The Roles of Police-Related versus Non-Police-Related Considerations in Shaping Diffused Support for the Police: Lessons from the COVID-19 Pandemic

Abstract Global views of the police (often termed “legitimacy”, “trust”, “satisfaction” or “support”) are frequently regarded as the product of micro-level, police-related consideration: beliefs about what the police are doing and how they are doing it. Studies taking a sociological or political perspective have revealed that views unrelated to policing, such as satisfaction with the government, may also be important predictors of global views of the police. However, police-related and non-police-related considerations are frequently not considered together as antecedents of macro-level views of the police. This study aims to illuminate the respective roles of police-related versus non-police-related considerations in shaping citizens’ diffused support for the police while utilizing the context of the COVID-19 pandemic. Data from a panel survey carried out during the first and third peaks of the pandemic in Israel reveal that the only factor which had a direct effect on the drop in diffused support for the police was the public’s assessment of the government’s (not the police’s) performance in handling the pandemic. We conclude by calling for more sophisticated measurement and interpretation of public approval of the police and its antecedents.


INTRODUCTION
Public attitudes toward the police have been measured as early as the era of August Vollmer when Bellman (1935) developed the "Police Service rating scale".Since then, and particularly since the social unrest of the 1960s in the USA, thousands of studies have assessed how citizens in different countries, communities, situations and policing contexts view their police (for reviews, see Brown and Benedict 2002;Cao and Wu 2019;Decker 1981;Mazerolle et al. 2013).The importance of these views lies in their socially desirable outcomes.Broad, favourable attitudes toward the police (often termed by researchers "legitimacy", "trust", "satisfaction" or "support") are expected to lead to a positive relationship between the police and the public, one that is characterized by cooperation, compliance, approval of police empowerment and, ultimately, compliance with the law (e.g.Mazerolle et al. 2013;Sunshine and Tyler 2003;Tyler 2004Tyler , 2009)).Negative sentiments, on the other hand, may have detrimental implications.These may include violent protests and rioting, and, recently witnessed in the USA, even calls to abolish the police (Lum et al. 2021;President's Task Force on 21st Century Policing 2015;Weisburd and Majmundar 2018).
A core assumption underlying much of the work on public attitudes toward the police, and particularly the work carried out within the social psychology framework of "police legitimacy" (for reviews, see Nagin and Telep 2017;Tyler and Nobo 2023), is that police conduct (as subjectively perceived) is the primary factor in determining why citizens evaluate their police the way they do (e.g.Cao, Frank, and Cullen 1996;Reisig and Parks 2000;Rosenbaum et al. 2005;Skogan 2009;Tyler 2004Tyler , 2009;;Weitzer and Tuch 2006).Studies taking a more sociological or political viewpoint have recognized that factors unrelated to police conduct, such as a sense of inclusion in the political system (e.g.Albrecht and Green 1977;Stack and Cao 1998), may have important effects on the public image of the police.However, the interplay between police-related and non-police-related considerations as antecedents of broad views of the police is rarely examined.In other words, it is uncommon to find studies that predict macro-level views of the police (general assessments of one's approval of the police) using both micro-level assessments of police conduct and other contextual characteristics beyond police influence (for some exceptions, see Karakus 2017;Wu, Poteyeva, and Sun 2012).
The implications are that the respective roles of police-related and non-policerelated considerations in the formation of global, macro-level views of the police remain unclear.Is it the case that the police are, by and large, responsible for their public image?Alternatively, do other factors, generally beyond police control, also play an important role?If the latter is correct, how should we interpret public attitudes toward the police?What do they reflect, and what conclusions can policymakers, police practitioners and the public more generally draw from them?In the present study, we seek to provide insight into these questions.We take advantage of the context of the COVID-19 pandemic and examine the effects of considerations associated with the pandemic, both related and unrelated to policing, on the public's diffused support for the Israel Police (IP) (as a macro-level assessment) during the pandemic period.
We begin by reviewing different bodies of work on public attitudes toward the police and demonstrate that while much of the research carried out within the social psychology framework presupposes that global, macro-level views of the police are determined primarily by perceived police conduct, there are good reasons to suspect that influences unrelated to policing may also have important effects.Importantly, the two types of considerations are typically not examined together in a single statistical model, and, thus, their reciprocal effects on macro-level views of the police still need to be clarified.Questions about the relative effects of police-related views versus considerations unrelated to policing can best be answered within a bounded set of circumstances where both types of factors are well defined.We present the COVID-19 pandemic as a useful setting for such an assessment and develop hypotheses about the factors that may have an impact on public support for the police during the pandemic based on the more general literature on policing emergencies.We then describe our panel survey, sample and analysis, which reveals that in our data, the only factor with a direct, statistically significant effect on the drop in diffused support for the IP over nine months during the pandemic was evaluations of how the Israeli government handled the pandemic.Surprisingly, assessments of pandemic policing and its outcomes had no direct effect on the drop in support.We conclude that there should be better recognition of the fact that sometimes, global views of the police may have little to do with perceived police conduct and argue for the importance of more complex analyses and interpretations of public attitudes toward the police.An important assumption underlying much of the work on public attitudes toward the police, and particularly the research taking the social psychology perspective (for a review, see Tyler and Nobo 2023), is that individuals develop their global, macrolevel views of the police (which are general assessments of approval or "liking" of the police, such as "trust", "legitimacy", "satisfaction" or "support") primarily based on micro-level assessments of what the police are (or are not) doing.By "micro-level" assessments, we are referring to concrete evaluations of specific aspects of policing, such as effectiveness in fighting crime and fair treatment of citizens.This bottom-up assumption is emphasized, for example, in the seminal work by Tom Tyler (2004Tyler ( , 2009) ) and many others (e.g.Bradford, Murphy, and Jackson 2014;Gau et al. 2012;Hinds and Murphy 2007;Reisig and Lloyd 2009) on police legitimacy."Legitimacy" has been defined as "the belief that the police are entitled to call upon the public to follow the law and help combat crime, and that members of the public have an obligation to engage in cooperative behaviors" (Tyler 2004, 86-7).It is frequently operationalized as "trust in the police", sometimes in combination with an obligation to obey police orders (e.g.Sunshine and Tyler 2003). 1 The model underlying this immense body of work postulates that people judge the legitimacy of their police based on both the outcomes the police deliver (the what) and on the fairness of the processes by which they exercise their authority (the how).Of the two, procedural fairness is typically more closely linked to legitimacy (for reviews, see Jackson et al. 2015;Mazerolle et al. 2013;Tyler, Goff, and MacCoun 2015;Weisburd and Majmundar 2018).The assumption at the model's core is that people first interpret police conduct in terms of both the what and the how.In turn, these concrete assessments then lead to global, macro-level views of the police, including "trust" and "legitimacy" (see Nagin and Telep 2017).

