Policing nightlife areas: comparing youths’ trust in police, door staff and CCTV

ABSTRACT

Against a background of the pluralisation of policing in contemporary city spaces, and sustained interests in the assessment of policing in the criminology and criminal justice literatures, the current study seeks to draw a comparative analysis in trust between policing actors, as experienced by nightlife consumers. While studies on trust in the police are numerous, this is much less the case for other actors involved in policing urban (nightlife) spaces. Neither is it very well understood how trust is distributed between policing actors. It is important to investigate this, taking into consideration the privatisation and technologisation of safety provision in contemporary cities, and the legitimacy of the actors involved. Using a survey, 894 youths enrolled in education were asked to evaluate their trust in actors involved in the policing of urban nightlife areas: the police, door staff, and CCTV. Results showed that people tend to trust human policing agents more compared to technological agents. A cluster analysis further indicated that alongside this general pattern, four additional groups can be found in the data: two groups that display the highest trust in either the police or door staff with intermediate trust in CCTV, and two groups expressing either overall low trust or overall high trust, independent of the policing actor. Employing logistic regression analyses, we find that demographic, victimisation, and contextual variables predict cluster membership. We end with suggestions for future research and reflect on whether the privatisation and technologisation of (nightlife) policing are desirable from a nightlife consumer point of view.

KEYWORDS:

• Trust
• confidence
• policing
• CCTV

Introduction

There has been a substantial and sustained interest in the interactions between police and the general public in criminological literatures. Part of these literatures has inquired the degree to which the general public assess, are confident of, and express trust in the police and their activities (e.g. Jackson and Bradford 2010). Markedly, however, interests in trust in other surveillance and policing actors have received less attention (O’Neill and Fyfe 2017). This is especially noteworthy considering that actors involved in the surveillance and policing of urban areas are increasing in number and diversity, as emphasised in the literatures on plural policing (Crawford et al. 2005, Jones and Newburn 2006, O’Neill and Fyfe 2017). Indeed, policing as a ‘pluralized, fragmented and differentiated patchwork has replaced the idea of the police as the monopolistic guardians of public order’ (Crawford 2003, p. 136). Apart from the police as a state actor, private security actors and policing technologies are (increasingly) involved in the provision of security in urban areas (Newburn and Jones 2006, Yarwood 2007). The presence of private security and policing technologies in (semi-)private spaces such as shopping malls and leisure complexes has been studied for quite some time now (Shearing and Stenning 1983, Button 2003), but more recently the discussion increasingly includes their presence in urban public spaces (Newburn 2001, Norris 2012, Germain et al. 2013, Boels and Verhage 2016).

The pluralisation of policing can be understood against the background of several general trends in late modern societies. Jones and Newburn (2006, p. 7) mention the rise of ‘mass private property’ (such as shopping centres and (semi)private residential developments), which are often privately owned and policed. Scholars also point at the apparent increases in public experiences of uncertainty, insecurity and anxiety, as characteristic to late modern societies, resulting in increasing (and often perceived as unmet) demands for security by citizens (Loader 1997, Crawford et al. 2005, Jones and Newburn 2006). Being at the crossroads of such escalating demands for security, as well as constraints experienced on public police expenditure by police forces (Jones and Newburn 2006), and a realisation among policymakers and senior police officers of ‘the limits of what the police alone can do to prevent crime’ (Loader 1997, p. 145), one outcome has been the adoption of, and outsourcing to, forms of private or commercial policing (Loader 1997, Livingstone and Hart 2003). Another is the rise of closed-circuit television (CCTV) as a means to prevent crime and meet the increasing demands for security by citizens, sometimes following heavy state sponsoring as (initially) in the UK (McCahill, 2008; Norris, 2012). Although estimates of the number of CCTV cameras in urban public spaces vary, scholars tend to agree that the instrument has become more-or-less a standard feature to urban life (Norris, 2012, Germain et al. 2013). Or as Norris puts it, ‘[i]n less than two decades, it has expanded from a local initiative in a few small towns in the UK (…) to penetrate every major city, in every country, on every continent’ (2012, p. 254).

Hence, by some once seen as especially a job for the police, the authority to ‘do’ surveillance and policing in city spaces is by now considered to be distributed amongst a variety of public, private, and technological policing actors. Taking into consideration the pluralisation of policing in urban spaces (Newburn 2001, Jones et al. 2009), it is important to consider the ways people perceive and experience the various actors involved in security provision. Assessments, attitudes, perceptions, and sentiments regarding policing actors can shape people’s willingness to accept the presence of these actors and to cooperate with them. That, in turn, determines the policing actor’s degree or ability to exercise control (Tyler 2001, Livingstone and Hart 2003, Jackson and Bradford 2010). More specifically, that those in the community being regulated believe that their authorities deserve to rule and make decisions that influence the outcomes of members of the community (also known as having legitimacy; Kelman and Hamilton 1989, Tyler et al. 2015).

Indeed, O’Neill and Fyfe (2017, p. 5) state that ‘[i]mportant conceptual and empirical questions include understanding levels of public acceptance for plural policing, [and] which types of security actors are perceived as legitimate (…)’. In response to this observation, then, and against the background of a shift in ‘the responsibility of policing from the state to an ever wider assortment of public, private and voluntary agencies’ (Yarwood 2007, p. 447) more generally, the current study seeks to investigate public perceptions of various policing actors. More specifically, the current paper compares people’s trust in the police, private security, and closed-circuit television (CCTV) as policing actors. To the best of our knowledge, research that empirically approaches a trust comparison of these policing actors is absent.

