The urban-rural divide in police trust: insights from Kenya

ABSTRACT The police occupy a central role in the functioning of the state by being tasked with upholding security, law and order. Across the African continent, the public has little trust in the police, but such perceptions are subject to considerable subnational variation. In this study, we are interested in how the different contexts in which the police operate affect police-citizen relations. We ask: How does an urban versus rural environment shape citizens’ trust in the police? We address this question within the context of Kenya, using geocoded survey data from Afrobarometer. We theorise that the rural versus urban environment will shape citizens’ experience with the police in ways that affect their attitudes toward the police. Specifically, we argue that in a context where the police have frequently been employed to repress specific sociopolitical groups, urban residents, living in denser and more diverse environments compared to rural residents, are more prone to have first- or second-hand experiences of the police that result in diminished trust towards them. Our results support these propositions: We find a strong and robust relationship between urban residence and lower levels of trust in the police. The relationship holds when controlling for respondents’ political alignment, which likely conditions people’s perceptions of state institutions. Qualitative evidence from interviews provide additional understanding of the urban-rural divide we identify. Our results provide important insights into the contextual dynamics that shape individuals’ trust in the police, and underline the importance of efforts to improve police-community relations in urbanising contexts.


Introduction
The police occupy a central role in the functioning of the state by being tasked with upholding security, law and order.In many contexts, the police also serve a conflict preventing function, by means of deterrence as well as dispute resolution before conflicts escalate into violence.However, the police have also been applied as a tool for intimidation, control and repression of dissent.In many African countries, demands for police reform have followed in the footsteps of recent reports about police misconduct.In Nigeria, the end-SARS movement gained widespread attention, and similar protests have taken place in, for example, South Africa and Ghana.Across the continent, the public has little trust in the police.In one study, the police are seen as the most corrupt state institution in 11 of 18 African countries surveyed (Sanny and Logan 2020).Research shows that experiences of corruption and other forms of misconduct affect how citizens assess police effectiveness, procedural justice and trustworthiness (Tankebe 2010).Studies also demonstrate that the broader conditions influencing relationships and trust between citizens and police vary sub-nationally and are contingent on socio-political circumstances.
We address one possible determinant of sub-national variation in how the police is perceived, posing the question: How does an urban versus rural environment shape citizens' trust towards the police?We argue that urban and rural areas feature different social and institutional settings that influence police-citizen relations, ultimately impacting public attitudes.Specifically, we suggest that in a setting where the police have frequently been used to repress specific sociopolitical groups, urban residents, living in denser and more diverse environments than rural residents, are more prone to have firstor second-hand experiences of the police that lead them to conclude that policing is unfair.These experiences in turn translate into mistrust.We analyse these propositions in the context of Kenya, using geocoded survey data from the .In addition, to validate and probe the findings, we draw on qualitative insights based on research conducted in Kenya. 1  The study advances research on police-citizen relations in contexts affected by a history of colonial policing and violent conflict in two important ways.First, recognising the importance of the police in both the prevention and production of violence (Eck et al. 2021), we focus on understanding variation in police-citizen relations across space and in the different environments the police operate.Specifically, we begin from the assumption that the capacity of the police to carry out their duties relies on the legitimacy they possess (Tyler 2003).Second, while some previous research has identified a rural-urban divide in how citizens interact with and perceive of the police (Boateng 2018, Dlamini 2020, Ndoma 2020, Sanny and Logan 2020), few studies have attempted to theorise and empirically investigate why and how the rural and urban environments that the police operate in matters.We build on previous work showing that when the police engage in repression, political partisanship shapes how people perceive the police (Curtice 2021).In line with these insights, we postulate that experiences of police repression and consequently citizens' trust in the police, are affected by the individual's political alignment and communal belonging.However, we develop and evaluate the argument that that in a context where policing is politicised, the urban-rural context will have an independent effect on trust among both opposition and incumbent supporters.
