Rapid Evidence Assessment on the use of DNA and body-worn cameras and the relationship with criminal justice system outcomes

Q: Annex A: Keywords and search terms

Increased use of new technology: body-worn cameras ( BWCs ) ‘body worn camera’ OR ‘ BWC ’ OR ‘bodycam’ OR ‘body worn video’ OR ‘BWV’ AND ‘Police’ OR ‘ CPS ’ OR ‘Prosecution’ OR ‘Courts’ OR ‘Prison’ OR ‘Probation’ OR ‘Criminal Justice System’ OR ‘ CJS ’ AND ‘Case Progression’ OR ‘Attrition’ OR ‘Prosecute’ OR ‘Charge’ OR ‘Convict’ OR ‘detect’ OR ‘clearance’ OR ‘arrest’ OR ‘outcome’ OR ‘impact’ OR ‘resources’ OR ‘costs’ OR ‘delay’ OR ‘backlog’ OR ‘guilty plea’ Use of forensic evidence by the police and prosecutors: forensics and impacts on case progression ‘Police Officer’ OR ‘Prosecute’ OR ‘SOCO’ AND ‘ DNA ’ OR ‘Forensic’ AND ‘Case Progression’ OR ‘attrition’ OR ‘Prosecution’ OR ‘Conviction’ OR ‘Delay’ OR ‘Sentence’ OR ‘Resource’ OR ‘charge’ OR ‘detect’ OR ‘clearance’ OR ‘arrests’ OR ‘criminalistics’ OR ‘proceeded against’ OR ‘case outcome’ OR ‘offences brought to justice OR ‘criminal justice system’ OR ‘ CJS ’

Question 1

Executive summary

Accepted Answer

The aim of this rapid evidence review was to assess research evidence on the relationship between forensic material, specifically trace DNA, and body-worn camera (BWC) footage and numbers of charges, prosecutions and convictions. The study also considers other possible impacts on the progression of cases (for example, processing times). Wider effects on the criminal justice system (CJS), such as possible deterrent effects, are also considered.

Keywords were used to search a range of bibliographic databases (4 for BWCs, and 5 for forensics), stakeholder sources, and one online search engine to identify relevant literature published between 1992 and 2023.

For forensics, 146 initial studies fitting the inclusion criteria were identified. Screening for quality and relevance reduced this to 4 reviews, 5 key studies and 28 secondary studies. The search results for BWC research identified 105 studies, which, after screening, yielded 6 reviews, 12 key studies (plus one review) and 10 secondary studies. Studies were assessed for quality using the Maryland Scale.

Relatively few high-quality empirical studies were found in DNA/forensics. While more robust study designs were found in BWC studies, many of these studies have methodological weaknesses.

Forensics and trace DNA and the relationship with CJS outcomes

Studies in this area cover 2 main themes: the impact of processing and analysis of trace DNA and other forensic material from crime scenes on CJS outcomes (mainly arrests and charges). And the impact of the establishment or expansion of DNA databases on the detection of offences and re-offending.

Robust evidence on the effectiveness of trace DNA on CJS outcomes is limited. Several less robust, cross-sectional studies on the collection and analysis of forensic material at crime scenes found cases with forensic evidence to be associated with more positive outcomes (for example, prosecutions). These studies are problematic because they often simply provide a ‘snapshot’ comparison of forensic versus non-forensic cases. They do not control for other characteristics that are likely to exist between these 2 groups. In addition, several re-analyses of the same datasets have not found significant differences between the 2 groups.

One multi-site randomised control trial (RCT) study (Roman et al., 2008) examined the impact of DNA trace material on the case progression of property crimes. This provides the most robust evidence that collecting and analysing DNA traces at volume crime scenes led to significantly more suspect identifications, arrests, and prosecutions. These effects are also quite large, with twice as many arrests and prosecutions made compared to control cases. This is likely to result in an increase in demand on the downstream with more suspects entering the CJS. The study did not, however, track cases beyond prosecutions (that is, they did not explore pleas, convictions and sentencing).

And while the results of this single multi-site study were positive, they relate only to that subset of property offences where DNA traces are present and are successfully retrieved.

The profile of suspects identified through DNA is different from those identified through traditional investigative methods. The RCT study found that DNA-identified suspects had at least twice as many arrests and convictions for both felony and misdemeanour crimes than for suspects identified under control conditions.

Since DNA trace material is not collected from crime scenes on a large scale, its overall impact in terms of detection of total crime is modest. However, for some offence types (for example, burglary) and arguably some sub-categories of offences (for example, stranger offences), the ability to link crimes to otherwise unknown offenders is likely to be important.

In summary, the retrieval of trace DNA from crime scenes appears to increase the volumes of cases progressing through the CJS, with a single robust study identifying a positive impact from DNA retrieval at property crimes. Cases with DNA traces are likely to lead to more arrests and more charges, but there is no strong evidence on the impact on pleas and convictions. Nor are there currently any robust studies examining the impact of DNA on violent and sexual offences.

DNA databases and the relationship with CJS outcomes

Several studies have examined the impact of the establishment, or expansion, of regional or national DNA databases. These are databases that store DNA profiles from convicted or charged offenders for searches against DNA retrieved from crime scenes. This review identified 4 key studies which focused on the impacts of DNA database expansion.

Quasi-experimental analysis of the specific deterrent effects of DNA databases suggests that they can reduce levels of re-offending (and therefore ease pressures on the CJS). Research in the US on the expansion of DNA databases found modest effects in reducing re-offending for violent crime (Doleac, 2017), robberies, and burglaries (Bhati, 2011; Bhati and Roman, 2014). In Denmark, a study of the more extensive and rapid expansion of a National DNA Database (NDNAD) found larger reductions in re-offending, especially for serious violent crimes (Anker et al., 2021).

One study also found that the incremental expansion of DNA databases in 7 US states also had a wider deterrent effect on the behaviour of the general population (that is individuals who are not currently on the database) in terms of reductions in violent and property crime rates. But here too effects are small (Doleac, 2017).

Finally, several of these studies have found that alongside deterrent effects, DNA database expansion can have positive effects on the detection of crime, increasing charges, as more offences are linked to more suspects. Bhati and Roman (2014) found this impact outweighed the deterrent effect, but other studies found the reverse (that is, the deterrent effect was stronger) (Doleac, 2017; Anker et al., 2021). The lack of consistent findings makes it hard to assess the net impact of DNA databases on the CJS downstream.

In summary, there may be some reductions in the demand on the CJS from the deterrent effects of introducing or expanding DNA databases for offenders whose DNA profiles are on the database. Databases may also have some small general deterrent effects on offending by the wider population. Database expansion may also increase the likelihood of detecting some offenders. The evidence on whether the overall net effect of database expansion on CJS demand is negative or positive is unclear from the current evidence.

The use of BWCs and its relationship with CJS outcomes

Compared to trace DNA, there are a much wider variety of possible impacts from the use of BWCs than simply the relationship with charges or prosecutions. For example, studies of the impact of BWCs have examined how they can affect a variety of outcomes (for example, complaints against the police, police behaviour, the behaviour of the public). There are currently few high-quality studies which assess the impact of BWCs on CJS outcomes.

This review identified 12 key studies which passed quality thresholds, and one focused review. Overall, the findings from the studies in relation to impacts on CJS outcome measures are inconsistent.

A particular limitation is that many studies examine the impacts on CJS measures without breaking down the impact by offence type. The current evidence suggests that crime type and the setting in which the offence takes place, are important in determining how the BWCs affect CJS outcomes. One focused review of the impact of BWC evidence on the downstream CJS concluded that studies of domestic abuse (DA) flagged offences tended to show increases in CJS volumes, while studies of total crime tended to find no effect.

