Public acceptability of containment measures during the COVID-19 pandemic in Italy: how institutional confidence and specific political support matter

PurposeThis article contributes to a better theoretical and empiric understanding of mixed results in the literature investigating the relationship between institutional confidence and adherence to recommended measures during a pandemic.Design/methodology/approachThe article relies on structural equation models (SEMs) based on data from ResPOnsE COVID-19, a rolling cross-section (RCS) survey carried out in Italy from April to June 2020.FindingsThe authors’ findings show the existence of multiple pathways of confidence at the national and local level. Confidence in the institutions is positively associated with support for the performance of the Prime Minister and that of the regional institutions in the North West, which in turn, raises the likelihood of following the restrictive measures. However, in the same regions, a good appraisal of the regional system's performance also had a direct positive effect on the perception of being safe from the virus, decreasing adherence to the restrictive measures. Finally, the direct effect of confidence in the institutions on compliance is negative.Social implicationsThe result enlightens the crucial role both of national and local institutions in promoting or inhibiting adherence to restrictive measures during a pandemic and suggests that “one size fits all” measures for increasing overall institutional confidence might not be sufficient to reach the desired goal of achieving compliance in pandemic times.Originality/valueThe authors theorize and test three cognitive mechanisms – (1) the “cascade of confidence”; (2) the “paradox of support” and (3) the “paradox of confidence” – to account for both the positive and negative links between measures of political support and public acceptability of COVID-19 containment measures.

the national and local level, higher confidence in the institutions is negatively associated with compliance with the preventive measures and weakly associated with willingness to restrict personal freedoms to prevent the spreading of the virus (in line with the "paradox of confidence").
These results are relevant from a theoretical and methodological point of view because they help push the field forward by offering a novel theoretical framework to explain mixed findings in existing literature. Furthermore, from a policy point of view they enlighten the crucial role both of national and local institutions in promoting or inhibiting adherence to restrictive measures during a pandemic and suggest that "one size fits all" measures for increasing overall institutional confidence might not be sufficient to reach the desired goal of achieving compliance in pandemic times. Indeed, our results show the existence of multiple, interrelated pathways of confidence that occur at the national and local level and that need to be taken into account for a full understanding of the acceptance of COVID-19 restrictive measures.
The article is structured as follows. In the next section we present the theoretical backbone of the research; we then provide a description of the unfolding of the pandemic in Italy, our case study, and present our hypotheses. Section 3 discusses the data and methods used in the article and in section 4 we present our results. Finally, section 5 outlines some tentative policy implications, and provides a conclusion.

The empirical and theoretical framework
Scholars referring to institutional theory (Baumol and Blinder, 2008) have pointed out the crucial role of the asymmetry between formal and informal institutions (norms, values and codes of conduct) in explaining the varying degrees of citizens' compliance with political measures or duties. The main argument is that it is not just a expected pay-off or coercion that shapes citizens' attitudes and behaviours, but also institutional confidence. For example, one of the main research topics in the tax morale literature is the influence of formal institutions (see Horodnic, 2018 for a recent and systematic review), but empirical evidence also comes from other fields in political science, such as healthcare, security and electoral behaviour (Peters, 2019). Not surprisingly, this relationship is also found in studies investigating what determines the public acceptability of non-medical measures to limit the spread of pandemics. Several interesting findings have emerged from studies on public reactions during epidemics of infectious diseases, such as Ebola (Vinck et al., 2019), SARS (Tang and Wong, 2003), H1N1 (Blendon et al., 2006;Quinn et al., 2009;van der Weerd et al. 2011;Prati et al., 2011). Most of these studies point out that confidence in political and health authorities plays a decisive role in successfully implementing containment measures. As far as COVID-19 is concerned, empirical evidence of this relationship is found both at the aggregate and individual level.
At the aggregate level, recent research shows that countries with higher institutional confidence report lower fatalities (Oksanen et al., 2020) or more social distancing behaviour (Barrios et al., 2020;Bargain and Aminjonov, 2020). Analysis of an original dataset revealed that confidence in the local government contributed to decreasing the COVID-19 infection rate in China, while confidence in the central government did not. Furthermore, the effect of confidence in local governments is mediated by risk perception (Maoxin Ye and Zeyu Lyu, 2020). However, due to the risk of "ecological fallacy", results concerning individual behaviour derived from aggregate data should be interpreted with caution.