We root this comparison of trust in policing actors in an urban nightlife context, as ‘[p]erceptions of the “hours of darkness” as a time of danger, fear, crime and sin seem to be persistent and deeply embedded’ in Western culture (Hobbs et al. 2003, p. 44, see also Edensor 2012). Beyond this general observation, urban nightlife is consistently problematised in the media and public discourse as cause of crime and disorder, among other things because of (excessive) consumption of alcohol and drugs (Hadfield et al. 2009, Shaw 2010). As a consequence, urban nightlife areas are often regarded space-times in need of (extensive) surveillance and policing, to which end we also observe a clear trend in the outsourcing of control in this specific context (Wadds 2013, Hadfield and Measham 2015). At the same time, the literatures on urban nightlife emphasise that such ‘social disorders’ may elicit different experiences amongst different publics; were some consider urban nightlife areas as disorderly and dangerous and indeed in need of greater surveillance and policing, they are experienced as fun and adventurous by others (Hadfield et al. 2009). As a consequence, the trust these different publics report in policing actors might be equally diverse. Given that policing actors’ ability to exercise control is grounded in perceptions of trust held by the general public, urban nightlife is therefore deemed an especially relevant setting to investigate the distribution of trust among surveillance and policing actors.

The current paper draws on the results of a survey distributed among Dutch youth enrolled in education, who were asked a series of questions about their trust in the police, (publicly installed) CCTV, and door staff. We chose to focus on youth as they – and students especially – participate in nightlife activities most frequently (Chatterton and Hollands 2002; Schwanen et al. 2012). On a more general level, we consider our focus on youth within a nightlife context one additional contribution to the literatures, which tends to focus on trust in policing actors at daytime, and among adults.

The public’s trust in policing actors

In the criminological field of study, and particularly in studies on policing, trust is an intensively researched topic. Especially in studies related to the police, an in-depth discussion exists on the conceptualisation and measurement of trust. Building on this literature, complemented with insights from studies on trust in private security and CCTV, we discuss the concept of trust in the first part of this section. The second part describes studies investigating trust in these policing actors. This is followed by a third part that highlights important determinants of trust as reported in the literatures.

Explaining the concept of trust

Peoples assessment of the police are measured with a wide and complex range of concepts (e.g. Brown and Reed Benedict 2002, Sun et al. 2013), many of which are at least partially overlapping. Following Sun et al. (2013, p. 645), these may roughly be divided into three groups. A first group consists of quite broad and neutral concepts ‘to describe the public’s general judgements and sentiments towards the police’ (e.g. attitude, perception, view, opinion, support, p. 645). On the other end of the spectrum, quite narrow concepts are utilised that ‘tap into perceptions of specific aspects of police performance and behaviors, such as respectfulness, fairness, effectiveness, shared values, priorities and integrity’ (p. 645). However, most studies, including the current, focus on a third group of intermediate ‘perceptual or attitudinal constructs’ (p. 645), including whether people have trust. But what exactly is meant by trust?

There is quite some discussion about the constitution of the concept ‘trust’. For example, Kääriäinen (2007) and Luhmann (2000) view trust as grounded in personal experiences and (face-to-face) encounters: we trust those people who in personal interaction have proved worthy of our trust. It thus seems that this notion of trust is closely related to the gist of the encounter. In the literatures this is also referred to as (encounter-based) interpersonal trust (Jackson and Bradford 2010). However, other scholars interpret the concept of trust beyond the encounter or a specific interaction partner. In such cases, trust is referred to as institutional trust (Jackson and Bradford 2010). This concept of trust involves how the system acts in general, which is more personally remote, and reflects what could be called a ‘job rating’ of the crime control system (Saarikkomäki 2018). Or as Jackson et al. (2011) define trust (in the police): not only the public belief that the police have the right intentions towards citizens, but also that they are competent to act in specific ways in specific situations (see also Hardin 2002). In various literatures such performance-focused, institutional trust is interchangeably referred to as confidence (e.g. Sindall et al. 2012).

Reported trust in the police, private security, and CCTV: a broad overview

The literatures tend to indicate that trust in the police in (North-Western) European countries can be considered sizeable (Blankenburg and Bruinsma 1994, Blankenburg 1998, Kääriäinen, 2007). Using a single item indicator, Kääriäinen (2007) studied trust in the police between 16 European countries. Highest trust is reported in the Nordic countries (approximately 8 on a scale ranging from 1 to 10), whereas countries in Eastern Europe tend to score somewhat lower. The Netherlands (on which the current study focuses) is however not part of this study. Based on the Eurobarometer Survey 2001, Hudson (2006) reports that 71.48% of the Dutch trust the police, scoring somewhat above average (69.13%, 15 countries). The Dutch Central Bureau of Statistics also measures trust in the police among the Dutch on a yearly basis. Measured on a 10-point scale (1 being very low trust, 10 being very high trust), they report a gradual increase from 6.1 in 2012 to 6.5 in 2017 (Centraal Bureau voor de Statistiek 2018). Although a large amount of studies use the single item measure of trust from the European Social Survey, studies measuring trust in European countries using multiple items have shown similar patterns: trust in the police is moderate to high (e.g. Bradford et al. 2009, van Damme 2017).

While (inter)national (comparative) statistics are widely available on the public’s reported trust in the police, less headway is made to study public assessment of private security – including door staff in urban nightlife areas. Moreira et al. (2015, p. 209) state that ‘we know little about how citizens view private security guards and what factors influence their trust in and satisfaction with private security guard services’. This is surprising, given that private security actors have repeatedly partnered up with the police (White 2014). Studies do note that private security ‘currently confront what might turn out to be a significant problem of trust’ (Loader, 1997, p. 152) resulting from the unregulated nature of the industry. Indeed, Livingstone and Hart (2003) mention that while the image of the police has for long been quite positive and ‘continues to epitomise legitimacy, stability and continuity’ (p. 162), greater scepticism seems to surround the image of private security. At the same time, the literatures indicate that a professionalisation offensive is underway in the private security sector, confronting the circulation of these images. Thumala et al. (2011) argue that the private security sector has invested in regulation, education, and licensing.