LITERATURE REVIEW
A similar assumption is reflected in a parallel body of work on public satisfaction with the police.For example, as a macro-level assessment, "satisfaction" presumably cultivates in interactions with police officers (both direct and vicarious), in which 1 See Gau (2011), Tankebe (2013), Tyler and Jackson (2013) and McLean and Nix (2021) for different views on the definition and measurement of "police legitimacy".

International Annals of Criminology
individuals develop micro-level views about what the police are doing and how they are doing it (e.g.Bradford, Jackson, and Stanko 2009;Dai, Hu, and Time 2019;Kochel 2012;Skogan 2006;Weitzer and Tuch 2002).The neighbourhood in which one resides was also found to affect satisfaction with the police, but similarly, neighbourhood effects are generally attributed to factors associated with police conduct in the neighbourhood, such as local levels of crime and social or physical disorder (e.g.Cao et al. 1996;Sampson and Bartusch 1998;Skogan 2009).Another important factor that was found to influence satisfaction with the police is concentrated disadvantage, which is typically made up of community-level characteristics such as income, racial composition, education and employment (e.g.Apple and O'Brien 1983;Dunham and Alpert 1988;Factor, Castilo, and Rattner 2014;Reisig and Parks 2000;Sampson and Bartusch 1998;Sampson, Raudenbush, and Earls 1997;Weitzer 1999).Such attributes in and of themselves are not within police control.However, scholars have again argued that their effects on public sentiments toward the police are mediated through police conduct because the police are more likely to use force or engage in misconduct in disadvantaged neighbourhoods (e.g.Cao et al. 1996;Fagan and Davies 2000;Gau and Brunson 2010;Roché and Roux 2017;Terrill and Reisig 2003;Wu, Sun, and Triplett 2009).In sum, different strands of research share the bottom-up assumption, according to which global, macro-level views of the police (whether it be "legitimacy", "trust", "satisfaction" or "support") develop from concrete, microlevel assessments of what the police are doing and how they are doing it. 2