At the individual level, despite the considerable proliferation of surveys aimed at studying individual COVID-19-related attitudes and behaviours, there are still few studies based on large samples and reliable data that have specifically focused on the relationship between confidence in political institutions and the willingness to adopt recommended preventive measures. The data at the individual level available at the time of writing (July 2020) provide some first interesting, albeit mixed, insights. For example, authors working on 1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  60   t  e  r  n  a  t data from the PsyCorona project (https://psycorona.org/), a web-based survey that included 23,733 participants from 23 countries, found a robust relationship between confidence in the national government and COVID-19 personal preventive behaviour (Han et al., 2020). By contrast, authors working on data from cross-sectional and panel surveys implemented in eight Western democracies found that confidence in the national government and positive evaluations of government responses were weakly and inconsistently correlated with compliance with protective advice (Jørgensen et al., 2020). A mixed picture is also revealed by two crosssectional studies implemented during the early phase of the COVID-19 pandemic in Germany (Dohle et al., 2020). Similar findings are drawn from two cross-sectional studies conducted after the lockdown and before the peak of the COVID-19 epidemic in France, when confidence in political institutions had no influence on compliance with behavioural recommendations (Raude et al., 2020).
Empirically, it seems that the sign of the relationship between confidence in institutions and adherence to preventive measures against pandemics varies based on contextual/contingent factors and methodological issues.
From a theoretical standpoint, although there is no systematic review of the literature on the subject, three theoretical streams can be identified which share the common argument that political confidence acts as a mechanism to reduce the complexity and uncertainty (Bish and Michie, 2010;Siegrist and Zingg, 2014;Luhmann, 1989) recurrent in times of (pandemic) crisis. The mechanism "reflects the positive evaluation by the subjects of specific objects in a situation of uncertainty or vulnerability" (van der Meer, 2017, p. 6). Beyond this common umbrella, each research stream stresses different aspects which we build on to theorize three mechanisms explaining the role of political support -both diffuse and specific (Easton, 1965) -in public adherence to COVID-19 restrictive measures. Specifically, these are the "cascade of confidence", the "paradox of support" and the "paradox of confidence".
The first mechanism, defined as the "cascade of confidence", explains the positive association between institutional confidence and adherence to COVID-19 restrictive measures through specific support for the government. It is inspired by to the so-called "rally 'round the flag" effect. This concept was developed in the 1970s (Mueller, 1973) to describe the growing support for national leaders in times of international crises due to the awakening of a sense of patriotism in response to the perception that the nation was under threat. Applications of this concept to the COVID-19 pandemic are growing and there is evidence that national incumbent and democratic institutions benefitted from a sort of this effect, especially during the first phase of the outbreak (Bol et al., 2020;Sibley et al., 2020;Baekgaard et al.;Yam et al., 2020). However, this mechanism only tells part of the story. The level of support for the government and other authorities was not invariant during the COVID-19 outbreak and the trends were highly context-dependent (Esaiasson et al., 2020;Bol et al., 2020;Sibley et al., 2020). 5 To explain these differences, some scholars have highlighted the key role of partisanship. The trade-off between a large number of deaths and the economic and social costs of lockdown nourished a heated political debate involving several salient political issues (such as civil liberties, economic insecurity, the right to health care…). On the one hand, there is evidence that elite and public polarization on salient political issues undermines compliance with social distancing and other preventive measures recommended by the national government (Barbieri et al., 2020;Cornelson and Miloucheva, 2020;Gadarian et al., 2020;Grossman et al., 2020). On the other hand, in some countries, the COVID-19 outbreak represented a "rare moment of cross-partisan consensus" (Merkley et al., 2020). In any case, the role of partisanship in influencing the relationship between institutional confidence, government support and adherence to COVID-19 preventive measures cannot be ignored. Moving to the second mechanism, the "paradox of support", scholars have theorized that a good (perceived) performance of authorities in the management of the pandemic decreases the perception of risk and, consequently, decreases compliance. This kind of speculation is frequent in the literature on risk management which proposed the "Trust, Confidence and Cooperation (TCC)" model (Early and Siegrist, 2008;Siegrist and Zingg, 