Among the first available studies that specifically and empirically study perceptions of private security are Nalla and Hereux (2003, p. 244), who report an overall ‘positive perception of security officers’ among college students. In part building on this study, van Steden and Nalla (2010) and van Steden et al. (2009) report a moderate level of satisfaction with security guards (about 2.8 out of a possible 1–5). Furthermore, Moreira et al. (2015, p. 209) show that 24.1% of their research participants (strongly) agree with the proposition: ‘Citizens can generally trust security guards to protect their lives and properties’ (trust), and 8.6% with the proposition: ‘Generally, I am satisfied with the way security guards conduct themselves’ (satisfaction). Still, 35.2% neither agree nor disagree with the first proposition, 30.2% with the second. Hence, the relatively scarce amount of research on private security seems to show that people are more or less neutral when it comes to trusting this particular policing actor. Given the degree of investment in regulation, education, and licensing by the private sector, as noted by Thumala et al. (2011), and their cooperation with the police, these levels of trust may well increase in the coming years.

Looking at research into CCTV, numerous studies have investigated the degree to which the public supports (the installation of) CCTV. When CCTV was introduced to the public on a larger scale, support tended to be substantial (see Gill et al. 2007). Spriggs et al. (2005), for instance, report that 82% of their research participants were happy with the prospect of the installation of a new CCTV system. The installation of CCTV was welcomed by around 90% of the sample drawn by Honess and Charman (1992). Reporting somewhat lower scores, in Bennett and Gelsthorpe (1996, p. 77) 64.2% of the research participants thought CCTV was a (fairly to very) good idea. However, (criminological) studies explicitly discussing trust in CCTV systems are scant, which according to Ellis et al. (2013, p. 2) is ‘astonishing’ since ‘trust is clearly one of the central components of surveillance systems’.

It thus seems that there was generally high support for the implementation of publicly installed CCTV surveillance amongst citizens at the time of, and following, its introduction. However, Ditton (1998) has underlined that survey design may also play an important role in explaining the high support figures. Others have argued that high (initial) support might in large part be linked to (politically instigated) general perceptions of CCTV as an effective instrument to control crime and disorder (Webster 2009). However, studies have called into question such substantial support figures, sometimes reporting them as overstated (Ditton 1998, Gill et al. 2007, Webster 2009). As Webster (2009, p. 18) states: ‘[p]resumably, as time passes and greater awareness of the limitations and implications of CCTV use becomes common knowledge public support will diminish’. Put differently, high support in studies may be an artefact of the cultural and political framing that crime reduction follows CCTV provision (Webster 2009). Sentiments that publicly installed CCTV is not, or only in part ‘keeping up to its promise’ are thus gradually more reported in the literatures.

In fact, the few studies that go beyond the seemingly positive picture painted by general polls and surveys on public support, seem to suggest that people’s perceptions of CCTV may also proliferate distrust in terms of the system being (in)capable to prevent crime (Ellis et al. 2013). As argued by Neyland (2006, p. 10): ‘trust in CCTV may involve the drawing together of multiple claims, based on multiple forms of information, forming ongoing assessments and decisions regarding what CCTV is doing, what it could do and whether or not it works’. Indeed, a recent study drawing on quantitative evidence from Thailand, reports that levels of institutional trust in CCTV are actually moderate to low (Trimek 2016).

In Ellis et al. (2013), research participants also raised concerns about the ways CCTV c/would be used for generating income through fining minor violations, and that the extensive power asymmetry between watcher and watched may be an important source of distrust in CCTV (see also Koskela 2002). In a more general sense this signals that public discussions of trust in CCTV are also closely related to issues of privacy (Neyland 2006). Trimek (2016) also assessed perceptions of rightful use of CCTV (footage), again finding moderate to low trust. It should be noted though, that these studies focused on the implementation of CCTV by, for example, the government. Sentiments might be different when it concerns CCTV being implemented by people themselves for monitoring places and property (see Mäkinen 2017). All in all, then, the above indicates that discussions on trust do proliferate in studies that take interest in the practice and governance of safety through CCTV surveillance. Still, limited effort has been made to start investigating this in a more systematic manner.

Finally, to our current knowledge, only Saarikkomäki (2018) empirically compares people’s trust between policing actors. Drawing on qualitative interviews with Helsinki youth, Saarikkomäki (2018, p. 6), reports that ‘[t]he police were typically described as more friendly, predictable, humorous and as acting in a more professional manner than the security guards’. The study also reports that research participants (and ethnic minorities in particular) questioned the neutrality of security guards more compared to police officers. Moreover, the study mentions that younger people may receive selective treatment more often by security personnel compared to police officers. Yet, the study also reports positive encounters with security guards, and negative encounters with police, nuancing the above. Interestingly, Saarikkomäki (2018, p. 9) mentions that ‘[t]ypically, participants stated they had general confidence in both police and security guards. However, a few participants had very low confidence in both of these policing agents’. This could of course mean that some have negative perceptions of those two agents of control independently; however, it could also mean that these negative perceptions reflect a more general negative sentiment about surveillance and control in society.

Determinants of trust

As trust is an intensively researched topic with regard to the police, but less with regard to private security and CCTV, it is not surprising that determinants of trust have mainly been discussed within the police literatures. For example, there is some consensus in the literatures on trust in the police that younger people, people with a non-western background, and people with lower income tend to be less satisfied with the police (e.g. Brown and Reed Benedict 2002, Ren et al. 2005, Kääriäinen 2007). Studies looking at the role of gender and the degree of urbanisation show mixed results when it comes to trust in the police (Ren et al. 2005).

