitude of this relationship is smaller than in the model that includes unvaccinated individuals. V. Putting the results into perspective: trait reactance and vaccination Our results suggest that trait reactance is the strongest predictor of vaccination behaviour. To put our finding into perspective, we identified nine studies that also investigate the relationship between trait reactance and attitudes towards vaccination (see Table 4for an overview). These studies support our finding by showing that trait reactance is a predictor of hesitancy to get vaccinated in samples from the USA (Albarracin et al.,2021), Poland (Dra˙zkowski and Trepanowski, 2022), Turkey (Salali et al.,2022), and Finland (Soveri et al.,2023). Compared to our study, which analyses actual vaccination behaviour against Covid-19 at several points in time, the focus of these studies was more on finding out what the hypothetical willingness to get vaccinated was before the vaccines against Covid-19 were available (Dra˙zkowski and Trepanowski, 2022; Soveri et al.,2023), how high the intention to be vaccinated is among those who were not vaccinated at the time of the study, i.e., the self-reported likelihood of being vaccinated in the future (Salali et al.,2022), or it was asked about the willingness to get vaccinated against a hypothetical new disease (Albarracin et al.,2021). Research has also shown that trait reactance not only predicts hesitancy to get vaccinated oneself in the context of the Covid-19 pandemic, but also correlates with opinions about childhood vaccination in general prior to the pandemic (Hornsey, Harris, and Fielding, 2018; Finkelstein et al.,2020; Soveri et al.,2020). Furthermore, two studies have investigated whether trait reactance moderates the effects of experimental manipulations on attitudes towards vaccination: Albarracin et al. (2021) found that individuals with higher trait reactance do not react differently to mandatory Covid-19 vaccine requirements than individuals with lower trait reactance. ©2024 The Author(s). Oxford Bulletin of Economics and Statistics published by Oxford University and John Wiley & Sons Ltd. 578 Bulletin TABLE 4 Literature overview on the relationship between trait reactance and vaccination behaviour and attitudes Reference Study Country Sample Size Outcome variable Relation between the outcome variable and trait reactance P-value Salali et al. (2022) 1 Turkey 1,013 Vaccination intention Negative 0.001 Albarracin et al. (2021) 1 USA 357 Vaccination intention N.a. <0.001 Dra˙zkowski and Trepanowski(2022) 1 Poland 551 Vaccination intention Negative <0.01 Finkelstein et al. (2020) 1 USA 300 Vaccination priority for children Negative <0.01 Hornsey et al. (2018) 1 International, 25 countries 5,323 Attitudes toward vaccination Negative <0.001 Soveri et al. (2020) 1 Finland 770 Influenza vaccination behaviour Negative 0.001 Vaccine attitudes (general) Negative <0.001 Vaccine attitudes (Influenza) Negative <0.001 Wright and Rune (2023) 1 Australia 1,050 Beliefs about vaccine safety and efficacy Negative <0.05 Soveri et al. (2023) 1 Finland 199 Vaccination willingness Negative <0.001 Vaccination attitudes Negative <0.001 2 Finland 293 Vaccination willingness Negative <0.001 3 Finland 398 State reactance Negative <0.001 Vaccination willingness in t=1 Negative <0.001 Change in vaccination willingness between t=1andt=2 Negative 0.782 Notes: For the outcome variables, higher scores indicate more positive attitudes and beliefs, and higher intentions and probabilities to be vaccinated. By contrast, Soveri et al. (2023) found that requirements to get vaccinated against a hypothetical disease as well as additional information on the vaccine increases the willingness to get vaccinated among individuals with low trait reactance, but not among those with high trait reactance. Neither study found evidence of vaccine requirements backfiring, i.e., vaccine requirements did not reduce the willingness to get vaccinated. The different and therefore puzzling findings of Albarracin et al. (2021) and Soveri et al. (2023) regarding reactance are consistent with the results of Laurin, Kay, and Fitzsimons (2012). Their findings indicate that reactance may be particularly