Abstract
Recent high-profile cases have sparked concern about strip searches following stop and search, but less attention has been paid to strip-searching in police custody. Using binary logistic regression on data (N = 25,676) from an English police force, we examine factors associated with strip-searching in custody for those detained under the Police and Criminal Evidence Act, 1984, finding increased odds of strip-searching among individuals who self-define as Black, disclose self-harm or mental ill-health or are arrested for drug-possession offences. Black males and Black children face higher odds than expected from either characteristic alone. As one of the very few, and quite possibly the only article in the last 20 years – either in England or elsewhere – to use multivariate analysis to consider factors associated with strip-searching of both adults and children in a single model, allowing us to control for age and interaction effects, it offers insights for theory, policy and practice.
Keywords
Introduction
The police’s use of strip-search powers has received considerable public, policy and press attention. In England and Wales, the focus has been on incidents in public. For example, the case involving ‘Child Q’, where a Black female child was strip-searched by police officers at her secondary school after teachers reported that she smelt of cannabis, suspecting that she may be in possession of drugs (Gamble and McCallum, 2022). This incident crystallised wider concerns around adultification and the racialised perception that Black children are more mature and culpable than their White peers and less deserving of protection (Davis, 2022). These perceptions are thought to intersect with broader stereotypes that cast Black individuals, especially males, as inherently dangerous.
While such concerns have been highlighted in public settings, there has been less scrutiny of the hidden context of police custody. There is little academic literature on strip searches in general (Duff and Kemp, 2024), and even less on strip searches by the police, and within police custody specifically (for exceptions, see Kemp et al., 2023, 2024). The work that does exist largely focuses on prisons (see Hutchison, 2020 for an overview) and, with a few exceptions discussed below, on North America.
Moreover, there is little systematic evidence about how strip-searching is patterned. Whether inside or outside of police custody, multivariate analyses that can isolate the influence of key variables are rare. To our knowledge, only one such article has been published in England, or indeed, further afield, that looks at child and adult populations in a single model (Newburn et al., 2004), which is important for our purposes as it allows us to control for the impact of age and to look at its interaction effects, among other factors. More recently, works by Kemp et al. (2023, 2024) have provided separate multivariate analyses of strip-searching in child, adult and vulnerable adult populations and found important issues around disproportionality, with Black children more likely to be strip-searched than White children.
Despite this handful of important studies, gaps remain due, in part, to ambiguity regarding legal definitions of strip-searching and data-collection difficulties (Balfour, 2018; Children’s Commissioner, 2024; Grewcock and Sentas, 2019; Güerri, 2023). Work in this area is particularly timely given calls by third-sector organisations, and others, for change (e.g. Bath, 2022) and the recent consultation on proposed alterations to the Police and Criminal Evidence Act 1984 (PACE), and its Codes of Practice A (Home Office 2023a) and C (Home Office 2023b), which purport to ‘strengthen the safeguards for children and vulnerable persons’, including safeguards in custody (UK Parliament, 2024: n.p.).
Our study addresses these gaps by using large-scale custody data, specifically unique administrative data (N = 25,676) from a police agency in England to examine patterns in the use of strip searches across custody suites. Informed by and aiming to contribute to broader theoretical debates on police practices (Bradford and Loader, 2016; Marenin, 1982), as well as more recent theories on strip-searching and ethnicity/race in particular (e.g. Duff and Kemp, 2024), our analysis seeks to (a) identify individual and contextual factors associated with the use of strip-searching in police custody, (b) assess whether ethnicity/race is associated with the likelihood of being strip-searched and (c) examine whether this relationship is moderated by gender and age – each important considerations in their own right. In so doing, it contributes to long-standing debates around whether racial/ethnic disproportionality in policing should be calculated with reference to the resident population (those who live in an area) or street population (those who are visible and available to the police) (Newburn et al., 2004). Our data, drawn from custody records rather than street-level encounters, helps cut through these debates by capturing people brought into custody under PACE, regardless of how the initial contact occurred (Newburn et al., 2004).
We start by detailing the legal and procedural framework of strip searches, before reviewing the literature, empirical findings and theoretical debates. Next, we introduce the data and methods used and present results. We then discuss the substantive and theoretical implications of the results, before concluding.