2014). Studies testing the hypothesis that confidence in political authorities influences the adherence to COVID-19 containment measures indirectly negatively, by shaping the public's judgments about the risks, are consistent with this approach (e.g., Mohammadi et al., 2020). Drawing on qualitative studies some scholars have even suggested the existence of a "paradox" during a pandemic: "public trust based on a perception of government competence, care and openness may in fact lead people to underestimate risks and thus reduce their belief in the need to take individual action to control the risks" (Wong, 2020). Indeed, this negative effect had also been found in quantitative studies carried out in previous pandemics. For example, van der Weerd and colleagues (2011) showed that at the start of the A/H1N1 influenza pandemic in the Netherlands, higher levels of institutional confidence were associated with lower intentions to adopt protective measures. Similarly, a study in Taiwan (Fong and Chang, 2011) showed that institutional confidence was positively related to the acceptance of recommended actions but only in cities with no SARS outbreak.
Finally, the third mechanism, the "paradox of confidence", is based on the psychological research that identifies COVID-19 as a stressful event (Brooks et al., 2020;Zettler et al., 2020) and argues that institutional confidence acts as a reducer of pandemic-related stress: the more confidence citizens have in the institutions, the less they perceive the need to take individual actions to avoid the contagion and limit the spread of the virus. Conversely, citizens distrusting the institutions may become even more anxious about the spread of the virus and decide to adopt all possible protective health behaviours. According to the literature, this behaviour occurs because the subjective perception of risk is one of the main predictors of adherence to authorities' recommendations, more than the objective severity of the risk (Khosravi, 2020).
The article contributes theoretically and empirically to the field by proposing a conceptualization and empirical testing of the three theorized mechanisms, while focusing on the Italian case, as illustrated below.

The Italian case
During the pandemic, Italy adopted very strict COVID-19 preventive measures. Starting on 23 February, the government restricted the movement of the people living in some northern towns where the virus was spreading fast, and by 9 March the entire country had entered full lockdown. This was the beginning of the socalled "Phase 1". Schools, commercial activities (except for essential services like supermarkets) and most economic activities were closed. Any form of gathering of people in public places was forbidden. "Phase 1" lasted for two months and, on 4 May, the government started to relax the restrictions. In "Phase 2" many people went back to work, bars and restaurants re-opened, and people were allowed to meet relatives living in the same region. "Phase 3" started on 3 June, when people were allowed to travel throughout the country and abroad. With the significant exception of schools, almost all activities have re-opened and citizens are requested to follow the rules laid down at the beginning of the pandemic: one-metre distancing, wearing protective masks in closed spaces and sanitizing hands. At the time of writing (end of July), Italy has counted a total of 245,000 positive cases, 35,000 deaths as well as 12,000 individuals that are still positive to the virus.
Italy represents a highly interesting case for the empirical investigation of the three theorized mechanisms. Firstly, it was the first European country to be severely hit by the pandemic. Consequently, it was the first to implement strict restrictive measures. At the same time, Italy is usually identified as one of the European countries with the lowest levels of institutional confidence (Torcal, 2017). Nonetheless, analysis from mobile phones revealed rather high compliance with the mobility restriction measures to contain the pandemic (Pepe et al., 2020). Secondly, the spread of the Coronavirus across the different Italian regions was extremely heterogeneous: almost 40% of the positive cases were concentrated in Lombardy, another 33% of cases were found in the other three northern regions hardest hit by the pandemic and the remaining cases were spread in the other sixteen Italian regions (data source: Italian Government, Ministry of Health). The localized epidemic combined with the decentralization of the health system 6 gave visibility to the key role of local authorities, alongside national ones, in implementing timely interventions and effectiveness against the Coronavirus (Armocida et al., 2020). Therefore, these are the ideal conditions to test the "paradox of support" not just at the national but also at the local level, i.e. the social and institutional perimeter in which Italian citizens mostly experienced both the daily exposure to the risk of contagion and the effectiveness of the containment measures. Moreover, the hard-hit regions, such as Lombardy and Piedmont, are governed by opposition parties, a situation which nourished both central-local tensions and a politicization of the outbreak (Barbieri et al., 2020).