important when rules are not absolute, i.e., when it is not certain whether a rule would come into effect or when a rule is not credibly enforced. By contrast, people tend to rationalize and ©2024 The Author(s). Oxford Bulletin of Economics and Statistics published by Oxford University and John Wiley & Sons Ltd. Predictors of individual behaviour in pandemics 579 thus accept rules that are absolute. Consequently, the association between the willingness to get vaccinated and reactance may be weaker under mandatory vaccination. During the Covid-19 pandemic, North Rhine-Westphalia came closest to mandatory vaccination with the so-called ‘2G regulation’, which restricted access to places and events of social life to only those who had been vaccinated or had recovered. This regulation first came into effect on 24 November 2021 (MAGS, 2021), i.e., after Wave 3 was conducted. During our study period, the pressure to get vaccinated was primarily based on non-absolute rules, i.e., on social pressure, public appeals, and increasing hurdles for unvaccinated individuals, e.g., compulsory tests. However, these measures still allowed unvaccinated individuals to participate in social life. Beyond the topic of vaccination behaviour, it is evident that trait reactance plays a significant role in determining behaviour in aspects of social life. Trait reactance, as operationalized by Hong and Faedda (1996), has been demonstrated to predict real behaviour in multiple domains such as risky sexual activity, drug use (Miller and Quick, 2010), and behaviour in close relationships (Chartrand, Dalton, and Fitzsimons, 2007). Furthermore, it has been shown that reactance moderates the relationship between electronic surveillance and counterproductive behaviour (Yost et al.,2019), between promotional favours and consumer spending (Bertini and Aydinli, 2020), and between hotels’ environmental initiatives and consumers’ intention to be environmentally conscious (Wang, Krishna, and McFerran, 2017). To summarize, this section supports two points. First, we note that our results are consistent with previous findings on the relationship between trait reactance and vaccination intentions, which further strengthens our conclusions. Second, we can highlight our contribution to the literature as we observe actual vaccination behaviour against Covid-19 rather than hypothetical behaviour or intentions. Furthermore, we look at different cognitive and non-cognitive characteristics in one study rather than focusing on a single trait. VI. Limitations This study is subject to limitations that are typical for survey-based correlation studies. These include the possibility that the sample differs from the population in certain observable and unobservable factors. However, as discussed in section II, we expect that the elaborate offline recruitment process of the survey panel from which the respondents were randomly drawn reduces the severity of this problem compared to pure online surveys. Furthermore, self-reported health measurements are susceptible to various types of reporting bias (see, e.g., Bound, 1991; Kerkhofs and Lindeboom, 1995; Kapteyn, Smith, and van Soest, 2007; Black et al.,2017a; Black, Johnston, and Suziedelyte, 2017b; Davillas, de Oliveira, and Jones, 2023), which may also be the case with our measure of vaccination status. For example, due to the high social pressure and public attention on vaccination against Covid-19 during the study period, self-reports on vaccination status may have been subject to social desirability bias (Wolter et al.,2022). However, we believe such biases to be unlikely to invalidate our results. First, the response options ‘Will get vaccinated as soon as I have the chance’ or ‘Will rather wait’ allow respondents ©2024 The Author(s). Oxford Bulletin of Economics and Statistics published by Oxford University and John Wiley & Sons Ltd. 580 Bulletin to answer truthfully that they have not yet been vaccinated without directly revealing that they actually do not want to get vaccinated at all. This would mean that we would expect some of those who chose one of these options to be actually not willing to get vaccinated and may only have chosen these options because they are