Strip searches in England and Wales
Arrests remain a core police function, with substantial volumes recorded each year. In England and Wales, there were 12 arrests for every 1,000 people in the year to March 2024 (Home Office, 2025b). Like other parts of the criminal justice system, this is disproportionately concentrated among certain demographic groups, with Black men almost twice as likely to be arrested as White men. Yet, there remains a striking absence of scrutiny around what happens after arrest, particularly concerning the use of strip-searching (Tiratelli, 2022). This is compounded by a complex legal framework, in which various powers permit different types of searches depending on the circumstances of an arrest and the detainee.
In England and Wales, searches conducted in police custody following arrest can be categorised as either intimate or strip searches, both of which are distinct from those conducted during a stop and search. 1 PACE Code C, Annex A (A) (Home Office 2023a) defines an intimate search as physical examination of body orifices other than the mouth. At the time of writing, it must be carried out by a registered medical professional, unless deemed impracticable by an officer of at least Inspector rank. However, as discussed earlier, the consultation on Code C of PACE aims to ‘strengthen the safeguards for children and vulnerable persons who are subject to searches involving the exposure of intimate parts’ (Home Office, 2024c; UK Parliament, 2024), and additional measures are yet to be put in place.
In contrast, a strip search, while less invasive, still represents a significant intrusion into personal privacy. It authorises the removal of more than outer clothing, shoes and socks to retrieve prohibited items the detainee is reasonably suspected of concealing (PACE Code C, Annex A (B), see Home Office 2023b). For instance, a strip search may involve removing multiple layers of upper and lower clothing, though not necessarily all at once. By contrast, an intimate search (sometimes also referred to as a strip search involving the Exposure of Intimate Parts, or EIP) permits the exposure of intimate body parts, such as the genitals, buttocks or female breasts.
However, despite the provisions of PACE, Code C, Annex A, what constitutes a strip search is not always clear-cut, and there has been ongoing debate regarding their classification, particularly in cases where individuals are required to swap clothing for safety reasons (e.g. those at risk of self-harm are asked to change into anti-rip clothing). This was clarified in Davies v CC Merseyside Police (2015), where the court ruled that the removal of more than outer clothing for safety reasons can constitute a strip search. This ruling has led to varied interpretations and practices across police forces. Additional ambiguity arises under Section 54(6) of PACE, which states: ‘a person who is in custody at a police station or is in police detention otherwise than at a police station may at any time be searched [. . .] to the extent that the custody officer considers necessary for that purpose’. This gives Custody Sergeants, who have responsibility for ensuring safety, legality and protection of detainees in a particular custody suite, considerable discretion over the level of invasiveness. The search could involve actions amounting to a strip search if authorised in accordance with PACE Code C, Annex A, (B). Breaches of PACE may occur if the legal thresholds are not clearly understood and officers act beyond the authorisation of the Custody Sergeant.
In response to concerns about inconsistent understandings and recording practices, the College of Policing (2023) updated its professional guidance on strip searches. 2 This came amid increasing public scrutiny and the release of national figures on strip search use for the first time in 2024. 3 These indicate there were 68,874 strip searches in custody across 41 police forces, equivalent to 9% of all detentions (Home Office, 2024b). Of those searched, 70% were from a White background (compared with 78% of all people in custody), 15% were from a Black or Black British background (compared to 8%) and 8% were Asian or Asian British background (same as overall proportion in custody), with the remainder from Mixed or Other backgrounds. However, the Home Office (2024b) notes these statistics do not account for the differing ethnic/racial composition of offence types, some of which may be more likely to result in a strip search than others – something we are able to address here.
The literature on strip searches
Much existing literature on strip searches is concerned with its use in prisons (see Hutchison, 2020 for an overview) and has produced key findings around the far-reaching use, impact and trauma of strip searches, often characterised as ‘sexual assault by the state’ (George, 1993: 31, Kilroy, 2003). It is often seen as patterned along ethnic/racial, Indigenous and gendered lines (e.g. Hutchison, 2024), and the literature often focuses on female detainees (see, for e.g. Balfour, 2018; Hutchison, 2020, 2024; Kilroy, 2003; Latty, 2023). These are important insights, but the literature is predominantly qualitative, and in rare cases where authors have managed to access quantitative data (e.g. Güerri, 2023), analyses tend to be descriptive.