One of the reviewed studies found a positive effect for BWCs on charges for all crime types. However, other studies which examined the impact on charges for all offences, but controlled for a wider range of factors, such as location, found no positive effect on CJS outcomes. Studies which have focused on impacts on charges for DA-flagged offences appear to have more positive results, with 2 finding BWC cases resulted in significant increases in charges and convictions for DA-flagged cases.

Findings in relation to time to outcome are mixed. Some studies found that BWC material was associated with longer case progression times. However, other studies suggest that BWCs can shorten overall disposition times, and lead to faster case dismissals. These effects were dependent on the type of offence.

There are some general limitations with the research evidence in this area; many studies of BWCs and their impact on the CJS do not consider the actual use of BWC material in the progression of cases. In other words, they do not consider whether the footage contributes to police or prosecutors’ decision-making or its use in the courts. If the use of BWC footage is not captured in the study, researchers are simply inferring the outcome due to the presence of BWC material (Petersen et al, 2023).

Few studies look at the impact of BWCs beyond charges (for example, guilty pleas) and the evidence is also mixed. Two US studies found a modest increase in guilty pleas for DA cases. However, other DA studies, and studies of other crime types, have found no effect.

In summary, although there is a growing evidence base on the use of BWCs, their impact on the CJS remains hard to assess. Only a few studies which examine CJS impacts account for officer compliance with carrying and activating their BWC, or test whether meaningful footage was captured, used to build a case, or provide evidence in court. The evidence does suggest that the impact of BWC footage on CJS outcomes varies by crime type, and that the impact may be most marked in DA-flagged cases. But even here the evidence is not the most robust.

Question 2

1.  Introduction and methodology

Accepted Answer

1.1 Introduction

This work was carried out as part of the CJS Demand Insights programme. This programme of analytical work was awarded funding through the Treasury’s Shared Outcomes Fund, and aimed to enhance the evidence base around CJS demands. The Home Office led on this work, with support and input from key partners such as the Ministry of Justice (MoJ) and the Crown Prosecution Service (CPS). The overarching aims of the programme were to explore what factors influence charge volumes and suspects proceeded against, and to improve understanding of the whole system dynamics between policing, CPS, courts and prisons.

There is a limited understanding of how changes in workforce size, composition, decision-making and the use of technology within the police and CPS, impact the downstream CJS. This presents significant challenges to the CJS and can have a detrimental impact on victims as well as the general public’s faith in the CJS (Home Affairs Committee, 2018). Consequently, case progression through the CJS has become an area of concern. This literature mapping exercise focuses on research into how technology (specifically BWCs) and forensic science (specifically trace DNA) have impacted charges, out of court resolutions and convictions.

Technological advances, including the use of forensic sciences and digital evidence, have affected many elements of the CJS. New technologies have been adopted at every stage of the CJS from the police (for example, BWCs, digital case preparation) to the courts (for example, virtual hearings and trials), and prisons (for example, virtual visits).

This literature mapping exercise focuses on 2 specific areas where there is a meaningful research evidence base: (1) DNA, initially used in the detection of serious crimes, but now commonly used across a wide range of crime types; and (2) the use of BWCs, one of the most rapidly diffused police technologies (Lum et al., 2020).

1.2 Conceptual framework

A Rapid Evidence Assessment (REA) methodology (see, for example, the Rapid Evidence Assessment Toolkit, Civil Service, no date.) was used to identify studies and reviews. The initial evidence review exercise was deliberately broad and covered a much wider range of potential factors and themes considered likely to impact CJS demands, including the role of technology, police workload, resources, and performance regimes. Although this research mapping exercise was valuable and identified some useful individual studies and clusters of themes, the searching parameters were too broad.

After the initial searching exercise and review, the initial long list of themes was narrowed to focus on only 2 areas: the application of forensic science (focusing particularly on the use of trace DNA), and BWCs. The choice of these areas was determined by the fact that, within the wider research into CJS demand flows, small bodies of research on BWCs and the use of trace DNA tended to typically capture how these technologies had impacted on charges, convictions, sentencing, and processing times.

Two research questions underpin the overarching aim of the project to examine whether the use of: 1) DNA (either through the collection of trace DNA at crime scenes or through the development/ expansion of DNA databases); or 2) BWCs impact the numbers of charges, convictions or other outcomes, as well as other metrics relevant to the progression of cases (for example, processing times). Other impacts relevant to the CJS, such as deterrent effects, where identified, were also to be considered.

1.3 Methodology: databases searched

A comprehensive search of academic and open-source databases relevant to the CJS was undertaken. Across the 2 areas, 4 bibliographic databases for BWCs and 5 bibliographic databases for forensics (DNA), a range of stakeholder sources, and one online search engine were searched to identify the most relevant literature published within the last 30 years (from 1992 to 2023). The primary focus was on literature from the UK, but due to the dearth of literature in the UK, studies from outside the UK – Europe, North America and Australia – were also included. The keywords used were formulated in partnership with relevant stakeholders and included terms relevant to each theme and to that stage in the CJS. The full lists of search terms can be found in Annex A.

Table 1.1: List of databases utilised

Bibliographic databases	Web of Science; PsycInfo; Criminal Justice Abstracts; National Criminal Justice Reference Service (NCJRS); Medline (only used for Forensics/DNA)
Stakeholders and relevant organisations	CPS; National Police Library/ College of Policing; Ministry of Justice; Home Office; Office of National Statistics
Online search engines	Google Scholar

1.4 Definitions and focus

Reference is made in this report to both the ‘upstream’ and ‘downstream’ within the CJS. These terms are not defined consistently in the literature. In this report, the term ’upstream’ relates to the police and the CPS/prosecutors, while ‘downstream’ relates to the courts and prisons. When we use the term ‘demand’ in this report, we are mainly referring to the number of suspects/cases which flow through the system. We are primarily interested in assessing how the introduction of these innovations impact the numbers of cases charged, suspects proceeded against, or individuals convicted (or, alternatively, the likelihood of these outcomes being achieved).

To be relevant for this literature review, the studies and reviews we focused on looked at the impact of BWC and DNA on:

CJS upstream: charges, detections, clearance, arrests (but only if charging or downstream outcomes were also covered in the analysis (see below), and cases accepted by the prosecution)
CJS downstream: suspects proceeded against at court, cases going to trial, convictions, prison places or sentence length

Some of the studies looked at the impact of BWCs on complaints against the police, as well as reporting on other CJS outcomes. A decision was made not to include findings on complaints because these will not always lead to downstream outcomes or demand.

We are also interested in other outcome measures, where they appear in the reviewed studies, which might yield further insights into the pressures on different elements of CJS. These included the likelihood of guilty pleas by defendants, or decisions around sentencing length. Finally, we are interested in study measures which may be indirectly associated with demand on the downstream. For instance, the impact that these technologies have on case processing times. If new technologies lead to faster case processing times, they may reduce victim attrition, increasing successful outcomes. Conversely, cases may have to queue for limited technological processing capacity, slowing down case progression and losing victims. Finally, we are interested in other case characteristics which may have relevance to outcomes (for example, sentencing) such as composition of the offender cohort.

1.5 Selection and screening of studies

The selection of articles was made using a quality assessment framework which was loosely based on the Maryland Scale (Sherman et al., 1998) (see Annex B). This generated the selection of key and secondary studies for each hypothesis.

First, in terms of the relevance to this research, particular emphasis was placed on whether a study was focused on DNA or BWC and on the subsequent impact on outcomes leading to upstream or downstream demands. The studies could be about an intervention, a change in context (for example, increase in resources or change in crime mix), policy or practice, or on any offence type and whether the evidence was sufficient to establish causal relationships between the impact of DNA or BWC on relevant outcomes. Studies were also prioritised if they measured ‘hard’ CJS outcomes such as changes in charges, detections, prosecutions, convictions, or sentences/disposals. Studies on arrests were included initially, but because arrests do not always lead to subsequent CJS outcomes (for instance, if the police make an arrest but then do not have sufficient evidence to charge or if they arrest the wrong suspect), these studies had to also include the impact post-arrest; for example, on charges or convictions.