Another determinant linked to trust, identified in the literature on police, is victimisation. Ren et al. (2005) showed that previous victimisation (whether one had been victim of any crime in the last twelve months) reduced confidence in the police, while voluntary contacts with the police increase confidence in the police. That negative police-citizen contacts affect judgments about the police has also been confirmed by other studies (Brown and Reed Benedict 2002, Skogan 2005, Kääriäinen 2007, Wells 2007), however, seems to have received less attention in studies into private security and CCTV trust. Relatedly, Bradford et al. (2009, p. 20) argue that ‘seeing regular police patrols and feeling informed about police activities are associated with higher opinions of effectiveness and community engagement’. From this it could be expected that people living in a large municipality, who are more frequently exposed to intensively policed places, might also in general value the presence of policing actors more (Steenbekkers et al. 2006). The same might hold for people who more frequently visit intensively policed space-times, such as urban nightlife areas.

Whether these determinants, as highlighted by the literature on trust in the police, also hold for trust in door staff and CCTV is unclear. Put differently, whether certain determinants transcend the type of policing actor is a question that remains to be answered in the current research.

Towards the empirical analysis

Based on previous studies we predict that, compared to private security (in this study, specifically pertaining to door staff) and CCTV, people have highest trust in the police. Research indicates that, overall, trust in the police tends to be rather high in (North-Western) Europe, and in their study Nalla and Hereux (2003) remark that it might be expected that ‘the public would have a slightly more negative perception of private security than of public policing’ (p. 238). A synthesis of the available literature further leads us to believe that a slightly more negative sentiment would exist for trust in CCTV as compared to both door staff and the police.

We further seek to perform a follow-up analysis to investigate whether multiple distributions of trust between policing actors exists (see Saarikkomäki 2018). Finally, we address if determinants for trust in the police also hold for other policing actors, and whether these predict people’s distribution of trust in policing actors. Below, then, we first address our methodological approach through which we seek to test our predictions, followed by a description of our results.

Method

Participants

The current study draws on questions embedded in a larger survey on nightlife consumption and safety experience among youth following ((public) secondary and tertiary), education in the Dutch cities of Utrecht and Rotterdam (but who were not necessarily living in these cities). The survey was part of a research project into experienced safety, surveillance and policing in the public spaces of these cities’ nightlife areas. In the current study, however, we do not differentiate between the two cities as such as we have found no theoretical grounds in our previous section that suggests this would be important.

We approached our participants through their educational institutions which we requested to promote our online survey by placing it on their social media channel(s) and/or website, online learning environments, by handing out leaflets in class and/or through direct mailings to students. We are aware of the fact that people enrolled in education do not necessarily generalise to the crowd that frequents nightlife areas and establishments. Nevertheless, youth and especially students tend to be overrepresented in urban nightlife areas (Chatterton and Hollands 2002; Schwanen et al. 2012), and for this reason are likely to have some experience with policing practices in these space-times.

A total of 1457 persons followed the hyperlink to the online survey; 894 persons completed it, constituting our sample. This sample represents both genders well (57% female). The mean age was 20.29 years (SD = 3.29). Taking into consideration that the minimum drinking age at the time of study was 16 years and older, students below this age were excluded from this study for ethical reasons, using a filter question (‘are you at least 16 years of age’). Fifty-two per cent of the participants live in a large Dutch municipality (≥100,000 inhabitants; Platform 31 2018), 48% in a small(er) Dutch municipality. Seventy-nine per cent of the participants did not have a migration background (i.e. father’s and mother’s country of birth was the Netherlands), whereas 21% father’s and/or mother’s country of birth was a country other than the Netherlands. Also, 52% of the participants said that they had been previously victimised in a nightlife context. That is, whether they had experienced an incident during a night out the last three years, with an incident being one of the following things: catcalling, staring, intimidation, scolding, being followed/watched, street brawling, or unwanted intimacies. Finally, in terms of frequency of going out, 38% indicated to go out about once a month or less; 27% about once every two weeks; and 35% about once a week or more. There were 67 people who indicated to never go out in general. Hence, they do not have a response on this variable, leaving a total of 827 valid responses for this variable.

Measuring trust

The primary objective of the current study is to examine patterns in terms of trust between various policing actors. Drawing specifically on the nightlife context, we focused on police, door staff and CCTV. Participants were asked (1) ‘In general I have a large degree of trust in [policing actor]’, and (2) ‘I have a large degree of trust in [policing actor] when they intervene during incidents in the nightlife area’, on a scale ranging from 1 (not at all) to 7 (very much). For CCTV the second question was framed slightly different taking into consideration CCTV itself cannot intervene (directly) with an incident taking place; ‘I have a large degree of trust in CCTV when I see an incident in the nightlife area’. We made composite scores of the two trust items for the police (Cronbach’s α = .88), door staff (Cronbach’s α = .84), and CCTV (Cronbach’s α = .91).

Analytic strategy

We seek to realise our aim in three steps. We firstly looked at the general pattern of trust in the policing actors, and compared these by means of paired samples t-tests. Instead of listing all separate values from the paired samples t-tests, we chose to summarise the results by reporting the lowest significant t and its accompanying p (see below). We then performed a Two-Step Cluster analysis to investigate whether, next to the general pattern of trust drawn from step 1, other trust patterns could be detected. In this cluster analysis, individual participants were selected as cases; the clustering variables were the trust composite scores introduced earlier. The log-likelihood distance measure was used, rendering a meaningful five-cluster solution using the auto-cluster option provided by the SPSS software package. Because activating the outlier handling option resulted in the detection of only one outlier, we ran our final solution without outlier-handling (Mooi and Sarsted 2011, Norusis 2011). In the final step, we performed logistic regression analyses to investigate, for each cluster separately, what group characteristics (independent variables) predict cluster membership (dependent variable).