Moreover, there are few works looking at its use by the police – and even fewer looking at police custody specifically despite recognition of this as a cause for concern (e.g. George, 1993) – with much of the literature focusing on North America and Australia (e.g. Grewcock and Sentas, 2021; Hutchison, 2020; Latty, 2023). It is crucial to look outside of these contexts – especially given the paucity of literature on police strip-searching in England and Wales. Although there have been important policy contributions on the topic (e.g. Bath, 2022), only two academic articles have focused on the practice in the jurisdiction 4 (Duff and Kemp, 2024 and Newburn et al., 2004), and of these, only Newburn et al. used multivariate analysis. That said, Kemp et al.’s (2023) report on the impact of PACE on child suspects included a multivariate analysis of strip searches on children in police custody, finding that drug offences, being handcuffed to the back and being Black increased the odds of the practice occurring, with similar patterns found in a later report by Kemp et al. (2024) who conducted separate analyses on child, adult and vulnerable adult populations.
Newburn et al. (2004), who analysed strip searches of adults and children in police custody, found that Black African-Caribbeans, male population and those between 17 and 23 years of age had increased odds of being strip-searched, while arrests for drug offences had the greatest explanatory power. These valuable findings are now somewhat dated and reliant on data from one police station across a 16-month period (Newburn et al., 2004). In contrast, we are able to analyse data from a 4-year period across multiple custody suites, 5 giving us a significantly larger sample size than previous studies (over 27,000 records, compared to around 7,000 in Newburn et al., 2004).
The second article (Duff and Kemp, 2024) sees strip-searching as ‘a form of normalised racial and sexual violence . . . (forming) part of a project of abjectification’ that aims to expel its targets from ‘the realm of the human’ (Duff and Kemp, 2024: 1). Under this theory, people are subjected to strip searches because of their group memberships and what this represents. Indeed, Duff and Kemp (2024: 20) note that, ‘while certain groups are systematically targeted, within those groups . . . (decisions) as to whom to strip-search . . . can be quite arbitrary’ as strip searches are conducted to perpetuate ‘state-inflicted harms’ (Duff and Kemp, 2024: 3) against ‘young people and people of colour, especially Black young men and boys’ (p. 1). While a valuable theoretical contribution, as the authors’ data came from Freedom Of Information requests, they were unable to test it using multivariate techniques.
This theory can be contrasted with a more conventional lens, consistent with accounts from consensus theorists and the police themselves. Such theoretical approaches emphasise what Bradford and Loader (2016: 241) call the ‘common-sense’ view of policing. This sees police action as driven by a desire to detect and prevent crimes and other forms of harm. Such approaches lend themselves to two rationales for strip-searching as discussed by Duff and Kemp (2024: 1): a ‘crime-detection rationale’ in which strip-searching is motivated and justified by attempts to detect illegal possessions, most prominently drugs and the ‘care rationale’, in which strip-searching is motivated by a concern to remove possessions, including potential weapons and intoxicants, that could be used to self-harm or harm others. Our data, which we turn to next, allows us to explore whether such factors are associated with strip searches and, in so doing, to examine these theories.
Methods
Data
The data was sourced from one police force area in England which serves a predominantly (approximately 95%) White, and rural, population. In 2023/24, crime rates ranged from 60 to 100 per 1,000 population, and approximately 5% of neighbourhoods within the force area ranked among the 10% most deprived in England (as defined by the Index of Multiple Deprivation). The data provided by the force included all custody records generated between January 2018 and December 2022 (N = 27,577). Each record is a live report created in the custody system when an individual arrives at a police station and contains key information on the detainee, including demographic details, health and vulnerability indicators and circumstances of the arrest. Although the police agency operates multiple custody suites, most records came from two sites (17,412 and 7,233 records, respectively).