Considerations Unrelated to Policing as Antecedents of Macro-Level Views of the Police
The above research notwithstanding, as early as the 1970s, scholars taking a more sociological or political point of view proposed that factors unrelated to policing may have important effects on how citizens view the police.For example, Chackerian and Barrett (1973) proposed a top-down model, according to which citizens' views of the police result not from a meticulous calculation of factors such as crime rates, arrests and police use of force but rather from the projection of a diffused feeling about the government onto the police as a more specific entity.Indeed, they found a correlation between perceived government accessibility and positive evaluations of law enforcement.
Similarly, Albrecht and Green (1977) argued that attitudes do not exist in isolation from broader value systems.While attitudes may appear to pertain to specific objects in the environment, they are likely to be organized into larger sets, within which they tend to be consistent.As expected, they found a close relationship between attitudes toward the police and evaluations of the courts and the legal system.Somewhat weaker correlations (but still stronger than expected by chance) were found between views of the police and feelings of alienation and powerlessness about the larger political system (also see Stack and Cao 1998).
2 While this assumption raises numerous questions (e.g.do citizens accurately interpret police conduct?; see Worden and McLean 2017), in the present study, we are concerned specifically with the relative roles of perceived police conduct versus other influences unrelated to policing (many of which are also perceptions) in the formation of diffused support for the police.
Such findings suggest there are good reasons to suspect that micro-level assessments of police action and demeanour are not sufficient to understand popular sentiments toward the police fully.Nevertheless, considerations unrelated to police conduct are often not accounted for in analyses predicting macro-level views of the police, and, what is more, the two types of evaluations are rarely considered together as part of a single analysis.Nevertheless, a few exceptions should be noted.Wu et al. (2012) analysed the predictors of public trust in the police in China and Taiwan and found that in addition to police-related considerations (e.g.sense of safety and accessibility of police help), evaluations of government performance (responsiveness, corruption, economic conditions) and cultural factors (e.g.interpersonal trust and orientation toward authority) also showed significant effects.
More recently, Roché and Roux (2017) assessed the effects of various factors, including prior experiences with the police, victimization, living conditions, ethnicity, political attitudes and punitive values, on the perception of police fairness among a French sample.Political attitudes and punitive values were the strongest predictors in their model, accounting for half of the explained variance.Left-wing political affiliation and low levels of punitiveness were associated with more negative evaluations of police fairness, while beliefs that the law should be strictly enforced and that crime is intolerable were associated with more positive assessments.Similarly, Karakus (2017) sought to illuminate the predictors of police legitimacy and cooperation with the police in Turkey and found that in addition to the "classic" predictors of legitimacypolice performance and procedural justicea sense of social cohesion and the performance of the local government also had a significant correlation with police legitimacy.
Echoing early arguments by Albrecht and Green (1977), Mehozay and Factor (2017) proposed that deeply embedded worldviews and normative values shape the legitimacy of the police in the eyes of the public.Their analysis of the responses of over 50,000 individuals from 27 countries to the 2011 European Social Survey revealed that three of the four value systems examined were statistically significant predictors of police legitimacy.They conclude that " : : : the effects of internalized core normative values extend beyond individuals' perceptions of enforcement institutions, meaning that even optimal procedural conduct or efficiency may not affect the attitudes of some populations" (Mehozay and Factor 2017, 172; also see Factor and Mehozay 2023).
Nevertheless, such integrations of police-related and non-police-related considerations as predictors of macro-level sentiments toward the police are not "standard practice".The implications are that our understanding of the roots of police legitimacy (and of other macro-level views of the police) is limited.This, in turn, may lead to an inaccurate interpretation of community surveys and to less than optimal decisions concerning policy and practice.For example, a police chief may assume that the conduct of her/his officers while responding to a new challenge (such as the COVID-19 pandemic) strengthened public support, but this may only appear to be the case because situational factors unrelated to policing (e.g. the "rally effect" in emergenciessee below) were not considered.
One reason why factors unrelated to policing are not routinely considered in studies of macro-level views of the police may be the inherent difficulty of defining the boundaries of any particular analysis.General surveys of public attitudes toward the police are usually not confined to a particular context or set of circumstances.
Thus, identifying all potentially relevant predictors, both related and unrelated to policing, is virtually impossible.The recent COVID-19 pandemic provides a unique opportunity to overcome this challenge by providing a bounded, clearly defined set of circumstances.Thus, it allows for examining the interplay between perceived police conduct during the pandemic (micro-level assessments) and pandemicrelated concerns not associated with policing on diffused support for the police during this period (a broad, macro-level view of the police).
In sum, a bird's-eye view of the literature on public attitudes toward the police suggests that studies investigating the roots of macro-level sentiments toward the police (legitimacy, trust, satisfaction, support) include two different perspectives: the perspective of social psychology (reflected primarily in the work carried out within the framework of Tyler's legitimacy model), which focuses on micro-level assessments of police conduct as the primary antecedents of police legitimacy; and the sociological or political perspective, which focuses on broader societal issues such as views of the government and the criminal justice system.While some unique examples can be found, police-related and non-police-related considerations are typically not examined together as potential sources of macro-level views of the police, and, thus, their reciprocal effects still need to be clarified.The present study seeks to address this gap.We take advantage of a specific, bounded set of circumstancesthe COVID-19 pandemic (see the next section)and ask: What is the role of public evaluations of police conduct in the pandemic, compared to nonpolice-related experiences and views concerning the pandemic in shaping diffused support for the police during this period?
The COVID-19 Pandemic as a Context for Developing Macro-Level Views of the Police By January 2023, research on police and police-community relations in the context of the pandemic was thriving. 3In this section we make no pretence to provide a comprehensive review of this immense body of work, but rather identify the context-specific factors that could have an impact on public support for the police in this period.Analyses of public attitudes toward the police in the context of COVID-19 have typically treated the pandemic as a type of emergency or threat that inevitably involves the police, similar to natural disasters, wars and terrorism (e.g.Perry and Jonathan-Zamir 2020; Sibley et al. 2020), and we follow this approach.

Police Conduct in Emergencies as an Antecedent of Macro-Level Views of the Police
Scholars have argued that police involvement in handling emergencies may both strengthen and weaken the public image of the police.On the one hand, extensive involvement in treating the acute, threatening situation may portray the police as heroic, highly relevant and attentive to public priorities, all of which strengthen public support.On the other hand, given that police resources are limited, excessive focus on the immediate situation could result in neglecting other police responsibilities, weakening public sentiments toward the police.Moreover, a sense of urgency and threat may encourage officers to adopt a "warrior" rather than "protector" policing style, which may, in turn, bring about public resentment (e.g.Jonathan 2010; Jonathan-Zamir, Weisburd, and Hasisi 2015;Perry and Jonathan-Zamir 2020).
Other Emergency-Related Factors as Antecedents of Macro-Level Views of the Police In addition to police conduct, scholars have recognized that factors associated with the emergency (but not directly with policing the emergency) may have an impact on the public image of the police.Jonathan (2010) argued that in situations of emergencies or threats, public views of the police may be influenced by in-group/ out-group theories of sociology (Coser 1956;Simmel 1955) and by the "rally 'round the flag effect" identified in political science research (Mueller 1970(Mueller , 1973)).This theoretical prism postulates that internal cohesion may increase in the face of an external conflict (for a review, see Stein 1976).Accordingly, the rally effect proposes that severe threats may bring about a temporary, short-term wave of support for the national leader and, by extension, to other public institutions (including the police).The rally effect is most likely in the face of a direct, severe and unprovoked threat to the national interest and fundamental national values (Lai and Reiter 2005) and when the levels of conflict and media coverage are high (Baker and Oneal 2001).It has been attributed to factors such as patriotism, a one-sided story in the media, the opposition temporarily muting its criticism of the leadership, and demonstrated leadership competence (Baker and Oneal 2001;Lai and Reiter 2005).
Empirical evidence supports the rally effect, generally (for reviews, see Baker and Oneal 2001;Lai and Reiter 2005) and specifically concerning the police (Jonathan 2010; LaFree and Adamczyk 2017).Interestingly, the rally effect was also identified in the context of pandemics: data from Switzerland reveal high levels of trust in the government at the early stages of the H1N1 ("swine flu") pandemic, followed by a subsequent decline in trust over time (Bangerter et al. 2012;Quinn et al. 2013).Specifically with regard to the COVID-19 pandemic in Israel, Perry and Jonathan-Zamir (2020) identified that during the first peak of the pandemic, public views of the IP about pandemic policing were more favourable than views of "regular" policing, suggesting that, in the short term, the pandemic strengthened the public image of the police.At the same time, Kimhi et al. (2020) and Perry, Jonathan-Zamir, and Factor (2021) identified a drop in support for the IP over time, a finding that is consistent with surveys conducted in the UK (Clements and Aitkenhead 2020;Shaw 2021).
Another path unrelated to policing by which emergencies may influence public views of the police pertains to concerns for the welfare of society and one's wellbeing (e.g.Sibley et al. 2020).These concerns may include, for example, worries over the financial ramifications of the emergency regarding unemployment or recession (Meltzer, Cox, and Fukuda 1999;Smith et al. 2009) and distress related to one's mental and physical health.A large body of literature confirms that experiencing a catastrophe, such as natural disasters, wars, fires and terrorist attacks, may significantly harm one's mental and physical health (Bolin and Kurtz 2018;Bonanno et al. 2010;Norris, Friedman, and Watson 2002).Focusing specifically on pandemics, high levels of fear and anxiety were identified following the 2003 global severe acute respiratory syndrome (SARS) outbreak (Kan et al. 2003;Yu et al. 2005), and slightly greater mental distress was reported following the COVID-19 lockdown in New Zealand (Sibley et al. 2020).In Israel, our study site, surveys reveal a rise in levels of distress and perceived threat, as well as a drop in subjective indicators of wellbeing, throughout the COVID-19 pandemic (Kimhi et al. 2020; also see Horesh, Kapel Lev-Ari, and Hasson-Ohayon 2020).Such worries may affect one's emotional state, mood, world views and expectations of the future more generally.In turn, these views may influence attitudes toward the police, for example, through a reverse halo effect (Nisbett and Wilson 1977;Thorndike 1920).Moreover, the government may be held responsible for such grievances (e.g.Olagoke, Olagoke, and Hughes 2020), which may, by extension, undermine support for the police.