Kenya is a suitable case for investigating urban-rural dimensions of citizens' perceptions of the police.First, baseline trust in police is low and like many other African states, policing institutions in Kenya have been strongly shaped by a colonial legacy, state repression, violent communal conflicts and spillover effects from armed conflict in neighbouring states.Both rural and urban areas have been affected by violence and it tends to be associated with electoral and macro-political dynamics.Linked to these processes, the police have been implicated in political violence undermining their legitimacy as a body protecting all citizens equally (Ruteere 2011, Saferworld 2021).In addition, the police have especially in urban settings been involved in killings disproportionally affecting poor, young men in marginalised settlements (Price et al. 2016, Kimari 2017, van Stapele 2020).Second, Kenyan politics have to a large extent revolved around rural interests, and the country still contains a strong rural base. 2 Compared to many other African countries, the urbanisation level is relatively low (27% in 2018 according to United Nations data (UNDESA 2018: 21)).At the same time, Kenya is urbanising rapidly, and the urban population is distributed across a relatively high number of large and medium-sized cities (OECD/SWAC 2023).These changing dynamicsmirrored in many other African statesmake the issue of rural-urban divides important to investigate.Third, Kenya has relatively strong state institutions, but their effective presence varies considerably across different parts of the country (Chopra 2009), and security functions are sometimes delegated to local informal or hybrid institutions (Mkutu 2018, Mutahi 2021).Such dynamics are likely to impact on police-citizen interactions and on perceptions of the police.Finally, Kenya is undergoing police reforms aimed at improving the interaction between police and local communities (Lid and Okwany 2020, Diphoorn and van Stapele 2021, Mutahi et al. 2023).Understanding how the contexts in which the police operate shape citizens' perceptions of the police provide important insights into the conditions for such reforms to be successful.

Previous research
Trust is central to policing: It affects both objective measures of 'police effectiveness' and subjective perceptions of safety among the public.A large body of research has found that public propensity to cooperate with the police and to obey the law is heavily affected by how the police behaves and whether citizens perceive their conduct to be fair and effective (e.g.Kelling and Coles 1997, Sunshine and Tyler 2003, Tyler 2003, Tankebe 2013).A key finding is that perceptions of fairness matter more than objective and subjective measures of effectiveness (Tyler 2004).However, knowledge about the conditions under which perceptions about the police are formed, and if they are subject to change, is limited.Further, the vast majority of this research has focused on Latin America and the West, and particularly on the US.It remains unsettled whether the findings from these contexts apply elsewhere.
Research from African contexts has underlined how a different historical and institutional context has shaped interactions between police and citizens.Notably, many African states have inherited or developed internal security apparatuses with low legitimacy, heavily relying on coercion and intimidation, with implications for trust in the police (Tankebe 2009).For instance, in Ghanawith a context of poor police performance and a legacy of colonial policingthe perceived legitimacy of the police is so low that a minimum level of effectiveness matters more than fairness (ibid.).Studies in other African contexts -Nigeria, Sierra Leone and South Africahave found that both fairness and effectiveness affect public willingness to cooperate with the police (Akinlabi andMurphy 2018, Boateng et al. 2022).
How do sub-national dynamics affect trust in the police?Studies from North America have highlighted demographic factors (notably, race) as conditioning the relationship between police and citizens and, in turn, the attitudes of the public towards the police (see e.g.Thomas and Hyman 1977, Prine et al. 2001, Nofziger and Williams 2005).In societies ravaged by armed conflict, the local intensity of conflict has been shown to influence public perceptions of how effective, fair and trustworthy the police is (Deglow and Sundberg 2021), and research has also assessed the impact of police reform intended to improve the effectiveness and legitimacy of post-war security functions, for example in Liberia (Blair et al. 2019, Blair and Morse 2021, Nilsson and Jonsson 2023). 3Some studies have also analysed if attitudes towards the police differ between urban and rural contexts, but results are mixed (Brown and Benedict 2002).Several studies find no, or unclear, effects of urban/rural location on trust in the police and the willingness to report crime (e.g.Skogan 1984, Jackson andBradford 2019).Other studies have identified differences between rural and urban contexts, which are partly conditional on aspects like socioeconomic status.For instance, one study compared different locations in the US, and found that the urban poor held the most negative attitudes toward police, while rural residents and the urban middle class held more positive attitudes to the police (Albrecht and Green 1977: 75).Another study, also from the US, found that students in urban areas had more negative attitudes to police than students in suburban areas (Hurst and Frank 2000).In Canada, by contrast, studies have also found rural residents to be less satisfied with the police and their ability to cooperate and ensure the safety of citizens (Ruddell and O'Connor 2022).