Studies were then defined as key studies if they were deemed to be of both high quality and relevance to the research question ( see Annex B). In terms of quality, this tended to include randomised control trials (RCT), quasi-experimental designs, and time series and panel data designs (both utilising longitudinal data). Studies were defined as secondary studies if they were assessed as being of lesser quality and/or relevance. These tended to be studies that had some direct relevance to the research questions, but were methodologically weaker in design, or from which researchers could not infer causality. A summary of the secondary studies for each theme are presented in Annex C and Annex D. The remaining papers read, reviewed, and screened from the initial searches were excluded due to their poor quality and/or lack of direct relevance (for example, little information on CJS outcomes, cross-sectional research with little or no comparison data).

Stages undertaken to select and screen articles:

Keywords were drafted and discussed with Home Office and stakeholder partners.
Preliminary searches undertaken using key words.
Feedback to the Home Office on the literature/studies found.
Key words were revised in accordance with the hypotheses (see Annex A).
Search of the databases was undertaken and articles were selected based on independent screening of abstracts by 2 researchers.
All literature/research was read and filtered/screened independently by at least 2 to 3 researchers who selected articles in terms of relevance and quality using the developed criteria (see Annex B).
At the end of the filtering, the researchers went back and checked for consistency within and across themes in terms of matching criteria.
Once quality had been assessed, the articles were arranged into key and secondary studies in consultation with the Home Office and stakeholder partners. These were then compiled in EndNote.

1.6 Search results

Forensics

Initially the approach was to search for studies on ‘forensic science’ more widely. This threw up useful research literature which examined the impact of forensic evidence on case outcomes in general but did not assess the impact of a specific forensic science technique. These studies provided a useful background to our enquiries. However, it quickly became apparent, given the wide range of forensic science techniques used, that the focus needed to be narrowed.

When searching for forensic studies of interest, studies which examined the impact of trace DNA on demand were most common. Trace DNA is DNA which has come from an unidentifiable body fluid (such as blood, sweat, saliva and semen) (Hoffmann et al., 2024). It is ‘pivotal as a vestige, or remnant, of an investigated activity’ (Roux et al., 2022). Other studies (for example, Doleac et al., 2017) have looked at the combination of the expansion in using trace DNA with the use, and expansion, of national DNA databases. So, this review includes studies which examined the downstream impact of forensic sciences more specifically, DNA-only studies as well as the expansion of DNA databases. The key studies focus only on DNA studies.

We use the term ‘trace DNA’ rather than ‘DNA evidence’ because it is the trace itself which when fully exploited by the forensic scientist, can lead to suspect identification, help the police investigate and progress a case as well as having other problem-solving potential. It is the forensic scientist’s role to decide whether the traces become ‘proofs’ for the decision maker at the CPS or court (Roux et al, 2022). Further, it is not the fact that DNA is present and recovered from a crime scene that is important but rather what the forensic scientist can elicit from the trace that is of most benefit in helping our understanding of the impact of this evidence on court outcomes (that is, the impact on the downstream CJS). See Hoffmann et al (2024) for a detailed explanation on the utility of trace DNA.

Searches across the selected databases yielded 146 initial studies (minus any duplications) that fitted the inclusion criteria. At this stage, 2 of the researchers checked the abstracts and/or executive summaries to remove any studies that were not primary empirical research and checked that the paper had a sufficient focus on downstream demand impacts. This initial screening reduced the search results to 65 studies for more detailed review. These papers were read in full, and notes were made on the sample size, design/method, and the outcome variables. The full research team then reviewed these studies together to assess both the relevance and quality of the research using the criteria outlined in Annex B. This process identified: 4 review papers, that is, systematic reviews, rapid evidence reviews or traditional literature reviews; 5 key studies; and 28 secondary studies (see Table 1.2). Although the initial search included both broader forensics research and trace DNA specifically, all 5 primary studies looked specifically at outcomes associated with trace DNA.

Body-worn cameras

As BWC technology is a relatively recent development, the research literature did not go back as far as the literature on forensics/DNA. Although the primary focus of this theme was on the use of BWCs in the UK, we also included some relevant literature from the US, Canada, Australia, and New Zealand. A total of 105 studies were initially identified from the selected databases. At this stage, 2 of the researchers checked the abstracts and/or executive summaries to remove any studies that were not primary empirical research and to check that the paper had a sufficient focus on downstream demand impacts. This initial screening reduced the search results to 36 studies for review. These papers were read in full, and notes were made on the design/method, sample size, and the outcome variables. The full research team then reviewed these studies together to assess both the relevance and quality of the research using the criteria outlined in Annex B. This resulted in the identification of 6 reviews, 12 key studies and 10 secondary studies (see Table 1.2).

Table 1.2: Search overview

Action and findings	BWC	Forensics
Studies returned and screened	105	146
Studies reviewed and read	36	65
Number of review papers	6	4
Key studies identified	12	5
Secondary studies identified	10	28

Searches were restricted to the past 30 years (January 1992 to December 2023).

Question 3

2.  Use of trace DNA by the police and prosecutors: Impact on the downstream CJS outcomes

Accepted Answer

Forensic scientists attempt to piece together the clues from an event under investigation based on the traces left behind from sources such as DNA, fingerprints, digital information and tool marks. The trace may help the scientist understand whether, when, where and how someone was present, and the sequence of events that took place. The clues gleaned from these trace findings then feed into the police investigation, intelligence, and court proceedings. The findings from these trace analyses only become evidence if they help the court decide the outcome of a case (Margot, 2017).

Forensic science can include any biological and physical traces collected from a crime scene, such as DNA, fingerprints, biological samples (for example, blood, semen), blood splatter patterns, ballistics, footwear markings, clothing fibres (Challinor et al., 2021). It can also include digital evidence such as that obtained from computer devices and mobile phones. Our primary emphasis was on the impacts of trace DNA. However, many studies of the downstream impact of forensic material, including some of the earliest studies in this area, do not disaggregate forensic science traces into specific types. We therefore present findings from studies of forensic science more broadly – mainly for what we have defined as secondary studies – and from studies which focus on traces of DNA. The latter dominates our key studies.

Digital forensic evidence was also of interest initially, due to the increased prominence of phone and electronic evidence in recent years (Wilson-Kovacs, 2021). Preliminary searches, however, revealed a lack of high-quality empirical work on downstream impacts in this area to date.

2.1 The use of DNA in criminal investigations

Using DNA profiling as part of a criminal investigation began in the UK in the early 1980s with the work of Sir Alec Jeffries. In 1983 and 1986, 2 young girls were raped and murdered in Leicestershire, and the DNA trace sample obtained from both crime scenes was successfully matched to the DNA profile of Colin Pitchfork, who was convicted for these offences in 1987. This was the first successful conviction that used DNA profiling (McDonald and Lehman, 2012). DNA profiling is now utilised in criminal cases across the world, and DNA traces are often recognised as the ‘gold standard’ of forensic science (Arnaud, 2017).

Implemented in 1995, the National DNA Database (NDNAD) is one of the oldest and most extensive DNA databases in the world (Amankwaa and McCartney, 2021). It is now run by the Home Office on behalf of UK police forces and holds over 6 and a half million individual profile records (Home Office and NPCC, 2022). In the year ending March 2021, DNA material from over 242,000 crime scenes was held on the database. The Home Office and NPCC (2022) estimate that there is a 66% chance overall that any sample uploaded from crime scene evidence will yield a suspect match and for some offence types, the match rate is higher (Home Office and NPCC, 2022).