perceived as less socially disapproved of than the option ‘Will not be vaccinated at all’. As a robustness check, we therefore re-estimated our main analyses using a binary dependent variable that takes the value 1 only if someone has already been vaccinated (Table A17), with all three other options coded as 0. The robustness check confirms our results. Additionally, the vaccine was still in short supply, particularly in May 2021, meaning that the majority of Germans had not yet been vaccinated. Therefore, it cannot be assumed that a respondent who was not vaccinated would have felt compelled by social desirability to make a false statement about their vaccination status. Third, the survey was conducted anonymously online, and participants were informed that their information could not be traced back to them. Finally, our main result on trait reactance remains evident when considering the timing of respondents’ vaccination, where social desirability or other motivated misreporting is even less likely (Table 3). Moreover, our results describe the associations between vaccination and the different cognitive and non-cognitive characteristics, rather than causal effects. Ideally, future research can provide causal evidence of the effects of specific characteristics on vaccination behaviour. However, for policy makers and also for society, it might be even more important to build upon the results of our and related studies and try to use knowledge on the predictors of vaccination behaviour to target vaccination campaigns. We show that trait reactance in particular is an important predictor of vaccination behaviour. Individuals with a high level of trait reactance display a significantly lower willingness to get vaccinated.11 Interventions could target especially those with high trait reactance and should, for example, try to overcome perceptions of freedom threat, which may be very pronounced among these individuals (see, for instance, the literature review by Reynolds-Tylus, 2019). Such approaches could be accompanied by scientific research to evaluate the causal effects of targeted interventions, for example by means of randomized controlled field experiments, and to investigate how these effects are moderated by individual differences in trait reactance. 11 Like probably all studies, we cannot observe all factors that may affect vaccination decisions. For example, we lack information on the health status and medical history of the participants, although it seems plausible that this has an influence on the willingness to be vaccinated. We would expect people in poor health to be more willing to be vaccinated (although this could also be heterogeneous and requires its own investigation). Assuming this is the case, it is an omitted variable that could distort our results. In particular, if people who are reactant are also more likely to have a poorer health status or, conversely, people with a poorer health status are more likely to be reactant, it could affect the estimation results. While we consider it useful to investigate these relationships, it is important to emphasize that this does not imply a limitation regarding the predictive power of trait reactance. We show a robust association that people with higher trait reactance are less likely to be vaccinated, which can be used to target interventions. This is true even if the underlying health status would actually be fully mediating this association. Ultimately, however, we would not assume that the correlation between trait reactance and the willingness to get vaccinated is only due to the health status. ©2024 The Author(s). Oxford Bulletin of Economics and Statistics published by Oxford University and John Wiley & Sons Ltd. Predictors of individual behaviour in pandemics 581 VII. Conclusions Using data obtained from three survey waves of about 2,000 individuals living in the German state of North Rhine-Westphalia, this paper investigates the association of different cognitive and non-cognitive characteristics with an individuals’ willingness to get vaccinated against Covid-19 as well as their vaccination status. The empirical results indicate that trait reactance is the strongest predictor of vaccination