Studies focusing on African contexts also give indications of an urban-rural divide in attitudes towards the police.For instance, a recent multilevel cross-national study by Boateng found that across Africa, urban residents were less likely to perceive the police as legitimate, than rural-based residents were (Boateng 2018(Boateng : 1114)).Studies from individual African countries add more nuance to these patterns.For instance, in Zimbabwe (2018), reported trust in both the police and army was '10 percentage points higher in rural than in urban areas' (Ndoma 2020: 3).In contrast, survey-based research from South Africa found no clear effect of urban-rural residence on attitudes towards the police (Fry 2013).Finally, some studies focus on the urban-rural divide as a spectrum rather than as a clear dichotomy (Gizelis et al. 2021, Meth et al. 2021).Dlamini (2020), for example, found that residents in areas close to the city centre displayed less trust and less positive attitudes toward the police than those living on the city outskirts.
Differences in attitudes toward the police may in turn relate to citizens' experiences with the police.Analysing recent survey data from 18 African countries, Sanny and Logan (2020: 10) find that urban residents interact more often with the police, whereas rural residents who do interact with police are more likely to face demands for bribes.Studies of police-citizen interactions in Kenya have focused on the varying outcomes, and challenges, of successive efforts of police reform in recent decades, including an increasing emphasis on community policing and moving away from a police force to a police service (Osse 2016, Gjelsvik 2020, Diphoorn and van Stapele 2021).This work underlines how efforts at police reform have often tended to reinforce existing oppressive structures (Ruteere and Pommerolle 2003), and that contextual dynamics including 'conflicting socio-economic and political interests at the community and national levels' condition the prospects for building trust toward the police (Gjelsvik 2020: 19).While these studies provide important insights into the factors that shape local attitudes towards the police in Kenya, they have not systematically analyzed sub-national patterns and their causes.
In sum, existing research has emphasised the importance of attitudes toward the police, but only to limited extent explained how and why such attitudes vary sub-nationally.For long, a focus on Western states dominated the field, but recent advances have made important contributions to knowledge about attitudes toward the police in African contexts.Within this literature, findings regarding how socio-political and institutional factors shape attitudes remain mixed.Several studies identify an urban-rural divide in attitudes to the police, but do not theorise why this is the case and/or do not control for the possibility that this relationship is driven by political dynamics in urban and rural areas.While several important studies of the factors shaping Kenyans' perceptions of the police exist, we know little about how the urban-rural contexts affect the propensity of citizens to mistrust the police.

Police-citizen relations and patterns of (mis)trust in the police
Our main theoretical interest is how the different contexts in which the police operate affect policecitizen relations, focusing on how an urban versus rural environment shape the public's perceptions of the police.Public perceptions of the police encompass a range of issues including citizens' subjective assessment of the effectiveness and fairness in how the police carry out their work and exercise their authority (Tyler 2003).In this study, we look at trust in the police as an important feature of police-citizen relations.We operate from the assumption of relatively low baseline levels of public trust in the police in a context with a history of state repression (including inherited colonial policing methods), ongoing insecurity, and politicised security institutions (Ruteere 2011, Hassan 2020).We also assume that important features of the social and institutional environment in which the police operate differ between rural and urban areas.Conceptually, we understand cities and other urban centres as sites characterised by high population density, frequent interaction between diverse groups and interests, and nodal points for economic accumulation and political networks (Simone 2004).By contrast, rural areas are characterised by lower population density and more social homogeneity (Magadi 2017).In practice, the distinction between urban and rural is not clear-cut, and the extent to which urban dynamics are present varies within cities (for instance, city centre vs more village-like suburbs) and across rural areas (for instance, at what point a small town can be considered 'urban') (Simon 2008, Agergaard et al. 2019).Acknowledging this complexity, we theorise about police-citizen interactions in what can be considered as ideal-type urban and rural contexts.In the methodological section, we discuss how this approach corresponds to our empirical strategy.
We expect urban residents to have lower trust in the police than rural residents.We theorise this to be the case, because the rural and urban environment create different conditions for police-citizen relations (Wang andSun 2020, Ruddell andO'Connor 2022).In settings where the police have been used as a tool for repression of specific socio-political groups, we argue that urban residents are more likely to have firstor second-hand experiences that lead them to conclude that policing is not fair.As noted above, previous research has suggested that perceptions of fairness are strongly correlated with perceived legitimacy and trust in the police (Tyler 2004).Fairness relates to police conduct and whether or not citizens perceive that they are treated equally.Citizens form an opinion of police fairness based on their own experience of police interactions, as well as observations of police behaviour and general knowledge about police behaviour in their community (Tyler 2004).When citizens are treated with respect by the police and are treated equally to other groups, as well as are allowed to provide their account of how and why a situation emerged, they are more likely to perceive the police as fair (Sunshine andTyler 2003, Tyler 2004).