In the year ending March 2021, this led to approximately 21,000 crime-scene-to-suspect matches. DNA trace analysis is also useful to link multiple crimes together when the same DNA profile is recovered from more than one crime scene. In the year ending March 2021, 814 crime scene DNA profiles were successfully matched to other crime scenes (Home Office and NPCC, 2022).

Although initially DNA trace analysis was utilised only for serious crime, especially homicide, it is now applied to a wide range of crime types. DNA profiles from burglary and vehicle crime along with sexual and violent offences against the person are the offences most likely to be uploaded onto the NDNAD. For example, in the year ending March 2021, 39% of DNA recovered from crime scenes were from burglaries (Home Office and NPCC, 2022). Some very common crimes, such as fraud, theft, and computer crimes, however, generate very few samples for loading onto the NDNAD.

While there are some useful applications of DNA trace analysis in individual cases (often those lacking any clear suspect, those deemed ‘hard-to-solve’, and/or ‘cold cases’), reviews of the literature tend to conclude that the overall contribution of DNA traces to positive case outcomes (for example, arrests, prosecutions, and convictions) is low as DNA traces are collected, analysed and loaded into the NDNAD in only a small proportion of total recorded crimes (Amankwaa and McCartney, 2021). Despite this, for some specific types of offences where DNA traces are more commonly retrieved (that is, burglary, stranger homicides and sexual offences) the use and analysis of DNA profiling can clearly be important in identifying suspects.

Analysis of DNA traces and the matching of DNA samples to a DNA database is now quite standard police practice in many countries. See Table C.1 in Annex C for a summary of relevant literature reviews.

When reviewing the literature, it was anticipated that studies looking at the impact of traces of DNA in the CJS would focus on case progression and CJS outcomes in several different ways. In summary, it was considered that traces of DNA, might:

aid in linking crimes to suspects and in linking different crimes to the same offender, which may lead to more crimes being solved, and therefore more suspects being arrested and charged
enhance the probative value of evidence, which may lead to further case progression and a higher volume of convictions
potentially impact on the type of conviction, as a higher evidential weight may be associated with a greater proportion of guilty pleas, saving time and money
either slow down or speed up time to case completion
in addition, the existence of DNA databases in combination with the collection and loading of trace DNA, could have a deterrent effect both on offender recidivism and a wider impact on crime rates

2.2 Summary of the secondary studies

There is an extensive literature examining the effects of forensic traces on case outcomes across different crime types and different countries. Some of this work has reported strong findings in terms of wider forensic science impacts. These studies tend to indicate cases with forensic material are associated with more arrests, prosecutions, guilty pleas, longer sentences and convictions.

Peterson et al. (2013) analysed data on aggravated assaults, rapes, homicide, robberies and burglaries (all these offence types were combined) from 5 areas in the US. They looked at the outcomes for these offence types depending on whether forensic evidence (not DNA specifically) had been collected.

They found positive significant effects from forensic evidence on arrests, prosecution referrals, charges and convictions. For example, cases were charged in 32% of cases where forensics were collected (compared with 14% where none were taken). Other researchers have disaggregated the data into specific crime types and have found a more mixed picture of the impact of forensic evidence on outcomes. In quasi-experimental studies in Australia, testing cases with and without DNA/DNA database expansion, it was found that DNA evidence improved the charge rates with more cases progressing to prosecution in property crimes (burglaries and vehicles offences) (Briody, 2005, 2006), robberies (Dunsmuir et al., 2008), homicides (Briody, 2004, 2005), and sexual offences (Briody, 2002, 2005; Dunsmuir et al., 2008). But other US studies have failed to find any significant effects of forensics on charge rates and prosecutions (Baskin and Sommers, 2010, 2011, 2012; Schroeder and Elink-Schuurman, 2017).

Most of the research in the area relies on cross-sectional or quasi-experimental designs without adequate control groups. These designs are problematic as they often provide a ‘snapshot’ comparison of forensic versus non-forensic cases at one point in time, in one or more places - they do not control for other differences that are likely to exist between those cases. Causality cannot therefore be established. Many of these studies also fail to unpack different types of biological trace analysis. And they often fail to ascertain whether trace material was just collected but not analysed or, if analysed, its utility in leading to a positive outcome (see McEwen and Regoeczi, 2015).

Furthermore, some of these studies’ findings have failed to be replicated by other researchers. Looking again at Peterson et al.’s studies (2010, 2013), the authors reported further case progression, more guilty outcomes, and longer sentences when comparing ‘forensic’ cases with ‘non-forensic’ cases for the combined crime types, as described above. But researchers from within the original research team (Baskin and Sommers, 2010, 2011, 2012), and later independent researchers (Nir and Griffiths, 2018; Steele, 2020), have both failed to fully replicate those findings in their own re-analysis of the data. Baskin and Sommers (2010, 2011, 2012, 2013) criticised the original findings as failing to correct for selection bias (treating each case progression event as separate and not interconnected) and relying on aggregated crime-type data.

In their re-analysis they found no significant effects for forensic evidence and highlight that the best predictor of case progression is victim and witness participation. Similarly, in a review of 600 homicide cases in the US, Schroeder and White (2009) found that detectives did not tend to rely on DNA evidence in their investigations, only submitting it for analysis in the absence of other lines of enquiry or collecting it in the anticipation that it would strengthen a prosecution at trial.

Therefore, while there is considerable research which attempts to examine the impact of forensics and, in some cases, traces of DNA, on case progression and outcomes, the findings are inconsistent and complex. Analysing differences by crime type is important. Some studies and reviews suggest DNA is useful at the investigative level, providing corroborative evidence and indirectly impacting on outcomes.

Wilson et al. (2011) undertook a Campbell Review of the effectiveness of traces of DNA and noted only one study (Roman et al., 2008) which they felt significantly contributed to the field while being methodologically robust. True experimental designs are, however, difficult to apply here, as the use of DNA in crime scene analysis has become more established and there are subsequently challenges about randomly assigning cases to conditions where DNA material would not be analysed. Wilson et al. (2011) therefore included some of the better quasi-experimental work in their review, highlighting the need for good quasi-experimental design to use adequate control comparison groups. They also noted the difficulties in establishing a causal relationship between the presence of DNA and case progression outcomes.

The strength and consistency of the impacts of DNA traces on case progression varies greatly by crime type (Bradbury and Feist, 2005). Wilson et al. (2011) concluded that DNA traces can be effective in increasing case outcomes, especially in the investigation of high-volume crime that tend to have low clearance rates. For more serious crimes, there is some evidence that DNA traces aid case progression, but the evidence is much weaker given a dearth of high-quality data in this area.

Quasi-experimental research evaluating various interventions for victims of sexual assault in the US and, to a lesser extent, the UK, can also provide insight into the potential impact of DNA trace evidence on outcomes. Overall, evaluations tend to show that involvement in these interventions, that is, not necessarily the DNA evidence itself, leads to higher charge volumes. In the US, the research is focused on Sexual Assault Nurse Examiners (SANEs), Sexual Assault Response Teams (SARTs) and Sexual Assault Medical Forensic Exam (SAMFE) programmes. In the UK, Kelly et al. (2005) evaluated the impact of Sexual Assault Referral Centres (SARCs) which support the victim through the CJS process as well as facilitating more timely recovery of high-quality DNA material. Because these interventions are multifaceted, it is impossible to determine the effects of DNA evidence alone.

Campbell et al. (2014a) evaluated SANE programmes across 6 sites, comprising almost 1,700 sexual assault cases, and concluded that cases with SANE intervention were statistically significantly more likely to be prosecuted. In the UK, Kelly et al. found rape cases were more likely to be discontinued during the investigative level if victims were not given SARC intervention, and 2 of the 3 SARC areas had higher conviction rates compared to the control areas (2005). However, an evaluation of the SAMFE programme, and not the whole intervention, concluded there were no direct effects from the DNA trace on case outcomes (Kettrey et al., 2022). The authors conclude that the DNA traces have indirect effects on case progression in terms of evidential weight, which may encourage the police and prosecutors to pursue the case.