behaviour. The probability of being vaccinated against Covid-19 is 9 percentage points lower for individuals with a high level of trait reactance, compared to those with a low or medium level of trait reactance. In addition, the time to vaccination is generally 9% higher for such individuals. Furthermore, we find tentative evidence that higher statistical numeracy, a higher internal locus of control, and a more pronounced level of neuroticism are related to vaccination intentions. However, these results are less robust than those regarding trait reactance. With regard to policy implications, our results indicate that vaccination campaigns and policies could be improved by specifically addressing certain groups of people, in particular those with a high level of trait reactance. To create such targeted measures, we can build on the literature predating the Covid-19 pandemic, which discusses ways to overcome reactance in health communication (see, for instance, the literature review by Reynolds-Tylus, 2019). For example, future vaccination campaigns could use narratives that promote empathy, as Shen (2010) has shown that state empathy can attenuate reactance regarding persuasive messages related to smoking and alcohol use. Moreover, persuasive messages may differ in the extent to which they are perceived as freedom threatening. According to the results of a meta-analysis by Rains (2013), the use of less freedom-threatening language may reduce reactant reactions to such messages. More generally, by demonstrating that reactance has high explanatory power, and by using an easy-to-implement measure of trait reactance, our study may provide an impetus for using this concept in other behavioural economic studies to explain why some groups are not reached by certain types of interventions. The Covid-19 pandemic was a situation in which the state strongly attempted to influence individual behaviour in the short term. A similar situation arose in the winter of 2022/2023 when Europe experienced an energy shortage due to the war in Ukraine, and European governments urged private households to save as much energy as possible. But even beyond immediate crises, there are many government interventions targeting individual behavioural changes, for example in the areas of preventive health care and retirement savings. In all these situations, reactance can be an important factor limiting the effectiveness of interventions. Acknowledgment Open Access funding enabled and organized by Projekt DEAL. ©2024 The Author(s). Oxford Bulletin of Economics and Statistics published by Oxford University and John Wiley & Sons Ltd. 582 Bulletin Final Manuscript Received: August 2023 References Albarracin, D., Jung, H., Song, W., Tan, A. and Fishman, J. (2021). ‘Rather than inducing psychological reactance, requiring vaccination strengthens intentions to vaccinate in us populations’, Scientific Reports, Vol. 11, p. 20796. Alfaro, L., Faia, E., Lamersdorf, N. and Saidi, F. (2022). ‘Health externalities and policy: the role of social preferences’, Management Science, Vol. 325, pp. 6751–6761. Allcott, H., Boxell, L., Conway, J., Gentzkow, M., Thaler, M. and Yang, D. (2020). ‘Polarization and public health: Partisan differences in social distancing during the coronavirus pandemic’, Journal of Public Economics, Vol. 191, 104254. Allison, P. D. (2010). Survival Analysis Using SAS: A Practical Guide 2nd ed., SAS Press, Cary, North Carolina. Andor, M. A. (2021). ‘Corona NRW – Further Analyses’, AEA RCT Registry, May 21. https://doi.org/10 .1257/rct.7710-1.0. Andor, M. A., Cox, J., Gerster, A., Price, M., Sommer, S. and Tomberg, L. (2022). Locus of Control and Prosocial Behavior, National Bureau of Economic Research (NBER) Working Paper No. w30359. Ball, H. and Wozniak, T. R. (2022). ‘Why do some americans resist covid-19 prevention behavior? An analysis of issue importance, message fatigue, and reactance regarding covid-19 messaging’, Health Communication, Vol. 37, pp. 1812–1819. Bargain, O. and Aminjonov, U. (2020). ‘Trust and compliance to public health policies in times of covid-19’, Journal of Public Economics, Vol. 192, 104316. Bazzi, S., Fiszbein, M. and Gebresilasse, M. (2021). ‘”Rugged individualism” and