Hypotheses
Following the theoretical arguments above, we developed the following hypotheses. H1: The "cascade of confidence" hypothesis: institutional confidence is positively associated with specific support for the Prime Minister and for the regional system, which in turn promote both compliance with the COVID-19 preventive measures and the willingness to restrict personal freedoms. H2: The "paradox of (local) support" hypothesis: only in those regions worst hit by the COVID-19 pandemic does specific support for the regional system increase the perception of COVID-19 safety, which in turn decreases both compliance with the COVID-19 preventive measures and the willingness to restrict personal freedoms. H3: The "paradox of institutional confidence" hypothesis: institutional confidence exhibits a negative direct effect on compliance with the COVID-19 preventive measures in all Italian regions.

Data
The present article relies on data from ResPOnsE Covid-19, a Rolling Cross-Section (RCS) survey carried out in Italy to monitor Italian public opinion during the Coronavirus crisis (Vezzoni et al., 2020). The online data collection was based on an opt-in panel survey carried out by a private survey agency (SWG), stratified by macro-area of residence and with quotas for gender and age class. Overall, this article relies on the sample collected from 6 April to 29 June 2020, corresponding to a gross sample size of 13,850 individuals.

Structural Equation Model
To test our hypotheses, we specified a Structural Equation Model (SEM) consisting of: 1) the measurement model, which includes five latent variables (willingness to restrict personal freedoms, compliance with the COVID-19 preventive measures, institutional confidence, specific support for the regional system and COVID-19 safety perception); 2) the structural/causal model: based on theoretical assumptions, it links the five latent variables seen above and one observed variable (specific support for the Prime Minister) 7 . The items used to measure the five latent variables are presented in Table 1 and were selected based on the theoretical and empirical framework discussed above. The measurement model aims to check the extent to which our conceptual refinement and operationalization of the constructs is consistent with the survey data collected. The causal model seeks to unravel which kind of political support (institutional confidence vs. appraisal of the performance of Prime Minister and local institutions), together with the COVID-19 risk perception, promoted 1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  60   t  e  r  n  a  t or inhibited compliance with the COVID-19 preventive measures and the public acceptability of restrictions on freedoms. Specifically, the model assumes that the willingness to restrict personal freedoms and compliance with the COVID-19 preventive measures, which are both at the end of the causal chain, are two different dimensions of a more general individual adherence to the authorities' COVID-19 recommendations. Their association is estimated as covariance/correlation given that directional effects cannot be established between them. The remaining latent and observed variables are related to them in a meaningful way, starting from institutional confidence. 8 The structural paths followed by the "cascade of confidence" mechanisms at the national and local level are specified as: institutional confidence directly and indirectly influences the respondents' willingness to restrict their personal freedoms and compliance with the COVID-19 preventive measures, through the mediation of i) specific support for the Prime Minister and ii) specific support for the regional system. An additional structural path is specified to reveal the "paradox of (local) support" mechanism: the model assumes that the willingness to restrict personal freedoms and compliance with the COVID-19 restrictive measures are influenced directly by specific support for the regional system/specific support for the Prime Minister during the pandemic, and indirectly, by the mediation of the perception of being safe from the virus. Finally, the "paradox of confidence" mechanism is revealed by the direct effect of institutional confidence on compliance with the COVID-19 preventive measures.
To assess the ability of the Full SEM to reproduce the data, we referred to the CFI (Comparative Fit Index), RMSEA (Root Mean Square Error of Approximation) and SRMR (Standardized Root Mean Square Residual). The cut-off values are: (a) CFI: if the value is equal to 0.95 the model can be accepted, values above 0.90 are satisfactory; (b) RMSEA: values between 0 and 0.05 are considered indicators of a good fit; c) SRMR, a value of less than 0.08 is generally considered a good fit.