For more detail on the secondary studies, see Annex C.2.

2.3 Key studies

As much of the research in the area fails to demonstrate strong causal links between traces of DNA at a crime scene and subsequent CJS outcomes, in most cases the findings are merely indicative of an association. As Wilson et al. (2011) conclude, for some crime types there are quite large positive effect sizes for DNA trace effectiveness (as well as negative relationships), but in most of the research there are plausible alternative explanations so the findings are not necessarily evidence of a direct effect of the traces of DNA on case progression. There are many other factors about the incident, the crime scene, the victim and offender, and investigative processes that are also likely to have significant impacts on case outcomes – it is unlikely that the only difference between cases solved and unsolved is the presence of forensic material.

The strongest evidence for the effects of trace DNA on case progression and in helping to increase CJS outcomes comes from one stand-out piece of experimental work by Roman et al. (2008). This study examines the impact of the presence, retrieval, and processing of DNA material on CJS outcomes while controlling for other confounding variables by randomising the allocation of cases to DNA trace analysis in the first instance. As it focuses specifically on volume crimes, in particular burglaries and other acquisitive crimes, any impacts on the demands going through the CJS are likely to be greater given the larger proportion of these crimes, and the increased likelihood that they yield DNA traces at the crime scene. Given the identifying strength of DNA and the ability of investigators to link suspects to crime scenes as well as connecting crime scenes through DNA databases, this section also includes 3 further studies that reliably address the issue of the deterrent effects of DNA databases (Anker et al., 2021; Bhati, 2011; Bhati and Roman, 2014; Doleac, 2017).

The study by Roman et al. (2008) examined the effects of a US field experiment whereby the police expanded DNA retrieval to non-violent crimes. Crime scenes across 5 US pilot jurisdictions (Orange County and Los Angeles, California; Topeka, Kansas; Denver, Colorado; and Phoenix, Arizona) were randomly assigned to either a test condition (initial DNA analysis) or a control condition (a standard non-DNA analysis for the first 60 days of the investigation). All 500 crime scenes included in the RCT had yielded traces of DNA, and rather than prevent the analysis of these traces in the control conditions, they merely delayed the analysis for 60 days and relied on traditional investigative techniques during that time. The aim was therefore to establish whether timely retrieval and processing of forensic traces improved case progression. The focus was on high-volume property crime which included residential burglary, commercial burglary, and theft from motor vehicles.

The collection of traces of DNA and their analysis had not been used in the investigation of volume crimes prior to this trial, and clearance rates of these crimes are typically very low. They compared the case outcomes in terms of whether the suspect was identified, whether an arrest was made, and whether the case went to trial. They also examined the cost-effectiveness of DNA analysis for volume crimes. They were unable to test the effects further downstream in terms of case outcomes as the DNA trace processing took longer than they had initially expected, and cases had therefore not been resolved by the end of the experiment. Consequently, final downstream outcomes, such as convictions and sentencing, were not examined.

The retrieval of traces of DNA from non-violent crime scenes was promising, with 78% of test cases resulting in a DNA profile that could be uploaded. Out of the cases that were uploaded, 43% of them yielded a hit on the database, and from this a suspect match was made in 88% of those cases. Seventy percent of suspects identified were later arrested, and of those arrested, 90% were prosecuted. Table 2.1 below summarises the key findings on the CJS outcomes of interest to this review from the Roman et al. study.

Table 2.1: Selected CJS outcomes for burglary and theft from vehicle offences: DNA versus control cases from all experimental and control sites

Findings	Experimental cases: DNA traces present and tested	Control cases: DNA traces present but not initially tested
Percentage of suspects identified	31%	12%
Percentage of suspects arrested	16%	8%
Number of arrests	173	86
Percentage of cases accepted for prosecution	19%	8%
No. of suspects identified who had prior felony arrests	5.6	1.7
No. of suspects identified who had prior felony convictions	2.9	0.9

Table constructed from the Roman et al. report - pp.140 and 146.

As can be seen from Table 2.1, cases in which DNA traces had been processed and analysed from the start showed more positive results in this area. They were statistically significantly more likely to lead to a suspect identification (31% of cases compared to 12% in the control group). Arrests were twice as likely to be made in the test condition, with a total of 173 arrests across the 5 sites in the test condition, compared with 86 in the control condition (16% of cases in the test versus 8% in the control). More than twice as many cases with early DNA analysis were also accepted for prosecution. They could also compare types of forensic analysis and found that DNA trace analysis was more effective than fingerprint trace analysis. Overall, traces from DNA were 5 times more likely to result in identifying the suspect, and 9 times more likely to result in an arrest, but fingerprint traces were collected in a much smaller sample of cases.

They therefore also analysed only cases where both types of forensic traces were obtained. The results were significant though more modest in proportion - suspects were identified by their DNA twice as often as by fingerprints (in 16% compared to 8% of cases) and arrested 3 times as often (9% of cases compared to 3%).

The researchers also found that the suspects identified through DNA trace analysis had markedly different criminal profiles to the control suspects. As they were already in the DNA database prior to the wholesale expansion to volume crimes, these offenders would have also perpetrated violent crimes. Analysis showed (see Table 2.1) that they had significantly more (at least twice as many) felony arrests and convictions, they also had significantly more misdemeanour arrests and convictions. They were therefore more prolific and versatile serial offenders than those suspects identified using more traditional methods. As there are very high recidivism rates among burglary offenders, the use of DNA traces to arrest and prosecute these most prolific offenders, it was argued, can therefore help to interrupt their criminal career and prevent future re-offending.

The expansion of DNA trace analysis to include high-volume property crime did successfully increase the number of suspect matches, arrests and prosecutions. These increases create additive downstream demand with more offenders potentially entering the CJS, and going to trial, although in this study, they were unable to test the extent to which traces of DNA impacts on convictions and sentences. As more prolific serial offenders were identified, the potential downstream impacts are more complex. As ‘DNA-present’ offenders had more serious criminal histories, this may well lead to more custodial and potentially longer sentences, their incapacitation also prevents them from committing further crime for the duration of their sentence.

The researchers also estimate the additional costs that DNA analysis adds to an investigation but highlight that these cases might not be solved at all using a traditional, non-DNA trace investigation. They do also, however, warn about the potential for backlogs in the CJS if sufficient investment is not made to increase the capacity for DNA trace analysis to match the increased demands. The large variation in success of the pilot between geographical areas also highlights the importance of best practice implementation strategies.

2.4 DNA databases: The impact of their introduction and expansion in offence coverage

The remaining key studies focus on the impact of DNA databases. These can have several indirect impacts on downstream demand on the CJS: they can either deter re-offending or increase the likelihood of detecting crimes (so called ‘probative impacts’). In total, 4 key studies were identified.

Bhati (2011) produced a government report – later written up as a journal article (Bhati and Roman, 2014) – which examined the expansion of the DNA database in Florida, US to include a much wider range of crime types. This quasi-experimental study used longitudinal data from 1996 to 2004 to compare released offenders’ patterns of re-offending over time, including pre-inclusion and post-inclusion in the DNA database. As with other research on DNA and deterrence, they did not establish precisely when each offender’s DNA was added to the database, but inferred this, using modelling techniques, from the various dates when state legislation incrementally extended the requirements for registering offenders on the DNA database.