collective (in)action during the covid-19 pandemic’, Journal of Public Economics, Vol. 195, 104357. Bertini, M. and Aydinli, A. (2020). ‘Consumer reactance to promotional favors’, Journal of Retailing,Vol. 96, pp. 578–589. Black, N., Johnston, D. W., Shields, M. A. and Suziedelyte, A. (2017a). ‘Who provides inconsistent reports of their health status? The importance of age, cognitive ability and socioeconomic status’, Social Science & Medicine, Vol. 191, pp. 9–18. Black, N., Johnston, D. W. and Suziedelyte, A. (2017b). ‘Justification bias in self-reported disability: new evidence from panel data’, Journal of Health Economics, Vol. 54, pp. 124–134. Borga, L. G., Clark, A. E., D’Ambrosio, C. and Lepinteur, A. (2022). ‘Characteristics associated with covid-19 vaccine hesitancy’, Scientific Reports, Vol. 12, p. 12435. Bound, J. (1991). ‘Self-reported versus objective measures of health in retirement models’, Journal of Human Resources, Vol. 26, pp. 106–138. Brehm, S. S. and Brehm, J. W. (1981). Psychological Reactance: A Theory of Freedom and Control, Academic Press, New York, New York. Cameron, A. C. and Trivedi, P. K. (2005). Microeconometrics: Methods and Applications, Cambridge University Press, New York, New York. Cascini, F., Pantovic, A., Al-Ajlouni, Y., Failla, G. and Ricciardi, W. (2021). ‘Attitudes, acceptance and hesitancy among the general population worldwide to receive the covid-19 vaccines and their contributing factors: A systematic review’, eClinicalMedicine, Vol. 40, 101113. CDC. (2022). Covid-19 Vaccines are Effective, Centers for Disease Control and Prevention (CDC), Atlanta, Georgia. last update: June 28, 2022. https://www.cdc.gov/coronavirus/2019-ncov/vaccines/effectiveness /work.html, retrieved on 12 December 2022. Chartrand, T. L., Dalton, A. N. and Fitzsimons, G. J. (2007). ‘Nonconscious relationship reactance: when significant others prime opposing goals’, Journal of Experimental Social Psychology, Vol. 43, pp. 719–726. Chen, C., Frey, C. B. and Presidente, G. (2021). ‘Culture and contagion: individualism and compliance with covid-19 policy’, Journal of Economic Behavior & Organization, Vol. 190, pp. 191–200. Cobb-Clark, D. A. and Schurer, S. (2013). ‘Two economists’ musings on the stability of locus of control’, The Economic Journal, Vol. 123, pp. F358–F400. ©2024 The Author(s). Oxford Bulletin of Economics and Statistics published by Oxford University and John Wiley & Sons Ltd. Predictors of individual behaviour in pandemics 583 Cokely, E. T., Galesic, M., Schulz, E., Ghazal, S. and Garcia-Retamero, R. (2012). ‘Measuring risk literacy: the Berlin Numeracy Test’, Judgment and Decision Making, Vol. 7, pp. 25–47. Cox, D. R. (1972). ‘Regression models and life-tables’, Journal of the Royal Statistical Society: Series B (Methodological), Vol. 34, pp. 187–202. Crossley, T. F., Fisher, P. and Low, H. (2021). ‘The heterogeneous and regressive consequences of covid-19: evidence from high quality panel data’, Journal of Public Economics, Vol. 193, 104334. Davillas, A., de Oliveira, V. H. and Jones, A. M. (2023). ‘Is inconsistent reporting of self-assessed health persistent and systematic? Evidence from the UKHLS’, Economics & Human Biology, Vol. 49, 101219. Devereux, P. G., Miller, M. K. and Kirshenbaum, J. M. (2021). ‘Moral disengagement, locus of control, and belief in a just world: individual differences relate to adherence to covid-19 guidelines’, Personality and Individual Differences, Vol. 182, 111069. D´ ıaz, R. and Cova, F. (2022). ‘Reactance, morality, and disgust: the relationship between affective dispositions and compliance with official health recommendations during the covid-19 pandemic’, Cognition and Emotion, Vol. 36, pp. 120–136. Dra˙zkowski, D. and Trepanowski, R. (2022). ‘Reactance and perceived disease severity as determinants of covid-19 vaccination intention: an application of the theory of planned behavior’, Psychology, Health & Medicine, Vol. 27, pp. 2171–2178. Fang, X., Freyer, T., Ho, C.