Mechanisms connecting institutional confidence to adherence to COVID-19 restrictive measures: a Structural Equation Model
To test the extent to which the three mechanisms were at work during the COVID-19 outbreak in Italy, we followed a two-step analytical strategy. First, we tested the goodness of the theorized SEM. The Full SEM on the pooled dataset exhibits a good fit (Chi Square=1229.048; DF=51; CFI=0.959; RMSEA: 0.041; SRMR: 0.021). We then fit the model for separate groups (North-West regions 9 vs. other regions) to test our hypotheses. Based on modification indices, we adjusted the model to avoid biased estimates of the relations between institutional confidence and specific support due to partisanship. Thus, the final model includes the exogenous "Political leaning" variable. The model proved to be a satisfactory fit in both groups. 10 The sign and intensity of the structural coefficients confirm how the three mechanisms worked (Figure 1). 8 Due to space constraints, the figure of the theorized model is not presented here. However, empirical testing of the model with an estimation of each structural path is depicted in Figure 1 1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  60   t  e  r  n  a  t Figure 1 here] Specifically, as far as the "cascade of confidence" (H1) is concerned, our results show that in the North-West regions, institutional confidence is positively associated with specific support for the Prime Minister (the standardized path coefficient is significant and equal to 0.588), which in turn promotes compliance with the COVID-19 preventive measures (β=0.24) and willingness to restrict personal freedoms (β=0.24). A similar pattern is found regarding the mediating role of specific support for the regional system, even though the intensity of the effects in each path of interest is smaller (respectively, β=0.27, β=0.15, β=0.17). Thus, these results indicate that in the areas worst hit by the pandemic confidence in both national and local institutions plays a role in predicting compliance with the preventive measures. In other regions, institutional confidence is also positively associated with specific support for the Prime Minister (the standardized path coefficient is significant and equal to 0.60), which in turn promotes compliance with the COVID-19 preventive measures (β=0.26) and willingness to restrict personal freedoms (β=0.28). In contrast, the mediating role of specific support for the regional system is not found here: its effect on compliance and willingness is negligible (β=0.09 in both paths), suggesting the limited salience of local institutions in regions not strongly hit by the pandemic.
Moving to the hypothesis of the "paradox of (local) support" (H2), the analysis revealed that in the North-West regions the effect of specific support for the regional system on COVID-19 safety perception is positive (β =0.17), while the latter reduces both willingness to restrict personal freedom (β=−0.21) and compliance with COVID-19 preventive measures (β=−0.21) 11 ; in other regions, as expected, specific support for the regional system has a negative (and very small) effect on COVID-19 safety perception; the effect of safety perception on willingness to restrict personal freedom is negative (β=−0.14) and on compliance negligible (β=−0.02).
Finally, the analysis confirmed the "paradox of institutional confidence" hypothesis (H3): the direct effect of institutional confidence on compliance with the COVID-19 preventive measures is negative in both groups (β=−0.16 in North-West regions and β=−0.13 in the other regions).

A further focus on the "cascade of confidence": seemingly unrelated regression models
The SEM showed that the Prime Minister played a key role in promoting compliance with the recommended preventive measures and boosting the willingness to restrict personal freedoms in Italy during the pandemic. Moreover, the analysis revealed that the direct effect of confidence in the institutions (such as the national parliament and European Union) on the public acceptability of these measures was negligible (or even negative as far compliance is concerned). Therefore, satisfaction with the government's handling of the pandemic seems to be one of the main circumstances under which the public is willing to sacrifice its own freedom for a "greater good", such as the limitation of the diffusion of viruses like SARS-CoV-2. These results are consistent with the empirical studies performed in other countries which have been presented above. Given the relevance from a public policy perspective, it is legitimate to ask whether this effect remains positive and statistically significant even when additional variables not present in the SEM model are taken into consideration. To this end, we ran a set of seemingly unrelated linear regression models whose dependent variables are compliance with the COVID-19 preventive measures and willingness to restrict personal freedom. Beyond the variables included in the SEM presented above, as controls these models also include: age (18-99), gender (women as the reference category), education (primary, secondary and tertiary as the reference), period of data collection (phases of the COVID-19 outbreak, from 6 April to 3 May as the reference, from 4 May to 14 June, and from 15 to 29 June), 11 It is worth noting that the effect of specific support for the Prime Minister on COVID-19 safety perception is negative, consistently with the Prime Minister's key role in reinforcing the "cascade of confidence". generalized trust, confidence in the police, confidence in science, institutional source of information during the pandemic (yes vs no), opinion on the appropriateness of following authorities' recommendations during a crisis and living in a region governed by opposition parties. The main results are reported in Table 2, while the full models are provided in Table A2 in the Online Appendix.