The argument for why we expect urban and rural residents to have divergent experiences of the police unfolds in two steps.First, we expect that police-citizen interactions are overall more visible and frequent in cities than in rural areas. 4In rural areas, police cover large jurisdictions and geographical distances which influence how fast communities and their citizens can be serviced (Weisheit et al. 1995).In many rural locations, police are simply 'spread thin' (Ruddell and O'Connor 2022: 117).While the distribution of police and their presence vary sub-nationally, presence is generally lower in rural areas of Kenya, as in most countries.The most visible security force in the most rural areas of Kenya is the Kenya Police Reserve (KPR), especially in arid and semi-arid regions, where they are the main source of security against bandits and cattle raiders (Agade 2015). 5Under the National Police Service Act (Government of Kenya 2012), the Reserve is mandated to recruit and train personnel that could be deployed to assist both the Kenya Police Service and the Administration Police Service in the implementation of their mandates.Although the KPR falls under the Officer Commanding Police Division, the Chiefs also play a role in the daily management of the KPR.In both urban and rural areas, police officers are deployed away from their home areas and frequently reshuffled (Hassan 2017, Mbuba 2018).However, KPR are more likely to work within their own communities of origin.
Recent studies emphasise that the role of the police in Kenya needs to be understood in the context of a broader spectrum of actors involved in policing or security governance (Mkutu 2018, Mutahi 2021).Overall, in many rural areas, formal policing is of relatively low importance, because other strong institutions for managing local security exist.Notably, the authority of traditional and religious conflict management structures is often more well-established than in urban areas (Adan and Pkalaya 2006), providing a pre-existing infrastructure for rural populations to solve disputes and seek recourse when faced with crime and insecurity. 6The Kenyan system in which rural police to a large extent worked through (neo)customary institutions is a legacy of the colonial period (Ruteere 2011).When the police constitute a less relevant body for rural populations than alternatives, the frequency of contacts between the police and rural residents is further reduced.As a result, rural residents are less likely to have direct contact with the police than urban residents.
Theoretically, low police visibility and interaction could translate into lower public trust.Yet in line with the 'asymmetry thesis' about police-citizen interaction, we expect the opposite in a context where baseline trust is generally low.The 'asymmetry thesis' posits that 'trustworthiness is easy to lose and hard to win' (Oliveira et al. 2021(Oliveira et al. : 1004)).In essence, it holds that negative encounters with the police (such as police brutality, unfair treatment, or inability or unwillingness of the police to address a crime one reports) tend to have a strong negative effect on attitudes toward the police, whereas positive encounters have only a weak positive effect, if any at all (Skogan 2006).Although several studies have found empirical support for this thesis, the evidence is not uniform.In a longitudinal study of survey respondents in Australia, Oliveira et al. (2021) found that negative encounters reduced trust in police fairness and effectiveness, whereas positive encounters increased trust in police fairness, but not effectiveness.However, given the overall poor police performance and corruption that has been documented in Kenya (Ruteere and Pommerolle 2003, Akech 2005, Hope 2019), we believe encounters are more likely to be perceived as negative.
Second, the heterogeneity of urban contexts makes it more likely for urban residents to directly observe unfair treatment of different groups (their own and others) by the police or having been informed by others about police misconduct.Urban areas are by definition more densely populated than rural areas, but also tend to be more heterogeneous in their social, economic and political configuration.This heterogeneity, alongside informal and often rapidly changing networks, contribute to the 'complexity of socio-spatial relations in the city' (Bjarnesen and Utas 2018: S1).While rural areas of Kenya tend to be relatively socio-politically homogenous (Magadi 2017), cities and urban areas feature high diversity and inequality.In Kenya's cities, people from different ethnic and religious groups live closely together and frequently interact across identity lines.Overall, there is ample evidence of how the police have been used by incumbents to repress opposition-aligned ethnopolitical groups, and of everyday bias in policing (Ruteere and Pommerolle 2003, Ruteere 2011, HRW 2013, Ruteere et al. 2013, HRW 2017, Mutahi 2018).In addition, police killings have been particularly common in urban areas, where young men from the informal settlements have been profiled as associated with crime, criminal gangs and terrorism (van Stapele 2016, Jones et al. 2017, van Stapele 2020).The police have also employed a shoot-to-kill policy against certain urban gangs, while supporting or colluding with others (Mutahi 2018).While many residents in urban informal settlements feel mistreated and harassed by the police (Price et al. 2016), the poor state of security has also given rise to support for vigilantism and on occasion even police involvement in vigilantism (Wairuri 2022).