First, they were interested in specific deterrent effects, or whether individual offenders were less likely to offend after being added to the database. Therefore, they tracked the re-offending of individual offenders to ascertain if future offending was prevented once the offenders were aware their DNA had been added to the database. It was hypothesised that if the individual offenders believed they had a greater chance of apprehension and punishment they would be less likely to re-offend compared to those offenders not on the database.

The researchers recognised that any measurable deterrent effects of being on the DNA database could be offset – or confounded – by the ‘probative effects’ of DNA databases. That is, offenders might well reduce their offending rate because of being on the database, but the probability of their detection for any given offence would actually rise. They therefore modelled the probative effects and separated out these effects from deterrent effects on recidivism using ‘multiple clock’ models in their analysis. Re-offending of offenders was measured as ‘time to failure’ or the time it takes the offender to commit another crime. Offenders were tracked for 3 years after their release from prison.

Furthermore, they considered that there were probably a lot of other differences between offenders on the database and those who were not, so the analysis controlled for criminal career trajectories, measured other offender characteristics, and used weighting to control for effects. As changes occurred incrementally over a period of years dependent on when specific crime types were included, they compared crime types at slightly different time periods. The analysis compared the expected effects on their re-offending if no intervention (change in law) had occurred and compared that with observed changes. As the follow-up period was 3 years, they only looked at medium term recidivism, looking for any post-intervention changes in their criminal career while also controlling for any expected changes due to desistance (that is, maturation and ageing). The sample was stratified by crime type and covered violent offences, robbery, burglary, other property offences, drug-related offences, and a residual ‘other’ category.

The specific deterrent effects on re-offending varied greatly between crime types. For violent offences, other property offences and ‘other’ offences recidivism increased, but this change was not statistically significant. There was no change to drug-related crime. Recidivism for burglary and robbery offences, once the probative effects were controlled for, did significantly decline. These offences were chosen because they considered them to be the most likely to leave easily obtainable DNA traces at the crime scene. The results suggest that being included on the database for robbery and burglary did have a specific deterrent effect on re-offending, although the size of the effect was modest with a predicted 2 to 3% reduction.

While there was only a modest significant impact on re-offending for 2 crime types, the probative effects were much greater. They measured both the likelihood of re-arrest and reconviction and found that offenders from all crime types (except for ‘other’ crimes) were between 30 to 40% more likely to be re-arrested and (to a lesser extent) reconvicted once on the DNA database. They concluded that the probative effects of being on the DNA databases were therefore much larger than any potential deterrent effects. Although some decreases in demand might be associated with small reductions in re-offending for robbery and burglary, the larger probative value of DNA traces and subsequent analysis are much more likely to lead to downstream increases in demand, especially in terms of offenders being re-arrested for further crimes within the next 3 years following release.

Doleac (2017) examined the deterrent impacts of the expansion of DNA databases across 7 states in the US. Similar to Bhati (2011; Bhati and Roman, 2014) the states’ databases were expanded incrementally over a period of years with the gradual inclusion of more crime types. In this quasi-experimental design, Doleac used longitudinal data from 2000 to 2010, to map re-offending of offenders for 5 years post-release and only included felony (that is, more serious) offences such as homicide, forcible rape, aggravated assault, robbery, burglary, larceny, vehicle theft, and arson. As the changes happened at different times in different states dependent on legal changes for the inclusion of new crime types, the author was able to compare recidivism changes between states for over a decade.

Doleac was unable to check whether a specific offender’s DNA was included in the database, but as with Bhati (2011) Doleac used the timing of mandatory inclusion per crime type to infer that they were included. While recognising the probative effects of DNA traces, the researcher did not attempt to model those effects specifically.

The study measures deterrence in 2 different ways. Firstly, there was an interest in the specific deterrent effects on individual re-offending – so whether the knowledge that an offender is included in the database deters them from committing future crimes due to the fear of apprehension, arrest, and punishment. Secondly, it was also proposed that DNA database expansions would have a general deterrent effect, that is, knowledge that the state has an inclusive DNA database will impact on all offenders’ willingness to commit crime, not just repeat offenders but first-time offenders, and other offenders who have not had their DNA collected and added to the DNA database. It was also speculated that DNA databases will have a larger net effect as a law enforcement tool. All of these wider societal changes should then be evidenced as a reduction in crime rates within the state.

Trying to ascertain a general reduction in crime and inferring a causal relationship to the database expansion, is however, more difficult, as many other wider changes impact on crime rates. The fact that changes happened at different times in different states, however, did allow Doleac to make some useful control comparisons. The author also controlled for many other potentially confounding variables in the models, including seasonal variations, criminal histories, and offender demographics.

Doleac found that the requirement to submit a DNA sample appeared to significantly lower recidivism of serious crimes, but again, the effects differed by crime type. In their final models, controlling for the widest range of other factors, recidivism for serious violent crime fell by 21% over the 5 years (p=<0.01). Recidivism for serious property offences fell by only 3% and this was not statistically significant. Although Doleac did not control for probative effects as effectively as the work by Bhati (2011; Bhati and Roman, 2014), Doleac did test to see whether DNA databases had a significant effect on arrest rates within the state but saw no significant changes. It was concluded that the reduced recidivism is more likely to be a specific deterrent effect than an evidential one.

In terms of any general deterrent effects on state crime rates, again, Doleac presented a range of models controlling for extraneous variables. In their final, best-controlled model, Doleac found that both serious violent crime and serious property crime offences significantly decreased which was consistent with the changes in state laws. These general deterrent effects were, however, quite small in magnitude and had large confidence intervals (suggesting that the estimates were imprecise). Nevertheless, Doleac estimated that for every additional DNA profile per 100,000 state residents, property crimes fell by between 5 to 35% (dependent on the type of property crime) from 2000 to 2010, whereas violent crimes fell by between 4 to 45%.

Doleac (2017) found evidence that DNA databases across different US states had both specific deterrent effects on violent re-offending rates and general deterrent effects on both violent and property crime rates. This suggests that DNA databases can reduce downstream demands, but the size of these impacts is quite minimal.

Although not included as a key study, it is worth noting that a related unpublished paper by Doleac (2016) also found that, as crime rates fell in a state with expansions of their DNA databases, crime rates increased in nearby states. This suggested that rather than an overall deterrent effect and prevention of crime, crime may be displaced to areas where the offenders have less chance of being detected. This finding is particularly significant in the US where DNA databases differ by state.

Doleac and colleagues later conducted a similar quasi-experimental study involving the expansion of the NDNAD in Denmark in 2005 (Anker et al., 2021). Unlike the state-level expansions in the US, in Denmark the law was changed to include the DNA profiles of all offenders who had committed any serious offences (equivalent to a felony offence in the US, all crimes that would receive a maximum custodial sentence of over 18 months). This was a dramatic and rapid increase in DNA database inclusion – the proportion of offenders included increased from 4% to 40%. They compared recidivism of ‘test’ offenders (those offenders on the database) with ‘control’ offenders (who were not on the database before the pre-expansion) and tracked their re-offending for a 3-year follow-up period.

They recognised the potential confounding impacts of the probative effects of DNA, and in an enhancement on their previous study, they attempted to separate out probative effects to estimate the ‘elasticity of criminal behaviour’. This controlled for the likelihood that if a suspect commits a crime, those present on the database are more likely to be arrested than those offenders not present. They also examined offender characteristics to look for any significant differences between the test and control population.

The analysis found that DNA expansion in Denmark resulted in a statistically significant reduction in recidivism rates, and that the effects were, at least in the 3 years reviewed, large in magnitude. Offenders were 43% less likely to be arrested for a further crime in the first year after their inclusion, and 42% less likely to be reconvicted. The deterrent effects persisted for the entire 3-year follow-up period. As with their previous study, the deterrent effects were largest on recidivism for serious violent crimes – reconviction for violent crimes decreased by 63% in the first year. DNA registration also reduced the likelihood of a property crime conviction, which saw a decrease of 34% in the first year, but this was not strongly statistically significant (p=<0.10).