[ Table 2 here] In the first model, which controls for the socio-demographic variables and the period of data collection, the association between institutional confidence and attitudes towards restrictions on individual freedoms is large and positive, ceteris paribus (β=0.25, p≤0.000). After the inclusion in sequence of each group of predictors, the intensity of the association gradually decreases, but remains positive and quite large. As expected by H1, when the appraisal of Conte's performance is included, the association between institutional confidence and the respondents' willingness to restrict their freedoms dramatically decreases (β=0.03, p≤0.000). In other words, the direct contribution of institutional confidence on the public's willingness to implement the COVID-19 restriction measures is negligible, even if positive and significant, once we control for the appraisal of Conte's performance. In contrast, institutional confidence only has a weak association with compliance, which becomes negative once all the controls are added to the models.
Crucially, the regression revealed that satisfaction with Conte's performance is one of the most important determinants of the public acceptability of the COVID-19 containment measures.

Conclusion
This article aimed to contribute theoretically and empirically to a better understanding of the puzzling results in the empirical literature investigating the relationship between institutional confidence and adherence with recommended measures during a pandemic.
From the theoretical and methodological point of view, we refined and systematized the construct concepts and measurements. We distinguished between institutional confidence and specific support and considered that institutions operate at different levels, nationally and locally. We theorized three cognitive mechanisms to account for both the positive link between measures of political support and compliance as well as counterintuitive and less explored negative associations. This analytical strategy allowed us to draw some empirically testable hypotheses. The COVID-19 outbreak in Italy, where the epidemic was localized and where local authorities are responsible for the organization and delivery of health services, provided an ideal setting in which to empirically validate our conceptualization.
The analysis provides interesting insights. To be sure, institutional confidence is related to individual experiences of the pandemic. However, the results invite caution in making simple claims such as "more confidence in institutions, more compliance" part of a strategy to flatten the epidemic curve. First, as far as the positive association is concerned, the results suggest an interesting "cascade of confidence" that goes from institutional confidence to the acceptability of preventive/restrictive measures. Powered by the popularity of the Prime Minister, step by step, this mechanism reinforces the pact between Italian citizens and the national institutions in times of crisis. The analysis suggested that confidence in institutions represented a "reservoir of favourable attitudes and good will" (Easton, 1965, p. 273) even during the COVID-19 outbreak. This reserve had little direct effect on adherence to the COVID-19 restrictive measures, but it largely fed support for the Prime Minister (and to a lesser extent for the regional institutions) in a hard time such as the pandemic. In the same way as other prime ministers and presidents in Europe, Giuseppe Conte enjoyed considerable popularity during the outbreak (Segatti, 2020). The data presented here show how this popularity was strongly rooted in a more general institutional confidence. Therefore, it was not only a "charismatic" and contingent popularity, but probably a sort of "offshoot" of diffuse support, which is more stable over time. However, the data showed  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  60   t  e  r  n  a  t that the direct effect of institutional confidence on citizens' willingness to restrict their freedom in order to limit the spread of the virus was negligible. The risk is that, once the "fuel" of the "rally 'round the flag" effect/Prime Minister's popularity runs out, the "cascade of confidence" mechanism gets stuck. It is a risk that could be heightened by the fact that institutional diffidence behaves as an amplifier of pandemic-related stress (as "the paradox of confidence" argues). Indeed, citizens distrusting institutions might become even more anxious about the spread of the virus and decide to adopt all possible protective health behaviours. Viewed from this perspective, the higher compliance with restrictions on mobility registered during the COVID-19 outbreak in Italy (Pepe et al., 2020), one of the European countries with lower institutional confidence, seems less puzzling.