Results

Comparing trust in nightlife policing actors

In general, our research participants report highest trust in the police (M = 4.53, SD = 1.43), followed by door staff (M = 4.34, SD = 1.49), and CCTV (M = 3.23, SD = 1.41), all t’s > 3.40, all p’s < .001. The distribution of the trust scores is presented in Figure 1. The first column of Figure 1 shows the clustering variables (trust in police, door staff, and CCTV), together with their rating scales (which display incremental steps of 0.5 points, because we used composite scores of two items; see method section). The second column provides a plot of the mean scores on trust in the police, door staff and CCTV. Comparing mean scores with the midpoint of our rating scale, participants can be regarded as positive about their trust in the police and door staff, but negative about their trust in CCTV, as all means differ significantly from the midpoint of the scale (4), all t’s > 6.90, p’s < .001. Higher trust is thus reported for human actors, compared to CCTV surveillance.

Figure 1 General (pre-cluster) trust in policing scores (column 2) and distinctive (post-cluster) types of trust in policing (columns 3–7).

Patterning trust in nightlife policing actors

Cluster analysis

A Two-step cluster analysis indeed shows us that additional, and different trust patterns exist. The five clusters (you may also read ‘groups of individuals’, instead) we find are also represented in Figure 1, columns 3–7, and will be discussed in further detail below. We named the clusters in terms of what we considered to be their defining characteristic, in order to facilitate the reading and understanding of the remainder of this section. From left to right: ‘CCTV adversary’, ‘high trust police’, ‘high trust door staff’, ‘overall limited trust’ and ‘overall high trust’.

Firstly, we find a ‘CCTV adversary’ cluster that is more or less in line with the general trust means reported in the previous section: we detect a pattern where trust in human actors is quite high, whereas trust in CCTV is quite low. Solely looking at the patterns, it seems that trust scores for CCTV among members of this cluster are somewhat lower compared to the average scores reported above (although the cluster analysis does not allow for testing statistical significance).

Secondly, where the means reported in the previous section indicate that people generally trust human actors more than CCTV, two other clusters differentiate individual trust ratings between the police and door staff. Members to the ‘high trust door staff’ cluster report high trust in door staff, but limited to average trust in police. The opposite is true for members to the ‘high trust police’ cluster who report high trust in police, but limited to average trust in door staff. Trust ratings for CCTV are roughly comparable for these two clusters and centre on the general average.

The ‘overall limited trust’ and ‘overall high trust’ groups consist of members who report quite opposing trust ratings across the policing actors. The ‘overall limited trust’ cluster represents individuals who report relatively low to limited trust, independent of the policing actor. This means that individuals member to this clusters are especially more negative about police and door staff, while somewhat more negative in their trust of CCTV. In the ‘overall high trust’ cluster we find a group of persons that report high trust in policing actors more generally. Scores for trust in police and door staff are somewhat more positive. Most interesting, however, is the high degree of trust in CCTV expressed among members to this cluster, which we observe in none of the other clusters reported above.

Logistic regression analyses

Table 1 shows the results from the logistic regression analysis using gender, age, migration background, municipality size, frequency of going out, and previous victimisation as predictors for cluster membership (for each cluster separately). This allows us to investigate the determinants of cluster membership. Or put differently, which groups of individuals are more likely to belong to each cluster.

Table 1. Cluster membership prediction using logistic regression.

Variable	CCTV adversary (n = 186)		High trust police (n = 144)		High trust door staff (n = 174)		Overall limited trust (n = 129)		Overall high trust (n = 186)
	OR (CI 95%)	SE	OR (CI 95%)	SE	OR (CI 95%)	SE	OR (CI 95%)	SE	OR (CI 95%)	SE
Gender
Men	0.68 (0.48/0.97)*	.18	1.12 (0.77/1.64)	.19	0.80 (0.56/1.15)	.18	1.75 (1.18/2.59)**	.20	1.07 (0.76/1.51)	.17
Women	1		1		1		1		1
Age	0.98 (0.92/1.03)	.03	1.14 (1.08/1.21)***	.03	0.88 (0.83/0.94)***	.03	1.01 (0.95/1.07)	.03	0.99 (0.94/1.05)	.03
Migration background
With	0.62 (0.39/0.98)*	.24	0.83 (0.52/1.32)	.24	1.04 (0.67/1.59)	.22	1.97 (1.26/3.08)**	.23	1.02 (0.68/1.55)	.21
Without	1		1		1		1		1
Municipality size
Small	1.27 (0.90/1.80)	.18	0.49 (0.33/0.73)***	.21	1.10 (0.77/1.56)	.18	1.00 (0.67/1.49)	.21	1.25 (0.88/1.76)	.18
Large	1		1		1		1		1
Frequency of going out
About once a month or less	0.99 (0.65/1.52)	.22	0.94 (0.59/1.51)	.24	0.82 (0.54/1.25)	.22	1.00 (0.62/1.60)	.24	1.23 (0.81/1.85)	.21
About once every two weeks	1.39 (0.91/2.11)	.21	1.24 (0.78/1.96)	.21	0.74 (0.47/1.16)	.23	0.77 (0.47/1.27)	.26	0.93 (0.60/1.45)	.23
About once a week or more	1		1		1		1		1
Previous victimisation
Non-victim	0.83 (0.59/1.18)	.18	1.21 (0.82/1.78)	.20	1.12 (0.78/1.60)	.18	0.49 (0.32/0.74)**	.22	1.53 (1.08/2.16)*	.18
Victim	1		1		1		1		1

Notes: OR = odds ratio, 95% CI = confidence interval, SE = standard error. Because the ‘frequency of going out’ variable contained 67 missing values, and the ‘municipality size’ variable contained 5 missing values, the total n of each cluster is lower than the original n.