Study Context: Policing the COVID-19 Pandemic in Israel
The COVID-19 pandemic began in Israel on 17 February 2020, with the first confirmed case.Since then, it has been characterized by multiple highs and lows in terms of the number of new cases per day, the basic reproduction number (the R), and the extent and severity of the restrictions imposed on the population (World Health Organization 2021; also see Horesh et al. 2020;Kimhi et al. 2020;Perry and Jonathan-Zamir 2020).Throughout the pandemic, Israeli citizens underwent three lockdowns of varying lengths, in which the police strictly enforced harsh restrictions on the population.These included, for example, the closure of most businesses and schools, strict limitations on gatherings, and the prohibition of leaving one's home except for particular purposes and within a limited radius (Perry and Jonathan-Zamir 2020).Based on the literature reviewed above, particularly the rally effect, we expected to find a drop in diffused support for the IP between initiating the first (8 April 2020) and third (27 December 2020) lockdowns.
More generally, we should note that local policing in Israel resembles local policing in many Western democracies with regard to its basic functions and restraints.At the same time, in terms of public support, cross-national comparisons reveal that public attitudes toward the IP are less favourable compared to other countries.This includes micro-level evaluations of police treatment, as well as macro-level expressions of trust in the police (Hough, Jackson, and Bradford 2013).Moreover, trust in the IP has been declining over the past decade, and, according to a recent survey carried out by the Israel Democracy Institute, in 2022, it hit its lowest point since 2008 (for this and additional information on trust in the IP by sector, age and religion; fluctuations in trust over the last two decades; and a comparison of trust across social institutions in Israel, see Hermann et al. 2022).At the same time, similar to the majority of studies originating from the USA, Australia and Europe, procedural justice has been identified as the primary antecedent of police legitimacy in Israel, while views regarding police effectiveness have been found to be weaker predictors of legitimacy (for a review, see Jackson 2018).
In terms of organization, a review of the IP's history, organizational structure and functions can be found elsewhere (e.g.Jonathan-Zamir and Harpaz 2018; Jonathan-Zamir et al. 2015, 2019;Weisburd, Jonathan, and Perry 2009).What is important to note here is that concerning COVID-19, the IP resembled police agencies in other Western democracies in terms of the challenges it faced (e.g. the enforcement of a constantly changing list of emergency regulations, changes in crime patterns, health threats to officers) and the tasks officers were required to carry out (e.g.frequent explanations of the situation and emergency regulations to citizens, issuance of warnings and fines, handling pandemic-related protests) (Bar-Tzvi 2020; Lum, Maupin, and Stoltz 2020;Neyroud 2020;Perry and Jonathan-Zamir 2020).All in all, and in line with the growing body of research on Israeli policing in recent years, we consider the Israeli setting a useful one for examining the interplay between the different factors that may have influenced public support for the police between the first and third peaks of the COVID-19 pandemic.

Sampling and Participants
Our analysis focuses on the majority population in Israel: non-ultra-Orthodox Jewish adults (self-defined).We clearly recognize the importance of examining the views of minority communities, including ultra-Orthodox Jews and Israeli Arabs; however, in 2020, these sectors were not well represented in web-based survey platforms like the one used here (see below), which were the only feasible surveying method during the pandemic lockdowns.Following the approval of the researchers' departmental ethics committee, the survey was carried out by "Midgam Project Web Panel",4 an online survey platform based in Israel and frequently used by social scientists (e.g.Gubler, Halperin, and Hirschberger 2015;Schori-Eyal et al. 2015).Participants are sampled by "Midgam" according to the distribution of gender and age in the population, as reported by the Israeli Central Bureau of Statistics.
Wave 1 of the survey was administered in the days preceding the first lockdown in Israel (6-8 April 2020).Of 1,798 complete questionnaires, 223 were excluded due to inattentiveness (survey completion within five minutes or less), resulting in a final wave 1 sample of 1,575 participants.Wave 2 was administered near the initiation of the third lockdown (31 December 2020-14 January 2021; 95% responded by 3 January 2021).All 1,575 respondents who provided valid questionnaires in wave 1 were contacted and asked to complete the second survey wave.Of the 1,085 who agreed (69% response rate), 101 were excluded due to inattentiveness, resulting in a final sample of 984 participants who provided valid questionnaires in both waves.Before analysis, this sample was weighted (calibrated) by age, gender and education to make the final sample match the distribution of these characteristics in the Israeli, Jewish, adult and non-ultra-Orthodox population according to national data reported by the Central Bureau of Statistics (2020).Full sample characteristics are provided in Appendix 1.