-Y., Chen, Z. and Goette, L. (2022). ‘Prosociality predicts individual behavior and collective outcomes in the covid-19 pandemic’, Social Science & Medicine, Vol. 308, 115192. Finkelstein, S. R., Boland, W. A., Vallen, B., Connell, P. M., Sherman, G. D. and Feemster, K. A. (2020). ‘Psychological reactance impacts ratings of pediatrician vaccine-related communication quality, perceived vaccine safety, and vaccination priority among US parents’, Human Vaccines & Immunotherapeutics,Vol. 16, pp. 1024–1029. Gatz, M. and Karel, M. J. (1993). ‘Individual change in perceived control over 20 years’, International Journal of Behavioral Development, Vol. 16, pp. 305–322. Greer, M. L., Sample, S., Jensen, H. K., McBain, S., Lipschitz, R. and Sexton, K. W. (2021). ‘Covid-19 is connected with lower health literacy in rural areas’, Studies in Health Technology and Informatics,Vol. 281, pp. 804–808. Hirsh, J. B., Kang, S. K. and Bodenhausen, G. V. (2012). ‘Personalized persuasion: tailoring persuasive appeals to recipients’ personality traits’, Psychological Science, Vol. 23, pp. 578–581. Hong, S.-M. and Faedda, S. (1996). ‘Refinement of the Hong psychological reactance scale’, Educational and Psychological Measurement, Vol. 56, pp. 173–182. Hornsey, M. J., Harris, E. A. and Fielding, K. S. (2018). ‘The psychological roots of anti-vaccination attitudes: a 24-nation investigation’, Health Psychology, Vol. 37, pp. 307–315. Humer, E., Jesser, A., Plener, P. L., Probst, T. and Pieh, C. (2021). ‘Education level and covid-19 vaccination willingness in adolescents’, European Child & Adolescent Psychiatry, Vol. 32, pp. 537–539. Judge, T. A., Erez, A., Bono, J. E. and Thoresen, C. J. (2002). ‘Are measures of self-esteem, neuroticism, locus of control, and generalized self-efficacy indicators of a common core construct?’, Journal of Personality and Social Psychology, Vol. 83, pp. 693–710. Kapteyn, A., Smith, J. P. and van Soest, A. (2007). ‘Vignettes and self-reports of work disability in the United States and The Netherlands’, American Economic Review, Vol. 97, pp. 461–473. Kerkhofs, M. and Lindeboom, M. (1995). ‘Subjective health measures and state dependent reporting errors’, Health Economics, Vol. 4, pp. 221–235. Kleinbaum, D. G. and Klein, M. (2012a). ‘Parametric survival models’, in Kleinbaum M. and Klein M. (eds), Survival Analysis: A Self-Learning Text, Springer, New York. Kleinbaum, D. G. and Klein, M. (2012b). ‘The stratified cox procedure’, in Kleinbaum M. and Klein M. (eds), Survival Analysis: A Self-Learning Text, Springer, New York. Kroencke, L., Geukes, K., Utesch, T., Kuper, N. and Back, M. (2020). ‘Neuroticism and emotional risk during the covid-19 pandemic’, Journal of Research in Personality, Vol. 89, 104038. Kroszner, R. S. and Strahan, P. E. (1999). ‘What drives deregulation? Economics and politics of the relaxation of bank branching restrictions’, Quarterly Journal of Economics, Vol. 114, pp. 1437–1467. Lau, N. T., Wilkey, E. D., Soltanlou, M., Lagac´ e Cusiac, R., Peters, L., Tremblay, P., Goffin, C., Alves, I. S., Ribner, A. D., Thompson, C., Van Hoof, J., Bahnmueller, J., Alvarez, A., Bellon, E., Coolen, I., Ollivier, F. ©2024 The Author(s). Oxford Bulletin of Economics and Statistics published by Oxford University and John Wiley & Sons Ltd. 584 Bulletin and Ansari, D. (2022). ‘Numeracy and covid-19: examining interrelationships between numeracy, health numeracy and behaviour’, Royal Society Open Science, Vol. 9, 201303. Laurin, K., Kay, A. C. and Fitzsimons, G. J. (2012). ‘Reactance versus rationalization: divergent responses to policies that constrain freedom’, Psychological Science, Vol. 23, pp. 205–209. Lipkus, I. M., Samsa, G. and Rimer, B. K. (2001). ‘General performance on a numeracy scale among highly educated samples’, Medical Decision Making, Vol. 21, pp. 37–44. MAGS. (2021). Neue Corona-Schutzverordnung ab 24. November 2021, Ministry of Labor, Health and Social Affairs of the State of North Rhine-Westphalia (MAGS), D¨ usseldorf. https://www.mags.nrw /pressemitteilung/neue-coronaschutzverordnung-ab-24-november-2021, retrieved on February 28, 2024. Miller, C. H. and Quick, B. L. (2010). ‘Sensation seeking and psychological reactance as health risk predictors for an emerging adult population’, Health Communication, Vol. 25, pp. 266–275. Olagoke, A. A., Olagoke, O. O. and Hughes, A. M. (2021). ‘Intention to vaccinate against the novel 2019 coronavirus disease: the role of health