Offender characteristics were important, and larger reductions in re-offending were associated with first-time offenders (those whose DNA had been recorded from their first offence), those with children, and those whose initial offence was a violent crime. They also measured subsequent positive life changes post-inclusion and found that offenders on the database were more likely to enter stable relationships, education, and work, than their pre-database controls. This suggests that not only can the DNA expansion deter future re-offending, it can also encourage engagement in more positive behaviours post-release.

They could also measure and compare the relative probative effects of DNA database inclusion. Offenders were significantly more likely, after inclusion on the database, to be detected if they offended again, but the deterrent effects of expanding the DNA database were much larger. They estimated for every 1% increase in the likelihood of future detection, the risk of re-offending was reduced by 2.7%. Although the primary aim of this work was to look for specific deterrent effects, there was also some evidence of more general crime reductions post expansion; for example, there was a 48% reduction in first-time offences.

The probative effects of the DNA expansion in Denmark did lead to some small increases in downstream demands (in terms of a higher risk of detection of offenders on the database if they re-offended). These increases are, however, much smaller than the deterrent effects on serious crime. Overall, the findings suggest that the rapid expansion of the Danish DNA database to include all felony/serious offences was much more likely to lead to reductions in demand on the CJS.

The deterrence work discussed here all found some reductions in demand associated with DNA database expansion, whether that expansion is incremental (Bhati, 2011; Bhati and Roman, 2014; Doleac, 2017) or happened with dramatic changes at one point in time (Anker et al., 2021). There is evidence for both specific and general deterrence. DNA databases appear to reduce the likelihood of re-offending by individual offenders present on the database, but they may also have a wider deterrent effect on crime rates.

All the research studies looked at more serious offences, but they still found differences between crime types. Bhati (2011; Bhati and Roman, 2014) only found significant reductions in violent crime recidivism. But they also found that these reductions were small compared to the probative impacts of DNA. In contrast, Doleac (2017) found significant, but small effects, for both serious violent and property crimes. Anker et al. (2021) found much larger effects that were more powerful than the probative effects of DNA. None of the research to date has, however, looked at potential longer-term deterrence effects.

Table 2.2 A summary of the findings on demand from the key and secondary studies

The impact of DNA traces on…	Increase/decrease or neutral demand on CJS	Offence types impacted	Evidence base
Suspect identification Arrest Acceptance for court (DNA retrieval)	Increase	High-volume property crime Burglary Theft from vehicle	Robust RCT – US (Roman et al., 2008). Cases with DNA traces were twice as likely to identify a suspect, end in an arrest and be prosecuted than fingerprints. More prolific offenders identified.
Recidivism of offenders (deterrent effect) (DNA database expansion)	Decrease	Burglary Robbery Serious violent crimes	Quasi-experimental – moderately robust evidence. US studies (Bhati and Roman, 2014; Anker et al., 2021; Doleac, 2017). Slight impact – 2 to 3% reduction in re-offending for those on DNA database (Bhati and Roman, 2014). 43% reduction in re-offending of those on database (Anker et al., 2021). 21% reduction in re-offending for those on the database (Doleac, 2017).
Re-arrest Reconviction (DNA database expansion)	Increase	Burglary Robbery Violent offences Drug-related crime Other property offences Serious violent crimes	Quasi-experimental – moderately robust evidence. US studies (Bhati and Roman, 2014). Large increase of 30 to 40% once on the DNA database –showing the probative value of DNA (Bhati and Roman, 2014).
Potential offenders’ deterrence (whose DNA is not on database) (DNA database expansion)	Decrease	Felony offences (homicide, rape, aggravated assault, robbery, burglary).	Quasi-experimental – moderately robust evidence. Probative nature of DNA traces not controlled for US (Doleac, 2017). A much smaller impact. Property crimes fell by 5-35%, violent crimes, 4 to 45%.
Cases charged and accepted for prosecution (Note, findings are from retrieval of biological traces more widely)	Increase Neutral	Property crimes Robberies Homicides Sexual offences	Weaker evidence based on quasi-experimental studies without adequate control groups. Mixed findings. Five studies found DNA improved the charge rates and had more cases progressing to prosecution (Peterson et al., 2010, 2013; Briody, 2004, 2005; Dunmuir et al., 2008). But other US studies (4) have failed to find any significant effects (Baskin and Sommers, 2010, 2011, 2012; Schroeder and Elink-Schuurman, 2017).
Charge volumes, accepted for prosecution, victim support, conviction rate. (Findings from Sexual Assault Intervention programmes)	Increase Neutral	Sexual assault	Quasi-experimental. Weaker evidence. Increase in cases being prosecuted if victim has been through an intervention programme (Campbell et al., 2014a). But an evaluation of the forensic examination found no impact from DNA on case outcomes (Kettrey et al., 2022). Broad range of interventions, so difficult to separate the effects of DNA from other factors.
Conviction rates (Retrieval of DNA)	Increase Neutral	Homicide Sexual assault/ offences	Weaker evidence. Mixed findings which differed by crime type. Other factors more likely to predict conviction (Briody, 2004, 2005; Tully, 1998).
Sentence length – longer for some crime types and shorter for others (Findings from pre- and post-DNA database cases)	Increase Decrease	Sexual offences Property crimes Homicide and manslaughter	Weaker evidence. Studies not adequately controlling for other differing factors in the cases (Peterson et al., 2010; Tully, 1998; Briody 2002, 2006).

2.5 Findings and implications

Evidence about the effectiveness of trace DNA on case progression and CJS outcomes is mixed, and many studies involve lower to moderate quality research designs; there is a significant gap in the literature in terms of high-quality experimental work; as forensic science analysis is now so well established in the field of crime investigation, it is also unlikely that further RCT research can be completed due to ethical considerations.

Some cross-sectional research in the US showed large increases in demands on the CJS in terms of numbers flowing through the CJS: more arrests, prosecutions, and convictions for cases with trace DNA/fingerprints across both aggregate crimes and a range of separate crime types. However, the results have not been replicated in re-analyses of the same datasets (and when the methodology has been replicated).

The single RCT study (Roman et al., 2008) examining the impacts of DNA traces on case progression in volume, property crimes provide strong evidence that rapid DNA trace analysis led to significantly more suspect identifications, arrests, and prosecutions. These effects are also quite large in magnitude – with twice as many arrests and prosecutions made compared to the control condition; this would inevitably lead to considerable increases in downstream demand, with more suspects entering the CJS and more cases being charged and prosecuted. They were not, however, able to track cases further in terms of case outcomes.

The profile of suspects identified through DNA analysis is also markedly different. Roman et al. (2008) found that they had at least twice as many arrests and convictions for both felony and misdemeanour crimes. They were therefore prolific and versatile offenders, committing a volume of crime across different crime types and severities.

Studies exploring changes in DNA databases, where comparisons are made between cases using traces of DNA and those without, have found the likelihood of charges and prosecutions increase for property crimes, robberies, homicides, and sexual offences. Less evidence exists for increases at other stages of the CJS.

Studies involving the evaluation of sexual assault interventions are designed to both provide timely and high-quality DNA traces. These studies make pre- and post-intervention comparisons, but they are unable to unpack the specific effects of DNA traces alone. It is likely, however, that the effects of DNA traces are indirect. For example, these interventions may primarily encourage victim engagement, while both the victim participation and substantive weight of the presence of DNA combine to impact on further case progression and better CJS outcomes.