Overall, while, through a combination of direct and indirect effects, widespread institutional and specific support fostered both compliance and the willingness to restrict individual freedoms during the pandemic, neither kind of political support at the national level had any relevant effect on the risk perception of COVID-19. In contrast, our data showed that in Italy a good appraisal of the performance of the regional system during the pandemic had a direct and positive effect on the perception of being safe from the virus, but only in the regions most affected by the pandemic. In turn, the perception of COVID-19 safety decreased adherence to the restrictive measures. We have labelled this mechanism "the paradox of support" to point out that good local management of the pandemic might satisfy citizens, but at the same time also reduce their willingness to follow the COVID-19 recommendations and accept the restrictions on freedoms. This mechanism might help to explain why the level of adoption of the COVID-19 preventive measures and willingness to restrict personal freedoms to limit the spread of the virus decreased over time.
As far as policy implications are concerned, they are mainly related to risk communication and community engagement. Specifically, the results give empirical support to some general recommendations of the Strategic Preparedness and Response Plan launched by the World Health Organization (WHO, 14 April 2020).
Regarding risk communication, the "cascade of confidence" mechanism offers a reason why "it is essential that international, national, and local authorities engage through participatory two-way communication efforts proactively, regularly, transparently and unambiguously with all affected and at-risk populations" (WHO, 2020, p.7). This communication strategy is expected to contrast the negative effect of infodemics (Zarocostas, 2020). The results in this article suggest that this policy can also boost compliance with COVID-19 containment measures by reinforcing trust across individuals, communities and institutions at local and national level. The main point to stress here is that, at least for the Italian case, greater satisfaction with the government's performance (at both the national and regional level) is directly associated with higher compliance, while institutional confidence plays a key role in the legitimization of the actors involved in controlling the pandemic. This suggests that, on the one hand, continuous, clear and reliable communication about the effectiveness of the implemented non-medical measures is essential to keep the crucial "cascade of confidence" mechanism alive. On the other hand, the messages from different institutions (local, national and international authorities and healthcare actors) should be consistent with each other, to prevent conflicting narratives and institutional distrust.
As regards community engagement, the Italian case sheds light on the crucial role of local communities in slowing down the transmission of the virus. The strategic plan launched by the WHO pointed out that "communities must be empowered to ensure that services and aid are planned and adapted based on their feedback and local contexts" in order to obtain "the support of every part of affected communities" (WHO, 2020, p. 6). Based on the data presented above, this recommendation is of major importance, precisely because the public risk perception of COVID-19 is influenced by the local management of the pandemic in affected regions. Furthermore, the findings allow us to make some considerations concerning the feared "second wave". In the face of the rise in infections since early August in Italy, Professor Galli (head of infectious diseases at the Sacco hospital in Milan) said: "The end of confinement has resulted in an excessive feeling of false 10 security" (quoted by La Repubblica newspaper). 12 This idea of "false security" is consistent with the expectations and empirical evidence discussed above: the lower the perceived risk of COVID-19 infection, the lower the public acceptability of non-medical containment measures. Therefore, to prevent the "second wave", it is essential that local stakeholders participate in defining a communication strategy that is able to explain why the implementation of targeted and time-limited containment measures is needed, even when the scenario seems to be sustainable and the transmission of COVID-19 under control.
To conclude, cooperation with physical distancing measures and other containment measures is the result of complex interactions between individuals, communities and political authorities. This article has offered empirical insights on a specific aspect, through its focus on the role of institutional confidence and political support. Further research is needed to investigate the role of horizontal trust and the gap between formal and informal institutions in explaining different levels of compliance (Horodnic, 2018). Moreover, it would be appropriate to validate the three mechanisms with different datasets, beyond the Italian case, especially concerning those countries where the introduction of sanitary measures was followed by competition between the local and central governments.