Based on these patterns regarding demography and police killings, urban residents belonging to both opposition and incumbent-aligned communities are more likely to witness how specific groups are treated differently by the police.In more homogenous rural areas, police misconduct may not to the same extent translate into perceptions of unfairness.Importantly, while opposition supporters in both urban and rural areas may be subject to police misconduct, our argument implies that urban residents regardless of political alignment will be more likely to perceive the police as unfair.
Taken together, the theoretical arguments above form the basis for the following main hypothesis: H1: Individuals living in urban areas hold lower levels of trust in the police than those living in rural areas.

Research design
Our main approach to assessing the urban-rural divide in the public's perceptions about the police relies on several rounds of Afrobarometer surveys conducted in Kenya 2011-2019(BenYishay et al. 2017, Afrobarometer 2019).The Afrobarometer is a useful and widely used source of information capturing citizens' attitudes on a range of socio-political issues.Our unit of analysis is the individual.We include rounds 5 through 8, which are all carried out after the adoption of Kenya's 2010 Constitution and available in geocoded format. 7The new Constitution was adopted after intense electoral violence in 2007-2008, and contains several features that create a new context for policecitizen interactions.One important element is the devolution of power to 47 Counties.Additionally, the Constitution introduced important changes to the structure of the police service, and ushered police reforms intended to alter the way that police and security forces interact with citizens and to improve public perceptions of the police (Hope 2015, Skilling 2016, Lid and Okwany 2020).

Dependent variable
Our main outcome variable of interest is trust in the police, which is measured based on the following question posed to survey respondents: 'How much do you trust each of the following, or haven't you heard enough about them to say: The Police?' (The question is posed in the context of a series of questions probing the degree of trust towards different institutions, governance structures and political parties).Respondents could choose to answer 'Not at all', 'Just a little', 'Somewhat', 'A lot', or 'Don't know/Haven't heard enough' (they could also refrain from answering).Based on this question, we construct a scale variable (the degree of trust) as well as dummy variables denoting different degrees of (mis)trust.
Kenya represents a context where levels of trust in the police are generally low.Of 34 African countries covered in round 8 of the Afrobarometer, Kenya was ranked on 23rd place in terms of levels of trust in the police based on aggregated results: only 38% reported that they trusted the police at least 'Somewhat' (Logan et al. 2022: 12). Figure 1 shows general levels of public trust in Kenya over the four Afrobarometer survey rounds we rely on, from 2011 to 2019.The graph shows low levels of public trust in police in Kenya, but also that trust levels have fluctuated somewhat over time. 8When aggregating the two categories of trust in police 'Not at all' and 'Just a little' into one, the pattern is consistent: for all four survey rounds, more than 60% of the population surveyed report no or just a little trust in the police.

Main independent variable
To assess if there are differences between trust in the police in urban and rural areas of Kenya, we rely on the binary measure provided in the Afrobarometer data which denotes whether the survey was conducted in an urban or rural area.The urban-rural indicator in Afrobarometer data is based on each country's official census classification of enumeration areas, which forms a basis for the sampling frame (Harding 2020: 38).Mirroring Kenya's relatively low level of urbanisation, around 1/3 of interviews in each survey round were conducted in urban areas.Survey locations coded as urban include neighbourhoods of major cities like Nairobi and Mombasa, but also smaller urban centres like Githurai (Kiambu County), Naivasha (Nakuru County), and Mandera (Mandera County) (Wiesmann et al. 2016).Areas on the 'urban fringe' of larger cities are coded as urban, including suburbs like Mararo, Ruiru and Kitengela on the outskirts of Nairobi.While the census classification is based on population density, it is acknowledged that some areas coded as urban display more rural characteristics (Wiesmann et al. 2016: 18).