Survey Procedure and Instrument
The questionnaire began with a standard consent form.Most subsequent items were phrased as statements, which respondents were asked to rank on a scale ranging from 1 ("strongly disagree") to 5 ("strongly agree").The questionnaire was based on previous surveys assessing public attitudes toward the police (e.g.Gau 2011;Jonathan-Zamir and Weisburd 2013;Reisig and Lloyd 2009;Sunshine and Tyler 2003) but also included questions about people's expectations of the police about the COVID-19 pandemic and perceived outcomes of pandemic policing, based on previous studies on policing emergencies (Jonathan and Weisburd 2010; also see Perry and Jonathan-Zamir 2020).In the second wave, several questions were added to measure public assessment of how the police and the government handled the pandemic, broad concerns about its potential effects on society, and personal hardships during this period.Importantly, these items were phrased to reflect the entire pandemic period thus far (e.g."The coronavirus period hindered my health") rather than ad hoc views.
In the present analysis, we use 16 items that capture the themes relevant to our research question.Our dependent variablea macro-level view of the police -"diffused support for the police" (measured in both waves) was operationalized using four items that have been used to measure global, macro-level views of the police in past research, such as "I have trust in the Israel Police" (Jonathan-Zamir and Weisburd 2013) and "I find myself defending the police in conversations with friends" (Sunshine and Tyler 2003).We should note that the literature on public attitudes toward the police is highly diffused and characterized by disagreements and ongoing debates about the definition, operationalization, and construct and discriminant validity of key terms (such as "police legitimacy", "trust in the police", "procedural justice" and more; see Gau 2011;Gau et al. 2012;Kochel 2012Kochel , 2013Kochel , 2018;;Maguire and Johnson 2010;McLean and Nix 2021;Reisig, Bratton, and Gertz 2007;Tankebe 2013;Tyler and Jackson 2013).Thus, in the present study, we take an all-inclusive approach and define our dependent variable in broad terms: macrolevel attitudes toward the police that are not limited to a specific encounter. 5ssessments of the outcomes of pandemic policing (which are micro-level evaluations of perceived police conduct in the context of the pandemic) were designed based on the literature on the potential negative implications of policing emergencies (reviewed above), both in terms of police-community relations ("The role of the Israel Police in enforcing the emergency regulations of the coronavirus crisis negatively affects the relationships between the police and citizens") and in terms of crime control ("Police handling of the coronavirus hampers its other responsibilities, such as handling property crime, violence, drugs and traffic").Through the prism of Tyler's legitimacy model (see above), the first statement echoes the notion of procedural fairness in the context of pandemic policing, while the second is associated with police effectiveness during the pandemic.
Evaluations of how the government has been handling the pandemic were measured using four items, such as "The decisions made by politicians during the period of the coronavirus are tainted by irrelevant motives".On this scale, high scores indicate inadequate performance.The scale of broad concerns about society reflects views regarding potential negative effects on children's education ("The education of the next generation in Israel was significantly hampered during the coronavirus period") and the financial ramifications of the pandemic ("The coronavirus had disastrous effects on the Israeli economy").Finally, the scale of personal hardships during the pandemic reflects one's physical and emotional wellbeing during the pandemic, using four items, such as "The coronavirus period led to much tension in my family and/or immediate environment" and "The coronavirus period hindered my health".Descriptive statistics of all survey items used in the analysis are available in Appendix 2.
The 16 relevant survey items (measured in wave 2), along with 10 covariances between the latent variables, were subjected to a confirmatory factor analysis (CFA) in order to validate the above themes (see Table 1).The CFA indicated a good fit (χ 2 (94) = 288.795;comparative fit index [CFI] = 0.958; root mean square error of approximation [RMSEA] = 0.046) and supported the use of these latent measures (Byrne 2009;Chen et al. 2008).As can be seen from Table 1, all factor loadings are significant (p < 0.001) and higher than the traditional threshold of 0.4 (Reisig et al. 2007).The Cronbach α value for most scales is higher than the commonly used threshold of 0.7, which supports the reliability of the scales (DeVellis 2003; Factor, Kawachi, and Williams 2013a).The Cronbach's α values of the "outcomes of the pandemic-policing" scale (α = 0.61) and "broad societal concerns" scale (α = 0.63) are somewhat lower, which is expected given that there are only two survey items in each of these scales (Keizer, Dykstra, and Jansen 2008).At the same time, their conceptual relationships, as well as Pearson's correlations between the items in each scale (r = 0.438 and r = 0.463, respectively; p < 0.001), support combining each pair into a single measure.