locus of control and religiosity’, Journal of Religion and Health, Vol. 60, pp. 65–80. Pearlin, L. I. and Schooler, C. (1978). ‘The structure of coping’, Journal of Health and Social Behavior,Vol. 19, pp. 2–21. Petersen, T. (1991). ‘Time-aggregation bias in continuous-time hazard-rate models’, Sociological Methodology, Vol. 21, pp. 263–290. Rains, S. A. (2013). ‘The nature of psychological reactance revisited: a meta-analytic review’, Human Communication Research, Vol. 39, pp. 47–73. Reynolds-Tylus, T. (2019). ‘Psychological reactance and persuasive health communication: a review of the literature’, Frontiers in Communication, Vol. 4, p. 56. Riiser, K., Helseth, S., Haraldstad, K., Torbjornse, A. and Richardsen, K. R. (2020). ‘Adolescents’ health literacy, health protective measures, and health-related quality of life during the covid-19 pandemic’, PLoS One, Vol. 18, e0238161. Rotter, J. B. (1966). ‘Generalized expectancies for internal versus external control of reinforcement’, Psychological Monographs: General and Applied, Vol. 80, pp. 1–28. Salali, G. D., Uysal, M. S., Bozyel, G., Akpinar, E. and Aksu, A. (2022). ‘Does social influence affect covid-19 vaccination intention among the unvaccinated?’, Evolutionary Human Sciences, Vol. 4, e32. Schupp, J. and Gerlitz, J.-Y. (2008). Big Five Inventory-SOEP (BFI-S), GESIS - Leibniz Institute for the Social Sciences, Mannheim/Cologne. Shen, L. (2010). ‘Mitigating psychological reactance: the role of message-induced empathy in persuasion’, Human Communication Research, Vol. 36, pp. 397–422. Shen, L. and Dillard, J. P. (2005). ‘Psychometric properties of the Hong psychological reactance scale’, Journal of Personality Assessment, Vol. 85, pp. 74–81. Solon, G., Haider, S. J. and Wooldridge, J. M. (2015). ‘What are we weighting for?’, Journal of Human Resources, Vol. 50, pp. 301–316. Soveri, A., Karlsson, L. C., M¨ aki, O., Antfolk, J., Waris, O., Karlsson, H., Karlsson, L., Lindfelt, M. and Lewandowsky, S. (2020). ‘Trait reactance and trust in doctors as predictors of vaccination behavior, vaccine attitudes, and use of complementary and alternative medicine in parents of young children’, PLoS One, Vol. 15, e0236527. Soveri, A., Karlsson, L. C., M¨ aki, K. O., Holford, D., Fasce, A., Schmid, P., Antfolk, J., Karlsson, L., Karlsson, H., Nolvi, S., Karukivi, M., Lindfelt, M. and Lewandowsky, S. (2023). ‘Trait reactance as psychological motivation to reject vaccination: two longitudinal studies and one experimental study’, Applied Psychology: Health and Well-Being, Vol. 16, pp. 597–614. Wang, W., Krishna, A. and McFerran, B. (2017). ‘Turning off the lights: consumers’ environmental efforts depend on visible efforts of firms’, Journal of Marketing Research, Vol. 54, pp. 478–494. Widiger, T. A. (2009). ‘Neuroticism’, in Leary M. R. and Hoyle R. H. (eds), Handbook of Individual Differences in Social Behavior, Guilford Press, New York, New York. Widiger, T. A. and Oltmanns, J. R. (2017). ‘Neuroticism is a fundamental domain of personality with enormous public health implications’, World Psychiatry, Vol. 16, pp. 144–145. ©2024 The Author(s). Oxford Bulletin of Economics and Statistics published by Oxford University and John Wiley & Sons Ltd. Predictors of individual behaviour in pandemics 585 Wolter, F., Mayerl, J., Andersen, H. K., Wieland, T. and Junkermann, J. (2022). ‘Overestimation of covid-19 vaccination coverage in population surveys due to social desirability bias: results of an experimental methods study in Germany’, Socius, Vol. 8, 23780231221094749. Wright, D. and Rune, K. T. (2023). ‘Underlying motivators for anti-vaccination attitudes amongst regional Sunshine Coast parents in Australia’, Health Promotion Journal of Australia, Vol. 34, pp. 579–586. Yost, A. B., Behrend, T. S., Howardson, G., Badger Darrow, J. and Jensen, J. M. (2019). ‘Reactance to electronic surveillance: a test of antecedents and outcomes’, Journal of Business and Psychology, Vol. 34, pp. 71–86. Supporting Information Additional Supporting Information may be found in the online Appendix: Appendix S1. Supplementary Information. ©2024 The Author(s). Oxford Bulletin of Economics and Statistics published by Oxford University and John Wiley & Sons Ltd.