Data was acquired through a formal data request, supported by the College of Policing. Requests were sent to all police forces across England and Wales, although several forces expressed interest, most were unable to provide data due to resource limitations or data availability constraints. During cleaning, we found 1,806 entries 6 for non-PACE offences, which were excluded from the analysis. To handle remaining missing data, we employed complete-case analysis, excluding any records with missing values. 7 The final sample included 25,676 observations.
Dependent variable
The dependent variable is a binary measure for strip searches, with custody records where an individual had undergone a strip search while in police custody (coded as 1) compared to cases where individuals were arrested but not strip-searched (coded as 0). 8
Independent variables
Demographic characteristics
We include dummy variables to represent the ethnic/racial identity of the detainees: Asian, Black, Mixed and Other, with White as the reference category. These categories are derived from self-defined ethnicity rather than officer-observed ethnicity, as the latter does not include a distinct ‘Mixed’ category and is therefore less granular. Using officer-observed ethnicity would risk obscuring substantively important differences between ethnic groups relevant to the analysis. Gender is coded as 1 for male and 0 for female, with no other categories recorded in the dataset. Age is included as a binary variable, where 1 represents individuals aged 18 or younger at the time of arrest, and 0 represents those older than 18.
Offence type
We include dummy variables to capture the offence type for which the detainee was arrested. Echoing the approach of Kemp et al. (2023, 2024), and drawing on the Home Office (2025a, 2025b) crime reporting rules, these are grouped into seven categories: acquisitive (burglary, fraud, theft of motor vehicles and other theft/handling), criminal damage, drugs, motoring, violence, sexual offences and other offences (including Public Order Act offences), with the latter as the reference category.
Use of force
We include a binary measure indicating whether force was applied against the detainees while in custody. In this dataset, the use of force includes the drawing and use of weapons and police dogs, as well as empty hand techniques, limb restraints and handcuffing. Detainees are categorised based on whether force was used against them (coded as 1) or not (coded 0). Observations where handcuffing was recorded as the only form of force (N = 697, 2.53%) are coded as 0 (i.e. non-use of force) to avoid conflating handcuffing – which can occur as a standard precaution when detainees arrive in custody – with force that may be applied while in custody.
Vulnerabilities
We include mental health (coded 1 if a detainee disclosed a diagnosed condition or concern upon arrival into custody, otherwise 0), self-harm (coded 1 if the detainee indicated any past self-harm or suicide attempts, otherwise 0) and intoxicant consumption (coded 1 if the detainee had disclosed consuming drugs, alcohol or other mind-altering substances at the time of arrest, otherwise 0). We also include request for interpretation services, as language barriers can increase detainees’ vulnerability by limiting their ability to communicate needs and understand custody processes (coded 1 if an interpreter was requested, otherwise 0).
Time of arrival
We include a measure for time of arrival, to reflect that arrests, and staffing patterns in custody, may vary systematically throughout the day. For example, individuals arrested late at night may be more likely to be intoxicated, whereas daytime arrests may more often involve planned operations. We create 6-hour time intervals categorised as Morning (06:00 am to 11:59 am), Afternoon (12:00 pm to 5:59 pm), Evening (6:00 pm to 11:59 pm) – the reference category – and Night (00:00 am to 05:59 am).
Analytical strategy
We employ a series of binomial logistic regression models, beginning with a baseline model that includes core demographic variables of the detainee: race, gender and age. This model examines whether individuals from different racial backgrounds face differing likelihoods of being subjected to a strip search. We then sequentially add blocks of conceptually grouped variables. Model 2 adds indicators of vulnerability, and model 3 introduces situational factors such as offence type, use of force and arrival time.
In model 4, we include an interaction term between being male and Black to test whether their intersection increases odds of strip-searching beyond the additive effects of gender and race. 9 Model 5 tests a further interaction between being under 18 and Black, to assess whether racial disparities are more pronounced among younger detainees. 10 This modelling strategy allows us to evaluate the extent to which racial disparities are explained by other factors, and whether the effects of race are moderated by other characteristics through the inclusion of interaction terms.