Analytic Strategy
The analysis centres on wave 2 data, while drawing from wave 1 the preliminary levels of diffused support for the police and sociodemographic information (wave 2 data were used for personal information that may have changed between the survey waves, such as contact with the police in the past year; see Appendix 1).Some of the variables in our model included observations with missing data (0.41-11.58%; see Appendices 1 and 2 for valid n of all variables included in the analysis).In order not to lose these observations, we employed maximum likelihood estimation with missing values, which is a direct and theory-based estimation (Arbuckle 1996;Byrne 2009;Rottweiler, Gill, and Bouhana 2020). 66The main concern with missing data is the question of whether the sub-sample of observations with missing values is different from the rest of the sample in significant ways relevant to the analysis.In order to address this concern, we conducted three sensitivity tests.First, we compared the mean of the dependent variable across observations with and without missing data and did not find a significant difference (p = 0.459).This indicates that the probability of having missing cases in our independent variables is not associated with the dependent variable (Allison 2001).Second, our main sociodemographic and control variables were subjected to Little's missing completely at random test, which examines whether missing cases are independent of the observed and unobserved data, and the results were insignificant (p = 0.404), indicating that there were no significant differences between the means of different missing value patterns and that the missing variables were completely random (Li 2013).Finally, an analysis of the observations with imputed data yielded similar results to the analysis of all cases.
Following the process of scale construction described above, we begin by examining whether diffused support for the IP indeed deteriorated between the two survey waves.For this purpose, we created indices of "diffused support for the police" in both waves using the weighted average of the items based on the loadings obtained from the CFA (Factor et al. 2013b). 7We then use structural equation modelling (SEM) to predict public support for the police in wave 2 while controlling  for support in wave 1. SEM was deemed particularly useful because it allows for the analysis of both latent and observed variables, provides explicit estimates of measurement errors, thereby increasing accuracy, and enables observing intermediate effects (Byrne 2009;Sunshine and Tyler 2003).Importantly, by accounting for initial support in wave 1, the model illuminates the contribution of our four main independent variables of interest to the change between the waves in levels of diffused support.Moreover, preliminary levels of support are expected to reflect (and thus control for) the various police-related considerations in the background (including preliminary levels of police fairness and effectiveness), thus allowing for a clear, well-defined assessment of pandemic-related influences.

RESULTS
Following the literature reviewed above on the potential effects of the COVID-19 pandemic on public attitudes toward the police, we begin by exploring whether diffused support for the IP eroded from wave 1 to wave 2. In wave 1, 21% of respondents ranked their support for the police at the lower end of the scale (1-2), indicating little police approval.In wave 2, this figure rose by nearly one-third to 27%.This change occurred within just nine months and among the same individuals.The drop in diffused support is also evidenced in a statistically significant decline of nearly 7% in the average "diffused support" score between wave 1 (2.75;SD = 0.85) and wave 2 (2.56; SD = 0.85) (t = 5.80, p < 0.001).Thus, our data confirm that in line with the rally effect (e.g.Jonathan 2010) and earlier findings (e.g.Kimhi et al. 2020), diffused support for the IP weakened over the nine months of the COVID-19 pandemic examined here.
Next, we turn to the main purpose of our analysisilluminating the factors behind the drop in diffused support.Using SEM, we estimate the effects of our four main independent variables of interest (outcomes of policing, government performance, broad societal concerns, and personal grievances, all in relation to the pandemic) on diffused support for the police as measured in wave 2, while controlling for both support in wave 1 and sociodemographic and other personallevel characteristics (included in all paths).The SEM is presented in Figure 1.For simplicity, the Figure only shows the main variables of interest and the standardized regression coefficients, while the regression weights of the statistically significant control variables are presented in Table 2.The SEM indicates a good fit to the data (χ 2 (307) = 960.632;CFI = 0.903; RMSEA = 0.047) (Chen et al. 2008;Cheung and Rensvold 2002;Hair et al. 2006).
Figure 1 reveals that evaluations of the way the government has been handling the COVID-19 pandemic have a statistically significant association with the drop in diffused support for the IP (β = -0.23,p < 0.01): the higher the evaluations of inadequate government performance, the lower the levels of support.Importantly, apart from preliminary views of the police, government performance during the pandemic is the only latent variable with a direct, statistically significant effect on diffused support for the police in wave 2. Assessments of the outcomes of pandemic policing (β = -0.04,p = 0.62), broad societal concerns about COVID-19 (β = 0.07, p = 0.36) and personal, pandemic-related hardships (β = 0.04, p = 0.41) have no direct, statistically significant effects on the drop in diffused support.
Nevertheless, it should be noted that there are multiple, statistically significant relationships amongst the four latent variables: pandemic-policing outcomes are significantly correlated with government performance during COVID-19 (β = 0.38, p < 0.001) and with societal concerns concerning the pandemic  (β = 0.36, p < 0.01).The performance of the government is also correlated with societal concerns (β = 0.52, p < 0.001) and with personal hardships as a result of the pandemic (β = 0.17, p < 0.05).Finally, there is a significant correlation between societal concerns and personal hardships (β = 0.34, p < 0.001).
As shown in Table 2, no control variablessociodemographic or other personal characteristicsdirectly contributed to the erosion in diffused support for the police.Nevertheless, there is a negative association between one's level of religiosity and viewing the government's performance during the pandemic as inadequate.In other words, the less religious viewed the government more critically,8 which, in turn, undermined their support for the police.We should also note that personal hardships during the pandemic are negatively associated with age, level of religiosity and income and positively correlated with belonging to a group at high risk of developing serious illness if infected with COVID-19.Concerns about the broad, societal, negative effects of the pandemic are more prevalent among women and the less religious.