Control variables
Our theoretical argument recognises the highly politicised role of the police in Kenya.Under successive regimes, the police have been used as an instrument to repress dissent and protect the incumbent (Ruteere 2011, Hassan 2020).A first factor we control for is therefore the political alignment of the respondent, as we have expectations that opposition-aligned citizens will have lower trust in the police (and other state institutions), all else equal.In addition, opposition support tends to be stronger in African cities, compared to in rural areas (Resnick 2011, Harding 2020) which could confound the results for our main hypothesis.We use the following question from the Afrobarometer to capture if the respondent is opposition or incumbent aligned: 'If a presidential election were held tomorrow, which party's candidate would you vote for?'We create the dummy variable incumbent aligned which takes the value 1 if the respondent listed the ruling party or a party in the ruling coalition, and 0 if they indicated other parties. 9The descriptive statistics (Table 1) indicate that across rounds, 4335 people answered they would vote for the ruling party or coalition, and 2513 that they would vote for an opposition party.There is a high amount of missing data for this variable (people who refused to answer or responded 'don't know').In combination with the very even election outcomes during the period under study, the large share of missing data suggests that opposition supporters may be less likely to answer this question.As a robustness check, we therefore re-run our main models keeping all non-answers in the 0 category (see Appendix 2).
We also control for a number of individual-level factors which previous research has shown to influence the degree of trust in the police: gender, education level, and socioeconomic status.All are based on the Afrobarometer data.For socioeconomic status, we use the following question: 'Over the past year, how often, if ever, have you or anyone in your family: Gone without a cash income?'In addition, research indicates that exposure to violence affects trust in the police, and violence patterns vary sub-nationally in Kenya which could confound our results.We control for the exposure to violence by leveraging the geocoded Afrobarometer data where the location of each interview is coded at town or village level (BenYishay et al. 2017).We match the survey data to georeferenced data from the Armed Conflict Location & Event Dataset (ACLED) (Raleigh et al. 2010) and control for the presence (dummy variable) and intensity (scale) of violence against civilians in the area wherein the respondent is living, in the years preceding the survey. 10To assess the robustness of our results, we include additional variables from the Afrobarometer data capturing trust in the local government authorities, the army, and perceptions about police corruption.Finally, to account for the fact that individuals living in the same geographical area are not independent from each other, we include county-clustered standard errors (models 1 and 4-5).Similarly, to account for the fact that there may be time-invariant county-specific confounders in the data, we include county fixed effects in all models.

Results
We test our theoretical expectations in a series of OLS regression models.Our results are reported in Table 2 below and provide strong support for our hypothesis that individuals living in urban areas hold lower levels of trust in the police than those living in rural areas.Across models, our 'urban' dummy variable is negatively associated with trust in the police, and this association is consistently statistically significant.These results indicate that, all else equal, respondents in urban areas hold lower levels of trust toward the police.In model 1, where we include no additional control variables, we can see that respondents in urban areas on average rated their trust in the police around 0.17 points lower on a scale from 0 to 3, compared to rural respondents.
Our results for our main independent variable hold when controlling for a set of potentially confounding factors.Importantly, the association remains statistically significant when controlling for the political alignment of the respondent, which we introduce as a control variable in models 2-5.As expected, respondents who stated that they would vote for the incumbent party also held higher levels of trust in the police, but the correlation between urban residence and lower trust in police remains statistically significant when this variable is included.This indicates that the urban-rural divide in police trust cannot be explained by subnational variation in support for the incumbent or the opposition.The association is also robust to inclusion of control variables such as gender, education level, and poverty level, included in models 2-5.In line with previous studies (Brown and Benedict 2002), higher education levels are associated with lower trust in the police, and poverty is correlated with lower trust in police.The results are robust to controlling for local exposure to violence in models 3 and 5. Substituting the dummy variable with a continuous measure of the amount of such violence (reported in Appendix 2) does not change the results.Our results are also robust to inclusion of round fixed effects, which accounts for the fact that there may be unobserved year-specific factors that affect individuals' trust in police.The coefficients (reported in Appendix 2) indicate that compared to round 5 (2011), respondents reported higher trust in the police in the later rounds 6 (2014), 7 (2016) and 8 (2019).This could be a sign that police reforms in that period helped improve the public's trust in police, but it could also reflect recovery from very low levels of trust following the 2007-2008 post-election violence.Finally, we noted above that there is reason to suspect that opposition supporters may be less likely to answer the question about political alignment truthfully, which could bias our results.However, the results are robust to including an alternative control for incumbent alignment, where all non-replies are included in the model and coded as not incumbent aligned (reported in Appendix 2).