All models report clustered standard errors to account for within-cluster correlation. This approach adjusts for similarities among units such as states or regions due to unobserved contextual factors (Abadie et al., 2023); in our case, differences across custody suites. This affects only the precision of estimates (i.e. the standard errors), not the estimated coefficients. To validate this approach, we estimated the full model both with and without clustered standard errors and observed notable differences, supporting their use. We also assessed multicollinearity using variance inflation factors (VIFs); all predictors exhibited low VIFs (from 1.01 to 1.50), indicating this is not a concern. We undertook sensitivity tests to account for the potential impact of the pandemic by including variables representing different lockdown periods 11 in a fully specified model. The inclusion of lockdown measures did not alter the effect sizes of the other predictors, indicating that our findings are robust and not significantly confounded by lockdown periods.
Results
Descriptive statistics
Table 1 presents descriptive statistics of the 25,676 detainees brought into custody under PACE. Around one in four detainees (24.97%) experienced a strip search, equivalent to 271 strip searches per 1,000 detainees. 12 However, these searches were not evenly distributed across the custody throughput.
Descriptive statistics.
Males made up the majority of the arrestee population (83.92%) consistent with their representation of those strip-searched (84.72%). Black detainees accounted for 3.04% of the total population, but 5.3% of those were strip-searched. Mixed and Asian individuals showed smaller differences between their representation in the overall sample and among those strip-searched. Individuals under the age of 18 made up 8.34% of the total detainee population, and 8.92% of those were subjected to a strip search. To put this in context, official statistics (Home Office, 2025b) indicate approximately 84.1% of those arrested are male, around 8% were under 18 (similar to our sample) and 7% described their ethnicity as Black, where self-described ethnicity was provided. To put this in context, official statistics (Home Office, 2025b) indicate that approximately 84.1% of those arrested are male, around 8% were under 18 (similar to our sample) and 7% described their ethnicity as Black where self-defined ethnicity was known. These national statistics also indicate persistent missingness in self-defined ethnicity, with approximately 15% of cases where this information was not recorded.
When looking at reason for arrest, violence (33.33%), ‘other’ (21.82%) and drugs offences (19.01%) made up the highest proportions of those who were strip-searched, while sexual offences (1.98%) and criminal damage (5.79%) accounted for the smallest shares. The use of force was recorded in 6.42% of records, but this was more than double (14.87%) among those who were strip-searched. Signs of vulnerability were common in both the full sample and among those strip-searched, though slightly more pronounced in the latter. Over half of strip-searched individuals had consumed intoxicants (57.87%) or presented with mental health concerns (53.72%), compared to 54.16% and 48.75%, respectively, in the overall sample. Having a history of self-harm was also slightly more prevalent among those strip-searched (39.03%) than in the full sample (33.62%). Request for an interpreter use was rare overall (2.28%) and even lower among those strip-searched (1.15%). Time of arrival in custody varied between the full and strip-search sample. Morning arrivals accounted for 15.82% of strip-searched individuals but 18.04% of the total custody population. In contrast, night arrivals constituted 24.44% of those strip-searched, compared to 26.76% of the full sample, indicating slight underrepresentation in those subjected to the practice.
Multivariate analysis
Table 2 presents the results of models 1–5, which are discussed in turn below. Model 1 begins with demographic factors and reveals significant disparities. The observed disparity for Black detainees (odds ratio (OR) = 2.39, p < .001) indicates that, in the absence of contextual information, race is strongly associated with the likelihood of a strip search. Detainees with a Mixed background also face elevated odds (OR = 1.62, p < .01). While these differences may be confounded by other factors, they point to an initial disparity that warrants closer examination. Neither being male (OR = 1.06, p > .05) or being under 18 (OR = 1.08, p > .05) is statistically associated with strip searches at this stage, suggesting that observed disparities cannot be attributed to age or gender alone.
Binomial logistic regression predicting strip searches.
The reference category for ethnicity is White/White British.
The reference category for offence type is ‘other’ offences.
The reference category for time of arrival is evening.
AIC = Akaike Information Criterion, BIC = Bayesian Information Criterion.
Statistically significant results are in bold, p < 0.05.