*p < .05.

**p < .01.

***p < .001.

CCTV adversary. The results show that gender and migration background are significant predictors of ‘CCTV adversary’ membership. Men (OR = 0.68) are less likely to be a member of the ‘CCTV adversary’ cluster than women. As can be seen in Table 2, relatively more women than men are present in the ‘CCTV adversary’ cluster. Also, people with a migration background (OR = 0.62) are less likely to be a member of the ‘CCTV adversary’ cluster than people without a migration background. Relatively less people with a migration background and more without a migration background are present in the ‘CCTV adversary’ cluster.

Table 2. Descriptive statistics on cluster membership.

	CCTV adversary	High trust police	High trust door staff	Overall limited trust	Overall high trust	Total
Gender
Men (% within gender)	67 (17.4%)	77 (20.0%)	76 (19.7%)	75 (19.5%)	90 (23.4%)	385 (100%)
Women (% within gender)	123 (24.2%)	96 (18.9%)	111 (21.8%)	66 (13.0%)	113 (22.2%)	509 (100%)
Total (% within gender)	190 (21.3%)	173 (19.4%)	187 (20.9%)	141 (15.8%)	203 (22.7%)	894 (100%)
Age	M = 20.02 (SD = 2.66)	M = 21.43 (SD = 4.18)	M = 19.59 (SD = 2.70)	M = 20.37 (SD = 3.44)	M = 20.16 (SD = 3.12)
Migration background
With (% within migration background)	28 (14.9%)	38 (20.2%)	37 (19.7%)	40 (21.3%)	45 (23.9%)	188 (100%)
Without (% within migration background)	162 (22.9%)	135 (19.1%)	150 (21.2%)	101 (14.3%)	158 (22.4%)	706 (100%)
Total (% within migration background)	190 (21.3%)	173 (19.4%)	187 (20.9%)	141 (15.8%)	203 (22.7%)	894 (100%)
Municipality size
Small (% within municipality size)	99 (23.2%)	62 (14.6%)	97 (22.8%)	65 (15.3%)	103 (24.2%)	426 (100%)
Large (% within municipality size)	91 (19.7%)	109 (23.5%)	89 (19.2%)	75 (16.2%)	99 (21.4%)	463 (100%)
Total (% within municipality size)	190 (21.4%)	171 (19.2%)	186 (20.9%)	140 (15.7%)	202 (22.7%)	889 (100%)
Frequency of going out
About once a month or less (% within frequency of going out)	65 (20.7%)	49 (15.6%)	70 (22.3%)	46 (14.6%)	84 (26.8%)	314 (100%)
About once every two weeks (% within frequency of going out)	63 (27.6%)	47 (20.6%)	43 (18.9%)	30 (13.2%)	45 (19.7%)	228 (100%)
About once a week or more (% within frequency of going out)	58 (20.4%)	54 (18.9%)	62 (21.8%)	53 (18.6%)	58 (20.4%)	258 (100%)
Total (% within frequency of going out)	186 (22.5%)	150 (18.1%)	175 (21.2%)	129 (15.6%)	187 (22.5%)	827 (100%)
Previous victimisation
Non-victim (% within previous victimisation)	80 (18.5%)	92 (21.3%)	93 (21.5%)	50 (11.6%)	117 (27.1%)	432 (100%)
Victim (% within municipality size)	110 (23.8%)	81 (17.5%)	94 (20.3%)	91 (19.7%)	86 (18.6%)	462 (100%)
Total (% within previous victimisation)	190 (21.3%)	173 (19.4%)	187 (20.9%)	141 (15.8%)	203 (22.7%)	894 (100%)

High trust police. The results show that age and municipality size significantly predict ‘high trust police’ cluster membership. Older participants (OR = 1.14) are more likely to be a member of this cluster than younger participants. Also, people who live in a small municipality (OR = 0.49) are less likely to be a ‘high trust police’ member than people who live in a large municipality. Relatively less people from a small municipality and more from a large municipality are present in the ‘high trust police’ cluster (see Table 2).

High trust door staff. Age was a significant predictor of ‘high trust door staff’ membership. Older participants (OR = 0.88) are less likely to be a member of this cluster than younger participants.

Overall limited trust. The results show that gender, migration background, and previous victimisation significantly predict ‘overall limited trust’ cluster membership. Men (OR = 1.75) and participants with a migration background (OR = 1.97) are more likely to be ‘overall limited trust’ cluster members compared to women and participants without a migration background, respectively. Participants who have not been victimised previously (OR = 0.49) are less likely to be a member of the ‘overall limited trust’ cluster. As can be seen in Table 2, relatively more men than women, more participants with than without a migration background, and more victims than non-victims are present in the ‘overall limited trust’ cluster.

Overall high trust. The results show that only previous victimisation significantly predicts ‘overall high trust’ cluster membership. People who have not been victimised previously (OR = 1.53) are more likely to have overall high trust in policing actors than people who have been a victim in the past. As displayed in Table 2, relatively more non-victims than victims are present in the ‘overall high trust’ cluster.