Discussion and extended analysis
There are important contextual factors to consider when interpreting these results.First, security challenges related to attacks by al-Shabaab and other transnational terrorist groups during the period under study likely affected policing in Kenya and public trust in the security forces.Cities are prime targets of terrorist attacks, which means that policing in urban areas may have been more affected by this security threat.However, militant activity by al-Shabaab as well as by cattle rustlers and other armed groups have to a large extent affected rural areas (Mkutu 2008, Greiner 2013, Lind et al. 2017).To probe whether our results are driven by different experiences and attitudes vis-à-vis the security forces in a broader sense, we include a variable denoting trust in the army in models 2 and 5.We find that higher levels of trust in the army is associated with higher trust in the police, but the results for our main independent variableurban vs rural residenceremains statistically significant.
Second, we study a time period during which significant reforms took place in the organisation and role of the police.Kenya adopted a new Constitution in 2010, which has entailed substantial devolution of power as well as police reform.The former could mean that political relationships at the local level may become relatively more important in mediating attitudes toward the police.While we control for the effect of survey round, reforms have been uneven and the pace of reform could be systematically different in urban and rural areas.Police reforms have taken place simultaneously with decentralisation to Kenya's 47 Counties, which introduces additional possible sub-national variation in political dynamics surrounding the police and civilian oversight.In models 3 and 5, we also include a variable denoting trust in the County government.This variable indicates that higher levels of trust in the County government is associated with higher trust in the police.Again, however, our main results are robust to inclusion of this variable: the association between urban residence and attitudes to police is not driven by trust in county-level politicians.In addition, when we control for whether respondents perceive the police as corrupt (models 3 and 5), our main independent variable remains statistically significant.In line with previous research, perceiving the police as corrupt correlates with lower levels of trust in the police.Taken together, our results are robust to the additional variables included in models 3 and 5, strengthening our confidence that what we find is not due to people who reside in rural areas being more trusting in general, or that low police trust is only explained by perceptions of them as being corrupt.
In sum, our analysis indicates a robust and negative relationship between urban residence and levels of trust in the police.This is in line with our hypothesis, and our broader theoretical argument that the rural and urban environment create different conditions for police-citizen relations, leading to different levels of trust.Specifically, we theorised that trust towards police would be lower in urban areas, because urban residents are more likely to have firstor second-hand experiences that lead them to conclude that policing is not fair.
Our quantitative analysis of observational data does not allow us to directly test the proposed causal mechanism connecting trust in police and urban-rural residence.However, our argument that the rural and urban environment create different conditions for police-citizen relations is supported by qualitative evidence from interviews we have conducted with different stakeholders in Kenya.One government official with experience of both rural and urban areas reflects: The local set up in a rural area is that you will have more homogenous groups.They can more easily agree due to similarities.It also makes policing work easier.Communities even help in policing, which makes it easier to track criminals there than in urban areas.Some like the pastoralists also have social sanctions.In towns, because of the nature of the population, organizing them is more difficult and they are more secretive.In the rural, people know everyone.In rural areas there is more natural respect for the police.They are seen as honest persons.Here [in the city] they profile the police and are acting from a defensive position. 11  In line with our expectation that police-citizen interactions are overall more visible and frequent in cities than in rural areas, another Kenyan government official noted that in his previous deployment in a rural area, one police station served a vast area with 100,000 residents, whereas in his current urban deployment there are four police stations in less than four kilometers. 12 Our argument emphasised fairness, but the results could also be reflective of different perceptions of police effectiveness.In Ghana, Tankebe (2009) found that police effectiveness is a more important predictor of public cooperation with police than perceptions related to fairness.He argued that in environments with poor police performance and a legacy of colonial policing, the perceived legitimacy of the police is so low that a minimum level of effectiveness matters more than fairness (Tankebe 2009).Our interviews in Kenya suggest that perceived improved efficiencyfor instance, frequent patrols in areas at risk of insecurity during the 2022 electionsimproved locals' trust in the police. 13Overall, our data does not allow us to adjudicate the relative role that fairness and effectiveness play in shaping Kenyans' trust in the police.However, the government official cited above suggests that in his experience, police can operate more effectively in rural areas.On the other hand, interviews with urban residents in Nakuru and Naivasha indicate that local police officers are often suspected to collaborate with criminal gangs. 14Such perceptions are likely lower in rural areas, where criminal networks are less present.