DISCUSSION
Much of the literature on public attitudes toward the police, and particularly the work carried out within the social-psychology framework, takes it as a given that global, macro-level views of the police develop in a bottom-up fashion, that is, they are primarily the outcome of concrete, micro-level assessments of police conduct (e.g.Nagin and Telep 2017).At the same time, studies taking a more sociological or political perspective have identified that factors unrelated to policing, such as views of the legal and political systems and normative values, may also have important effects on broad views of the police (e.g.Mehozay and Factor 2017).Nevertheless, the two types of considerations are rarely examined together in a single analysis; thus, we know little about their reciprocal effects.In the present study, we have taken advantage of a specific, bounded, well-defined set of circumstancesthe COVID-19 pandemicand examined the factors behind the drop in citizens' diffused support for the police (a macro-level assessment) between the first and third peaks of the pandemic in Israel.We analysed the relative effects of perceived police conduct in relation to the pandemic and its outcomes (micro-level evaluations of police conduct) and other factors associated with the pandemic but unrelated to policing (the way the government has been handling the pandemic, concerns for the wellbeing of society, personal hardships as a result of the pandemic) on the drop in diffused support for the IP, to come to more general conclusions regarding the relative roles of perceived police conduct versus non-policing considerations in the formation of macro-level views of the police.
In addition to confirming that, indeed, there was a deterioration in majority communities' support for the police during the nine months examined here, our findings show that the drop in support was directly influenced only by assessments of the way the government handled the situation.Citizens who expressed little trust in the figures leading the efforts against the pandemic, thought that decisions were tainted by irrelevant motives, and believed that overall the government had failed to do a good job addressing the COVID-19 situation, were more likely to express little support for the police in the third peak of the pandemic (end of December 2020) compared to the first peak (beginning of April 2020).Surprisingly, assessments of pandemic policing and its outcomes in terms of crime control and policecommunity relations had no direct effects on the erosion of public support for the police.
Clearly, evaluations of the government's performance do not exist in isolation, but, as argued by Albrecht and Green (1977), appear to be part of a broader attitudinal structure.They were found to be correlated with perceived outcomes of policing the pandemic, personal hardships in this period, and concerns over the broader effects of the pandemic on society.Moreover, our model reveals multiple correlations amongst the four main explanatory variables and between them and background characteristics.Nevertheless, we view the main finding of the analysisthat evaluations of police conduct during the pandemic (micro-level assessments of police action/behaviour) had no direct effects on the drop in diffused support for the police (macro-level views of the police)as highly important.It suggests that the erosion in diffused support was not the result of a bottom-up effect (whereby specific, micro-level assessments of police conduct influenced global views of the police) but rather of a top-down effect, whereby evaluations of the government at large influenced views of the police as a particular arm of the government.Put differently, beyond anything the police did (or did not do) over the course of the pandemic and the outcomes of their actions (in the eyes of the public), it was the loss of trust in the government (irrespective of its causes) that eventually led to more negative views of the police.Notably, the Israeli Democracy Index shows a significant drop in trust in the government in 2020 (Hermann et al. 2020).
As reviewed earlier, this top-down interpretation of our findings conforms with the observations of Chackerian and Barrett (1973), who argued that citizens' views of the police are actually a projection of a diffused feeling about the government onto the police as a more specific object.It also corresponds with the body of research illuminating the importance of factors such as broad views of the legal and political systems (Albrecht and Green 1977), political alienation (Brown and Coulter 1983), satisfaction with the quality of the local government (Brown and Coulter 1983), political affiliation and punitive values (Roché and Roux 2017), and normative values (Mehozay and Factor 2017) in the formation of the public image of the police.It also aligns with previous propositions linking public attitudes toward the police in emergencies or threats to the "rally 'round the flag effect" (Jonathan 2010; LaFree and Adamczyk 2017; Perry and Jonathan-Zamir 2020).
What are the implications of our findings?Recent years have witnessed unprecedented public criticism and protest against the police, particularly in the USA (e.g.Cobbina-Dungy et al. 2022;Lum et al. 2021).While few would disagree that all citizens in democratic societies deserve fair and unbiased policing services (as argued by the protestors), public protests of such magnitude raise important questions about how citizens develop their beliefs about the police and the extent to which they reflect police conduct as opposed to various other factors, such as the conduct of the government more generally.Clearly, we are not suggesting that what the police do (or do not do) has no bearing on public sentiments.The importance of police conduct is reflected not only in prior work but also in the present analysis.Our main goal was to explain the drop in support for the police and, thus, we have focused on the effects of the four main predictors that reflect pandemic-related experiences after controlling for baseline levels of support.At the same time, a more general consideration of the analysis reveals that, not surprisingly, diffused support for the police in wave 2 was primarily influenced by diffused support in wave 1.These baseline evaluations, in turn, are expected to develop from concrete assessments of police conduct, including procedural justice and police effectiveness.
Nevertheless, our findings do draw attention to the fact that a scenario where perceived police conduct has little to do with public support (or lack of support) for the police is not unreasonablethe drop in support for the police during the COVID-19 pandemic was not the result of evaluations about the police.At times, both satisfaction and dissatisfaction with the police may be only loosely tied to policing.Thus, we argue that the bottom-up approach to macro-level views of the police should not be taken for granted.Researchers and practitioners would be well advised to interpret public attitudes toward the police with caution and strive to develop a more sophisticated understanding and measurement of these views (for a recent example, which also finds that non-policing considerations have important effects on a "scorecard" of evaluations of the police, see Weisburd et al. 2022).
More specifically, we encourage future research to illuminate the various contextspecific, non-policing considerations that may have an impact on global views of the police and take them into consideration in analyses predicting broad public views of the police.Furthermore, our findings call for inquiries into potential interactions between the circumstances and the magnitude of both types of influences.For example, when developing global views of the police, do citizens assign less weight to police conduct, specifically in emergencies (compared to non-emergency situations)?In what contexts or situations does the top-down approach better explain global, macro-level views of the police than the bottom-up model?In terms of policy, policymakers and practitioners would be well advised not to congratulate themselves too much when public support for the police is on the rise or, alternatively, berate themselves when support is deteriorating.In both cases, attitude fluctuations may not reflect what the police are actually doing or are perceived to be doing.Before drawing conclusions regarding the success or failure of the police, it is recommended to carefully analyse the situation to disentangle the effects of police conduct from those of other situational characteristics that are outside the influence of the police.
Before concluding, the limitations of our study should be acknowledged.First, we recognize the shortcomings of our sample as a non-probability convenience sample of registered panellists.We should note, however, that this survey method has become common in the study of public attitudes towards the police (particularly using Amazon's Mechanical Turk; see Pickett, Nix, and Roche 2018).Moreover, it was the only feasible method during the pandemic lockdowns, and our final sample was weighted to correspond with the sociodemographic makeup of the population.Second, as noted earlier, for the same reason, our sample only reflects the views of majority communities in Israel.At the same time, although Israeli minority communities tend to view the police more negatively (e.g.Factor et al. 2014;Hasisi and Weitzer 2007;Hermann et al. 2022), we have no reason to suspect that the relationship between police-related and non-police-related considerations, and broad support for the police, would be different for them.
Third, additional survey items reflecting other potentially relevant views, such as more nuanced political ideologies, would have strengthened the analysis.Moreover, additional aspects of policing the pandemic (such as more direct measures of fairness and effectiveness in pandemic policing) may have given more weight to the "pandemic-policing outcomes" scale in the analysis.Finally, we argued earlier that a well-defined set of circumstances allows researchers to identify the relevant factors (both related and unrelated to policing) that may influence support for the police within that context.However, the flip side of that coin concerns generalizabilitycan our findings be generalized to other settings, such as those where there is no emergency?These limitations can only be addressed in future studies, replicating our analysis with additional survey items and in different populations, countries, policing contexts and circumstances.Despite these inherent limitations, our analysis offers a useful example demonstrating how public support for the police may sometimes have little to do with policing.