General discussion

The current study investigated Dutch youngsters’ perceptions of surveillance and policing actors in the public spaces of urban nightlife areas. More specifically, levels of trust in the police, door staff, and publicly installed CCTV were measured and a comparison between these actors was made. Results show that, generally speaking, our research participants have considerable trust in surveillance and policing actors. This was especially the case for police and door staff, as the general trust means showed that these human actors were rated as more trusting than the non-human actor CCTV. Possibly, this is explained by perceived differences in the temporal ordering of effects between the surveillance and policing actors. Brands et al. (2016) for instance show that research participants consider CCTV of help in the aftermath of an incident, but not so much during the incident. This might, in part, be interpreted as an expression of (limited) institutional trust: the degree to which our participants believe CCTV is competent to act in specific ways (see Hardin 2002) in the specific situation of the urban nightlife area might be limited.

To allow for a more nuanced understanding of the trust ratings, we also performed a cluster analysis subsequent to our general measurement of trust. That is, we checked if there were individuals that rated, or distributed, trust in ways that were more or less similar to others and classified these as groups (of individuals). Five different groups followed from our analysis. We termed these ‘CCTV adversary’, ‘high trust police’; ‘high trust door staff’; ‘overall limited trust’; and ‘overall high trust’. The CCTV adversary cluster displays a trust pattern that more or less resembles the pattern found in the general mean scores: this group has lower trust in CCTV than in human actors. With both trust in police and private security scores being well above the average of the rating scales, results for this cluster also seem in line with Saarikkomäki (2018, p. 9), stating that ‘[t]ypically, participants stated they had general confidence in both police and security guards (…)’. One thing these human actors have in common is that they are able to intervene directly and on the spot, which might be especially valued with regard to the tensions and risks that characterise the night-time economy.

Still, two other clusters differentiate individual trust ratings between these human actors. With trust ratings for CCTV being moderate in these clusters, members of the ‘high trust door staff’ cluster report high trust in door staff, but limited to average trust in police. The opposite is true for members of the ‘high trust police’ cluster who report high trust in police, but limited to average trust in door staff. This differentiation between trust in police and trust in door staff clearly illustrates the added value of our cluster analysis. While the expectation drawn from the existing literatures that – overall – trust in police is likely to be somewhat higher than trust in private security (e.g. Nalla and Heraux 2003) is confirmed, at the same time we do find a group of persons who are more trusting of door staff compared to the police.

Interestingly, in the ‘overall limited trust’ and ‘overall high trust’ clusters, we find two groups of youth enrolled in education who report quite opposing trust ratings across the policing actors. The ‘overall limited trust’ cluster represents individuals who report relatively low to limited trust, independent of the policing actor. A group of individuals expressing overall low trust in policing was also recognised in the study by Saarikkomäki (2018). The ‘overall high trust’ cluster, on the other hand, represents a group of persons that report high trust in policing actors. Most interesting here is the high degree of trust in CCTV expressed among members to this cluster, which we observe in none of the other clusters.

Identifying different clusters (or, groups of individuals) that each have their own and unique pattern of trust in policing actors might explain why previous literature has put forward a group of people expressing fairly limited trust in CCTV (Trimek 2016), while at the same time putting forward a group of people trusting the police more than door staff (Nalla and Heraux 2003), and a group of people displaying low trust in policing in general (Saarikkomäki 2018).

We also find from our logistic regression analysis that gender, age, migration background, municipality size, and victimisation (but not frequency of going out) predict cluster membership. For example, men were more likely to be a member of the ‘overall limited trust’ cluster and less likely to be a member of the ‘CCTV adversary cluster’ than women. However, gender was not a significant predictor for the ‘high trust door staff’, ‘high trust police’, and ‘overall high trust’ clusters. This might explain why previous research has been inconsistent when it comes to – in this case – finding a gender effect in trust (Ren et al. 2005).

Results flowing from our logistic regression analysis also indicate that people with a migration background were more likely to be a member of the ‘overall limited trust’ group, but less likely to be part of the ‘CCTV adversary’ group, than people without a migration background. The notion that minorities display lower trust in the police (and explanations for why that is the case) has been put forward by previous research (Brown and Reed Benedict 2002, Ackaert and van Craen 2005, van Craen 2013). The literatures on urban nightlife have also indicated that people with a migration background often experience (routine) exclusionary practices ‘at the door’ of nightlife venues (Böse 2005, Hadfield 2008, Measham and Hadfield 2009, Søgaard 2014, 2017). Although this seems to be reflected in the finding that membership to the ‘overall limited trust’ cluster is predicted by migration background, we do not find migration background to be (also, and inversely) related to ‘high trust police’ and ‘high trust door staff’. Possibly, people with a minority background generally share lower trust in these human actors, making them fit the overall limited trust cluster better. At the same time, the finding that participants with a migration background are less likely to be in the CCTV adversary group, might be due to the perception of CCTV as a more anonymous policing actor. As a consequence, perceptions of social exclusion might also be less likely. It should at the same time be noted that, if this would be the case, this likely is limited to public (night-time) spaces in cities, as other research illustrates exclusionary strategies in semi-public spaces (such as leisure centres) (McCahill 2008).

In line with research on police-citizen contacts (Skogan 2005, Kääriäinen 2007, Wells 2007), we also see that people who have not been victimised previously are more likely to be a member of the ‘overall high trust’ group, and less likely to be a member of the ‘overall low trust’ group than previous victims. Furthermore, people living in a small municipality were less likely to be a member of the ‘high trust police’ group than people living in a large municipality. This seems to be in line with research by Bradford et al. (2009) arguing that the confrontation with regular police patrols and feeling informed about police activities increases trust. It could be postulated that both might more likely be the case in large municipalities, which in turn might be positively related to people’s judgments about the police’s effectiveness and engagement. In that sense, one might also expect that frequency of going out would predict cluster membership. However, in the current study, no significant results were found. An explanation for this null finding might be that it is not so much the confrontation with regular police control, but the personal police encounter itself that one has during a night out, that matters for one’s trust. One of the limitations of the current study is that we did not include a variable tapping into personal prior experiences with police or security agents. Hence, we can merely speculate about the explanation of this null finding.