Model 2 introduces vulnerability indicators. Mental health need (OR = 1.20, p < .001), self-harm (OR = 1.28, p < .001) and consumption of intoxicants (OR = 1.19, p < .001) are all associated with higher odds of being strip-searched. In other words, individuals perceived as vulnerable are more likely to be subjected to this form of intrusive policing. By contrast, requiring an interpreter is associated with lower odds of strip search (OR = 0.48, p < .001), suggesting it may function as a protective factor. Conditioning on the other variables shown, the odds of strip search for Black detainees increase slightly (OR = 2.66, p < .001) when compared with Model 1.
Model 3 adds situational factors, specifically the reason for arrest, use of force and time of arrival, resulting in a fully specified model that includes all the main effects. Arrests for drug offences are associated with more than twice the odds of a strip search (OR = 2.45, p < .001), while the use of force is highly predictive with this police practice (OR = 4.97, p < .001). Although the odds for Black detainees are reduced slightly (OR = 2.18, p < .001), the disparity remains statistically significant. In other words, situational context only partially explains racial differences in strip search likelihood. Time of arrival also appears to be an important factor, with individuals arriving in the afternoon having significantly higher odds of being strip-searched than those arriving in the evening (OR = 1.24, p < .001).
Model 4 extends the previous model to examine moderating effects by including an interaction item between Black and male. In this specification, the main effects for Black (OR = 1.06, p > .05) is no longer significant. However, the interaction term is both positive and statistically significant (OR = 2.14, p < .001), indicating that Black men face substantially higher odds of being strip-searched than would be expected if the effects of race and gender were simply additive. In other words, the combination of these identities produces a heightened likelihood of strip search that is not explained by either characteristic alone.
Model 5 tests a second interaction, this time between being Black and under 18. The interaction term is again positive and statistically significant (OR = 1.69, p < .001), indicating that Black children face significantly higher odds of strip search. This may reflect a tendency for Black children to be treated differently from their peers.
Figures 1 and 2 13 display differences in predicted probabilities. The effect of gender is negligible among non-Black detainees but substantially larger among Black individuals, with a similarly pronounced pattern observed when comparisons are made by age.
Differences in probability of strip search by gender and ethnicity.
Differences in probability of strip search by age and ethnicity.
Discussion
The results provide some evidence in support of the crime-detection and care rationales: the odds of being strip-searched increase when the individual involved has been arrested for drug offences (indeed, in keeping with Kemp et al., 2023, 2024; Newburn et al., 2004, this often had the biggest effect size of all the variables studied), has a prior history of self-harm and has consumed intoxicants. Yet strip-searching is scarcely reducible to, or excused by, such rationales. Rather, as Newburn et al. (2004: 693) concluded over two decades ago, our findings also point to the ‘spectre of police racism’ – particularly, although not exclusively, anti-Black racism.
Self-reporting as Black emerges as consistently associated with increased odds of strip-searching, after controlling for the nature of the suspected offence, arrest characteristics, vulnerabilities and characteristics of custody (see also Kemp et al., 2023, 2024). Similar – although less pronounced – effects are also found for those who describe their racial background as ‘Mixed’. The results also highlight the racially patterned nature of strip-searching has remained a consistent feature over time, place (appearing both in inner-city custody and in a largely rural police force) and irrespective of the method of reporting used (with Newburn et al., 2004 using officer perceptions, versus our use of self-reporting).
Moreover, our results show that gender and age significantly moderate the relationship between ethnicity/race and strip-searching. While the literature on strip-searching in prison has focused on female experiences and how this intersects with anti-Black racism (Hutchison, 2024), our findings highlight important ways in which male and Black identities intersect. Black men have higher odds of being strip-searched than would be expected from the effects of independent effects of race and and gender alone. In the same way, Black children face significantly higher odds of strip search than White minors. These findings provide some evidence for ‘adultification’, defined by Davis (2022: 5) as ‘a persistent and ongoing act of dehumanisation, which explicitly impacts Black children, and influences how they are safeguarded and protected’, with Black children being treated less favourably than their White counterparts. They also provide some support for Duff and Kemp’s (2024) abjectification theory which posits important interactions between Blackness, age and gender – a point we return to shortly.