CONCLUSIONS
In the present study, we set out to illuminate the relative roles of perceived police conduct (micro-level assessments of the police) and situational influences unrelated to policing in forming global, macro-level public views of the police.In doing so, we have taken advantage of the drop in diffused support for the IP over nine months during the COVID-19 pandemic.We found that the only variable which directly influenced this drop was evaluations of the way the government had been handling the pandemic.Surprisingly, assessments of pandemic policing and its outcomes had no direct effects.This finding provides an important example of the fact that, sometimes, public views of the police may have little to do with perceived police conduct.We thus propose that policing scholars and practitioners do not take the bottom-up approach to macro-level views of the police for granted, interpret the sources of these views with caution, and strive to develop a more complex understanding of how these views develop in particular contexts.
Tal Jonathan-Zamir is an Associate Professor at the Institute of Criminology, Faculty of Law, the Hebrew University of Jerusalem.Her work, published in leading journals, focuses on policing, particularly policecommunity relations and evidence-based policing.She has investigated police legitimacy and procedural justice from the perspective of citizens, communities, police officers and neutral observers in diverse contexts such as routine encounters, security threats, protest events, airport security and street level.She has also examined the psychological mechanisms underlying police officers' orientation to evidence-based policing, effective mechanisms for police training, and the effects of COVID-19 on police-community relations in Israel.
the same sense of "right" and "wrong" as I do 0.68 I find myself defending the police in conversations with friends 0.51 Pandemic-policing outcomes (α = 0.61) The role of the Israel Police in enforcing the emergency regulations of the coronavirus crisis negatively affects the relationship between the police and citizens 0.55 Police handling of the coronavirus hampers its other responsibilities, such as handling property crime, violence, drugs and traffic 0.78 Inadequate government performance during the COVID-19 pandemic (α = 0.90) I have trust in the figures managing the coronavirus crisis in Israel (reversed) 0.82 The decisions made by politicians during the period of the coronavirus are tainted by irrelevant motives 0.75 The state of Israel deserves the score of "zero" for the management of the coronavirus crisis 0.72 The government in Israel does the maximum possible in handling the coronavirus crisis (reversed) 0.81 Broad societal concerns in relation to the COVID-19 pandemic (α = 0.63) The education of the next generation in Israel was significantly hampered during the coronavirus period 0.58 The coronavirus had disastrous effects on the Israeli economy 0.79 Personal hardships during the COVID-19 pandemic (α = 0.80)The coronavirus period led to much tension in my family and/or immediate the standardized confirmatory factor analysis loadings.All items are significant at the p < 0.001 level.Survey items are translated from Hebrew.

Figure 1 .
Figure 1.Structural equation model estimating the decline in diffused support for the police between wave 1 and wave 2. *p < 0.05; **p < 0.01; ***p < 0.001.Statistically significant relationships are marked by bold arrows.Values represent the standardized regression coefficients.
the Israel Police in enforcing the emergency regulations of the coronavirus crisis negatively affects the relationship between the police and citizens 0.62 Police handling of the coronavirus hampers its other responsibilities, such as handling property crime, violence, drugs and traffic 0.70 Inadequate government performance during the COVID-19 pandemic I have trust in the figures managing the coronavirus crisis in Israel (reversed) 0.81 The decisions made by politicians during the period of the coronavirus are tainted by irrelevant motives 0.72 The state of Israel deserves the score of "zero" for the management of the coronavirus crisis 0.70 The government in Israel does the maximum possible in handling the coronavirus crisis (reversed) 0.80 Broad societal concerns in relation to the COVID-19 pandemic The education of the next generation in Israel was significantly hampered during the coronavirus period, I am much more nervous and angry 0.87 During the coronavirus period, I feel more stress and anxiety 0.84 a The values represent the standardized confirmatory factor analysis loadings.All items are significant at the p < 0.001 level.Comparative fit index = 0.963; root mean square error of approximation = 0.04.Survey items are translated from Hebrew.
How Do Macro-Level Views of the Police Develop?Policing-Related Considerations as Antecedents of Macro-Level Views of the Police

Table 1 .
Confirmatory Factor Analysis of Main Themes a

Table 2 .
Structural Equation Modelling Regression Weights for Sociodemographic Characteristics with Statistically Significant Effects