We consider our analysis on patterns of trust across policing actors important for several reasons. First of all, our finding that trust tends to differ substantially between policing actors is important against the background of the pluralisation of policing trend in urban public spaces. While other policing actors are increasingly taking over roles that traditionally belonged to the police force, this study shows that trusting the police does not necessarily mean that people also express trust in other policing actors. Not only does this call into question the desirability of transferals of power, it also probes questions of effectiveness, as the literatures clearly illustrate links between trust in policing and people’s willingness to cooperate. Especially within a nightlife context, often considered a liminal space–time in need of extensive surveillance and policing, these questions deserve attention. If the aim would be to close a ‘between actor gap’, our results hold some indications whom to approach. For instance, if municipalities see benefits in a technocratic approach to securing their public nightlife spaces, it would especially be helpful to approach both ‘CCTV adversaries’ (expressing low trust in especially CCTV) as well as ‘overall high trustees’ (the only cluster in which high trust in CCTV is expressed) to better understand why it is a helpful instrument to some, but not to others. As mentioned, we also identified a selective group of individuals holding very little trust across policing actors. It is especially relevant to understand what these persons hold in common and to understand their sentiments when considering policing legitimacy. While a selection of explanatory variables was already outlined in our results, a more in-depth study is needed to explain group membership.

Secondly, and relatedly, our findings are relevant against the background of the theorisation and increasing implementation of CCTV surveillance in urban space, among others in nightlife spaces. While we would argue that our findings largely align with more recent studies that have drawn into question the substantial support the implementation of CCTV surveillance has received (see for instance Saetnan et al. 2004, Taylor 2011, Brands et al. 2016), asking members of the ‘CCTV adversary’ cluster may provide additional explanations why perceptions of CCTV dwindled and differ from those of police and door staff.

Tapping into the discussion of the securitisation of our society, our results suggest that surveillance and policing may, but not necessarily does, have a positive outcome on people’s perceptions. It is therefore crucial that (increases in) surveillance and policing, and transferal of powers to do policing, are implemented with sufficient public support backing them. This might especially be challenging within the nightlife context: while it is important to make sure that nightlife consumers can enjoy a safe night out, authors have at the same time suggested that some risk, excess and peril may actually be an important aspect to the popularity of and the excitement experienced on a night out (Williams 2008, van Liempt and van Aalst 2012), with excessive surveillance and policing running the risk of creating non-stimulating and sterile environments. This again brings us at the importance of studying the different clusters in more depth, to see why support is lacking in some cases.

At the same time, our findings should be interpreted with care. Most importantly, we would call for future research to do greater right to the complexity of the concepts under study. Quite aware of the sizeable discussion surrounding the measurement of trust in the literatures, we readily acknowledge the generality of our trust items. That is, we are unable to specifically tell whether or not participants believe that the institution or individual actors are effective, fair, as representing certain community norms, values, and standards, or a combination of these specific perceptions (Ren et al. 2005, Jackson and Bradford 2010, Cao 2015). As our trust items are mainly focused on the performance of the institution, we believe that the items tap more into the idea of institutional trust. However, they might as well include some interpersonal trust. The above notwithstanding, we chose to work with these items. As our main goal is to provide a (general) comparison of trust in different policing actors, we sought to include items that make such a comparison feasible. By using two items measuring trust per policing actor – as opposed to quite some other studies that use a single item – we sought to compromise between sophistication and feasibility. At the same time, to the best of our knowledge, no validated scales for between actor measurements exist that approach the degree of sophistication provided in the literatures on the police alone (but see van Steden and Nalla (2010) and Moreira et al. (2015) on private security). Neither was it our research objective to develop such a scale. We also considered it quite possible that the (more in-depth) operationalisation of trust – as provided in police studies – might not be one on one transferable to other actors (especially when it comes to CCTV).

Furthermore, our study has employed a limited number of explanatory variables, among which some contextual, but mostly demographic ones. To this end it is also important to note, as Moreira et al. (2015) explain, that it is quite likely that demographic explanations of trust in policing are mediated by other variables. Future research would also do good to consider this. It also means that one should be cautious in interpreting the results from our logistic regression. Finally, the focus on a nightlife context in the current study also means that the results cannot be easily generalised to other contexts. It would be interesting to investigate if comparable patterns in trust in surveillance and policing are present in contexts other than urban nightlife areas.

In general, it can be concluded that people have considerable trust in human policing actors (police and door staff), and somewhat lower trust in the non-human actor CCTV. This general observation became more nuanced when our data was subjected to cluster analysis, on which basis we observed five groups of persons: one displaying high trust in human actors; one displaying highest trust in the police; one displaying highest trust in door staff; one displaying limited trust overall; and one displaying high trust overall. These nuances highlight the importance of research into the pluralisation of authority to ‘do’ surveillance, and at the same time confirms that the concerns regarding the legitimacy of more recent surveillance technologies are valid. The current research further contributes to the existing literature on trust in policing actors by focusing on a rather understudied group, namely youth (enrolled in education) within a nightlife context. More in-depth analyses are necessary though in order to gain more insight into the foundation behind youths’ patterns of trust, on which the current study only offers a first, exploratory, approach.

Disclosure statement

No potential conflict of interest was reported by the authors.

Additional information

Funding

This work was supported by the Netherlands Organization for Scientific Research (NWO) [grant number MVI-313-99-140].