Moreover, there is a significant association between force being used in custody and increased odds of strip-searching; in our data, approximately 14% of people strip-searched had force used on them, compared to 6% of the arrested population. Some of this may be explained by police feeling the need to use force to occasion the search. However, given that many strip searches do not involve (other) uses of force, it is also possible that strip-searching could be used as punishment for previous perceived (mis)behaviour necessitating the use of force, and refraining from strip-searching could be a ‘reward for compliance’ (Savigar-Shaw et al., 2022, see also Hutchison, 2020).
These patterns are important in their own right but also for the light they shed on ‘crime-detection’ and ‘care’ rationales. While, as discussed earlier, there is some evidence of these rationales being associated with strip searches, the end does not necessarily justify the means. Even if some seek to justify strip searches on these grounds, the practice nevertheless results in Black men and children being over-policed, in this case by being subjected to strip searches and degradation, dehumanisation and trauma associated with the practice (Duff and Kemp, 2024).
Justifying the practice with regards to the ‘care’ rationale is also problematic; as the Baird report (2024: 127) notes, ‘stripping a man or a woman naked by brute force, however clinically done, would be likely to make a depressed person feel worse . . . (while) not remov(ing) the ability to self-harm’. The impact of the practice is particularly concerning given our finding that there is a positive association between disclosing a mental health problem and being strip-searched – even after controlling for a previous history of self-harm – and that the odds of Black people being strip-searched were more pronounced after vulnerability indicators were introduced. Perhaps, for White detainees, vulnerability indicators come to be seen as a reason not to strip-search, essentially operating as a protective factor. Conversely, Black detainees do not experience such protection; instead, they may experience what Shadravan et al. (2021: 623) refer to as ‘the compounding risk of having mental illness and being Black’.
Finally, before concluding, we want to pull out some theoretical implications of these findings. First, the evidence in support of both consensus and abjectification theories, with variables associated with each perspective reaching statistical significance, suggests that they are not necessarily mutually exclusive. This recalls Marenin’s (1982: 241) important insight that police officers may engage in both ‘parking tickets and class repression’, which we might want to extend to encompass other forms of repression and biased policing too. Indeed, our findings are consistent with racialised patterns of policing found elsewhere (e.g. Bradford and Loader, 2016), highlighting the possibility of police racism. Even if this may co-exist with other rationales and practices, the persistence of race cannot be ignored.
Second, and relatedly, evidence for both theories does not negate the value that more critical approaches can bring; indeed, it is not enough that ‘care’ and ‘crime’ rationales have a part to play. The drivers of a practice such as strip-searching might be less important than the consequences; whatever be the intentions in individual cases, the cumulative result is the over-use of an extensive and traumatic police power on minoritised communities, primarily those identifying as Black.
Third, these findings might lead us to provide some support for Duff and Kemp’s abjectification theory. However, while our findings of the importance of self-identifying as Black and interactions with age and gender are in keeping what this theory predicts, this is necessary but not sufficient. While we can rule out other potential explanations for racial disparities, we are less able to adjudicate between competing explanations for the continued persistence of race (see also Newburn et al 2004: 693). Duff and Kemp (2024: 1) argue that strip-searching ‘aim(s) to exclude the individuals and groups it targets from social and political subjecthood’; this might well be the case, but the patterns we identify could also be explained by a range of other officer motivations. For example, Bradford and Loader’s (2016: 254) analysis of stop and search found that the police use this power disproportionately due to ‘multiple, intersecting causes . . . (including) “various forms of bias, stereotyping and institutional racism” as well as organisational incentives’ (p. 247). Ultimately, however, our findings provide further evidence in support of racialised policing, irrespective of their causal processes.
Conclusion
This article has used a dataset from an English police force to examine factors associated with the odds of being strip-searched in police custody. As such, it is one of the very few, and quite possibly the only, article in the last 20 years – either in England or elsewhere – to use a multivariate analysis to consider factors associated with strip-searching of both adults and children in a single model, either in police custody or, indeed, in other contexts.