Relatedly, residents in rural and urban areas are prone to interact with different branches of the police.During colonial rule, separate policing structures were set up for policing the cities (where European settlers lived, and Kenyans were only allowed transitory residence) and the rural 'native areas'; these parallel structures were retained after independence (Ruteere 2011: 13).While formally reintegrated as parts of the National Police Service (Government of Kenya, 2012), the branches retain partly different mandates and citizens may hold different degrees of trust toward them.Notably, the Kenya Police Service has a key responsibility for crime prevention and reduction (Government of Kenya, 2012: Part III, section 24), while the Administration Police is tasked with border security, preventing stock theft, and 'coordinating and complementing Government agencies in conflict management and peace building' (Government of Kenya, 2012: Part IV, section 27).These different tasks, and the public's assessment of them, are likely to shape their overall trust in the police.However, the provincial administration has historically been a key tool for rulers to 'suppress regime opponents, rig elections, and control civilian protests throughout the country' (Hassan, 2015: 588) and there is a history of presidential control of the Administrative Police.Thus, the Administrative Police, more visible in rural areas, would arguably be subject to higher mistrust, in contrast to our findings.
Finally, our theoretical argument is based on the premise that in a setting of repressive policing and low overall trust, interactions with the police will generally have a negative rather than positive effect on citizens' attitudes.This argument is plausible in a context where 'the culture of impunity in the police service has contributed to too many cases of insecurity, gross violation of human rights, mistrust by citizens and derailment of key achievements in democratic governance' (Kivoi and Mbae 2013: 189).Ongoing police reforms under the 2010 Constitution are aimed at increasing accountability and oversight.There are mixed results as to whether these reforms increase public trust in police.Interviews conducted in 2022 indicated that whereas building trust is a slow process, citizens no longer fear the police in the same way. 15One woman working in an urban area with high poverty and crime noted: 'Wethe massesdo not fear the police, but we also do not respect them.In the past when we saw a person in uniform we would turn and go the other way if we were in the right or in the wrong.Now we see them as human beings'. 16Discouragingly, however, a report based on the most recent Afrobarometer survey, 17 indicates the lowest levels of aggregate trust since 2014, with a full 42% responding they do not trust the police at all (Kamau et al. 2022).

Conclusion
A functioning and legitimate police force is essential for ensuring the safety of the public.In addition to dealing with crime, the police also serve a conflict preventing and dispute resolution function in countries like Kenya.Enjoying the trust of the citizens they are intended to serve is vital for the police to perform these tasks effectively.Conversely, mistrust in the policein Kenya and elsewhereindicates that ordinary people do not see themselves as enjoying full security, or that they even fear being ill-treated in their interactions with police.
Our study suggests that an important contextual factor affecting citizens' trust in the police is whether they reside in a primarily urban or rural settling.Our empirical analysis provides evidence that urban residents in Kenya have lower levels of trust in police than rural residents.While other factors commonly associated with trust in state institutions, such as political partisanship, also matter, urban or rural residence has a robust independent effect on police trust.This urban-rural divide in attitudes towards the police, we argue, can be linked to differences in how residents in these two settings are expected to interact with the police and the type of perceptions they gain based on first-hand and second-hand information about police conduct.
Our findings are relevant for understanding police-citizen relations in contexts with similar historical legacies and insecurity, and where police brutality is a common feature of informal urban settlements.They have important implications for reforms seeking to improve police conduct and policecommunity relationship in Kenya and beyond.On the one hand, the case could be made that efforts to improve police-community relations should be concentrated to informal settlements in urban areas, where police have been largely ineffective in curbing high rates of crime and insecurity and where police harassments and killings have undermined citizen trust in the police (Kimari 2017, van Stapele 2020, Wairuri 2022).The pervasive trend of urbanisation across the continent suggests that police mistrust will only increase as urban centres grow and the police operate in an ever more challenging context.On the other hand, in Kenya and many other African countries a large part of the population still resides in non-metropolitan areas and rural considerations shape social, economic, and political development in important ways.For this reason, the policy and scholarly community should retain focus also on rural settings.Research on attitudes towards the police and police performance generally suffers from an urban bias, and more effort is needed to uncover the dynamics that unfold outside of the cities.For example, research suggests that expectations of police to be more problem-oriented and also address non-criminal issues are greater in rural areas (Jiao 2001).Patterns of police performance as well as perceptions of the police are likely to differ between rural and urban areas, but also expectations about the core tasks the police is intended to fill.
Research also needs to consider other important subnational variations and how they intersect with the urban and rural environments the police operate in.Notably, countries affected by violent conflict experience very different security challenges depending on the type and intensity of the violence, and local manifestations of violence will play out differently in urban or rural areas.For example, scholars argue that urban violence and conflict is more nebulous in character than rural violence (Moser and McIlwaine 2014), and state-insurgent relations are fundamentally