Nevertheless, there are several limitations to the dataset that should be considered. One well-documented challenge in research using police-recorded data is inconsistencies, errors and gaps in record-keeping, as well as its status as a police-generated account (Rojek et al., 2012), and strip-search data, and associated variables, are no exception to this (see Nickolls and Allen, 2022: 6 for a recent discussion in the English and Welsh context, as well as Balfour, 2018; Children’s Commissioner, 2024; Grewcock and Sentas, 2019; Güerri, 2023). This may mean that some strip searches were not recorded and are therefore missing from the analysis and also highlights the limitations of police accounts more broadly. Future research could usefully look not just at police accounts but also at the accounts of those who have experienced strip searches to gain a fuller perspective.
Second, we are unable to provide detailed information about the nature or extent of the strip searches conducted. We use a binary measure to capture the occurrence of a strip search since the dataset lacks detail on the level of invasiveness for each search. We are unable to categorise strip searches based on whether intimate parts were exposed, as this is not systematically recorded by the police agency. In addition, recording practices for strip searches can vary across forces, both in terms of consistency and the level of detail captured. This variability presents challenges in assessing the full extent of these practices and may contribute to underreporting or misclassification. While our reliance on data from a single police force area ensures consistency, it also limits the generalisability of the findings. Policymakers should consider implementing mandatory nationally standardised data-collection practices that ensure greater transparency and allow for meaningful comparisons across police force areas.
Third, we do not have information on any pre-existing Police National Computer warning markers from previous arrests (e.g. past incidents involving concealing a prohibited item) which could play a significant role in influencing the decision to conduct a strip search. Their absence in the data may limit our understanding of the full range of factors associated with such process. Future research should seek to understand the role of warning markers in authorising more intrusive forms of policing such as strip and intimate searches.
Fourth, a key omission from the dataset is whether the grounds for authorising the strip search, aside from those related to safeguarding (e.g. to prevent self-harm), align with the outcomes of the search. This limits our ability to assess whether the justification for the strip search was validated by its results. To our knowledge, such information is either not routinely captured in England and Wales or is not easily retrievable, raising concerns about monitoring and oversight. Moreover, while we have comprehensive information on the detainee, we lack data on the officers involved in authorising or conducting the strip search, particularly the custody sergeant.
Despite these limitations, the reliance on strip searches in custody, and its patterning along the lines of race, as well as other variables such as mental health, raises critical questions about its necessity, particularly when those subjected to them are often individuals with pre-existing vulnerabilities. While the proposed changes to PACE Code C (Home Office, 2024c; UK Parliament, 2024) may seek to introduce some positive additional safeguards around intimate searches, in and of themselves, they may do little to challenge the concerning patterns in strip-search use (detailed earlier), or the ways in which they remain embedded in risk-management practices in police custody (e.g. preventing self-harm or suicide). If policing is to be trauma-informed, the use of strip searches on already vulnerable individuals must be critically and more fundamentally examined, particularly when there is little empirical evidence to justify their utility and when less-intrusive tactics were considered first (see also His Majesty’s Inspectorate of Constabulary and Fire & Rescue Services (HMICFRS), 2022).
More broadly, the responsibility for managing risk cannot be divorced from the responsibility for inflicting harm. Any policy or legal framework that legitimises strip-searching without evidence-based safeguards risks being fundamentally incompatible with principles of dignity and care and arguably does more to perpetuate harm than prevent it. When that harm falls disproportionately on Black individuals, it becomes impossible to ignore the role of race in the exercise of power in police custody.
Footnotes
Acknowledgements
We would like to thank the police force, their data analysts and Dr Paul Quinton (College of Policing) for facilitating access to the data. We would also like to thank Inspector Paul Thompson (Essex Police) for his invaluable insights into the legal framework and Professor Jonathan Jackson and Professor Ben Bradford for comments on an earlier version of this article.
Funding
The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by the Economic and Social Research Council [grant number ES/P000622/1] and the LSE Scholarship.
Declaration of conflicting interests
The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Data availability statement
Further information on the dataset used and findings may be available on request from the corresponding author. The data are not publicly available due to containing information that could compromise the privacy of research participants.
Open access
For the purpose of open access, the author has applied a Creative Commons Attribution (CC BY) license to any Author Accepted Manuscript version arising from this submission.