Quasi-experimental work examining the specific deterrent effects of DNA databases have shown that they can reduce downstream demands by reducing re-offending. Research in the US showed modest but significant effects have been found for both violent crime (Doleac, 2017), robberies, and burglaries (Bhati, 2011; Bhati and Roman, 2014). These impacts are, however, small in magnitude (Bhati, 2011; Bhati and Roman, 2014). In Denmark, where a more dramatic and rapid expansion of the NDNAD occurred, effects were stronger in magnitude, especially for serious violent crimes (Anker et al., 2021).

Limited evidence also suggests that the expansion of DNA databases to cover a broader range of offences may have a significant crime reduction effect (a broader deterrent effect on the population) on both violent and property crime rates. But again, these effects, while statistically significant, are very small in size (Doleac, 2017).

There was also a probative impact from the DNA database expansion studies where offenders with DNA on the database were linked to other crime scenes and were reconvicted. Bhati and Roman (2014) found a larger probative effect than deterrent effect. But other studies found the reverse and that the deterrent effect was stronger than the probative effect (Doleac, 2017; Anker et al., 2021).

In summary, trace DNA can increase the volumes of cases progressing through the CJS.The evidence is particularly strong for its impact on volume crimes. Cases with DNA traces are likely to have an increased demand impact on the CJS in terms of more detections and arrests, more suspects entering the courts, more prosecutions and potentially more convictions.

The effects of the impacts are also additive, so there is the cumulative increase of demands at all stages of the CJS. The most robust evidence we have also suggests that trace DNA impacts most on high-volume offences such as burglaries. Any increases in demand therefore have greater effects due to the high proportion of these crime types. And while DNA traces are not collected from crime scenes on a large scale, and their overall impact is modest, for some offence types (burglary) and some sub-categories of offences (stranger offences), it can be important.

There may additionally be some reductions in demand from the deterrent effects of DNA databases on the future offending of released offenders whose DNA is recorded.There may also be some general deterrent effects on subsequent crime rates, although these effects are difficult to isolate from other changes in crime patterns over time.

Overall, cases with DNA traces appear to have some impact on increasing the numbers of detections, charges, prosecutions and convictions, and thus increasing the number of offenders entering the system. Convictions are also associated with higher sentence lengths – because of being able to link crime scenes to potentially prolific offenders leading to prosecutors laying more serious charges. However, the impact is highly variable by offence type.

Question 4

3.  Use of new technology: body-worn cameras and the relationship with CJS outcomes

Accepted Answer

This section focuses on the use of BWCs by the police, examining how changes in BWC technology and its use might affect CJS outcomes and subsequent demands.

3.1 Use of BWCs

BWCs have been one of the most rapidly diffused police technologies this century, albeit one that involves huge financial investments in terms of equipment, training, and infrastructure (Lum et al., 2020). National Police Chiefs Council (NPCC) guidance explains that BWCs are small, visible devices attached to an officer’s uniform used to capture both video and audio evidence as an incident unfolds. The BWC “acts as an independent witness to police actions and interactions in an unbiased and secure recording” (NPCC 2022:8).

BWC use in England and Wales started on a trial basis in 2006. There then followed a period of more widespread force level trials. For instance, Hampshire Police initiated a large-scale project called Operation Hyperion in 2013 where they equipped every frontline police officer on the Isle of Wight with a personal issue BWC device (Jameel and Bunn, 2015). The first trial with a robust evaluation design on the use of BWC for domestic abuse incidents began during 2014 in Essex (Jameel and Bunn, 2015). A 2017 His Majesty’s Inspectorate of Constabulary and Fire & Rescue Services (HMICFRS) inspection focusing on domestic abuse identified 35 forces using body-worn video cameras on a force-wide basis and an additional 6 forces as part of a pilot scheme. Only 2 forces did not use a body-worn video camera at the time of the 2017 inspection (HMICFRS, 2019).

The success of the early small-scale pilots appeared promising, and BWCs were later trialled in Canada (Toronto Police Service, 2016), the US (Ariel, 2016a; Ariel et al., 2014), and Australia (Clare et al., 2021). It was claimed that BWCs could reduce local crime rates, improve police detection rates, improve officer behaviour, and enhance police-community relations. Police forces across the UK, North America, and Australia widely adopted them, despite there being only a relatively limited and evidentially weak research base. There has since been an increase in studies assessing BWC impacts across a broad range of variables, including some high-quality RCTs and quasi-experimental studies. Recently, there have also been several narrative and systematic literature reviews (Cubitt et al., 2017; Lum et al., 2020; Lum et al., 2019; Maskaly et al., 2017).

Although the focus of much of this literature has been on how BWCs impact on police and citizen perceptions and behaviours (a summary of these studies are in Annex D) – some of which will have demand consequences – there is also a small but growing body of evidence on outcomes which have more direct CJS impacts (charges, prosecutions, and convictions). This culminated in a recent meta-analysis (Petersen and Lu, 2023). Table D.1 in Annex D summarises some of the key evidence reviews in the area.

The adoption of BWCs by the police was anticipated to generate a range of benefits, some of which were expected to lead to downstream CJS outcomes, and potentially on CJS demand. However, the relationship between BWCs and CJS demand is complex. Some of the impacts might be expected to increase demands, whereas others might work in the opposite direction. As might be expected, given the rapid introduction and diffusion of BWCs into mainstream policing, there may also be unanticipated effects. In summary, BWC has been proposed to have the following effects:

an increase in demand in terms of numbers of cases in the CJS
- BWCs may enhance evidence against suspects, resulting in the police and prosecutors charging more suspects
- BWCs may enhance evidence against suspects, which might result in more robust cases and an increase in convictions
decrease in demands due to time saved
- stronger evidence from BWCs may result in more guilty plea convictions, thus leading to savings in time and cost to the CJS (since guilty pleas require less police, prosecutor and court time, and tend to attract a sentencing discount)
complex, mixed impact if police/court capacity for processing BWC leads to a slowdown in time to charge
- BWCs may slow down case completion due to the processing of large amounts of data. This could lead to higher victim attrition, resulting in less demand

3.2 Key studies

In the following section we will explore the CJS downstream impacts in more detail using our key studies that have been screened to ensure a high-quality standard of research design (that is, RCTs and more robust quasi-experimental work), while also capturing a range of CJS outcome measures. We have also included some of the findings from a meta-analysis review undertaken by Petersen and Lu (2023). While not strictly a key study, the objectives of the review were similar to the objectives of this literature review, and we felt they were worth including in this section. Summaries of what we have considered to be secondary studies are given in Annex D. We could identity 12 key studies: 2 from the UK, 9 from the US and one from Australia.

The literature into BWC impacts has grown substantially in recent years, and while some of the early intervention evaluations suffered flaws in design which potentially impacted on the generalisability of their findings and reliability, there are now several good quality experimental and quasi-experimental studies. There are still gaps, however, in terms of number of peer reviewed studies with robust evaluative methods, while some of the grey literature can be somewhat lacking in terms of design specifics and depth of analysis.

Furthermore, in terms of CJS impacts, many of the high-quality RCTs in the field have focused on police and public behaviours (such as police complaints, use of force, and assaults against the police) which do not necessarily lead to impacts on the CJS downstream. And although some of these studies capture the impact on arrest volumes, they rarely measure downstream impacts post-arrest. Some of this work is presented in Annex D, Table D.2, while Table D.3 gives a more detailed summary of our key studies.

Katz et al. (2014) and Morrow et al. (2016) present the findings from a small-scale pilot of BWCs in Maryvale, Phoenix, Arizona in the US. This involved the implementation of 56 BWCs over 15 months. The study had a quasi-experimental design and involved both pre- and post-intervention comparisons, as well as comparisons between the test precinct in Phoenix and other non-BWC areas. As an unadjusted quasi-experimental design, this work is methodologically weaker than some of the other key studies, but the authors present a range of comparisons, and unlike many other studies present both the findings in terms of aggregate crimes (Katz et al., 2014) and domestic violence (DV)/ intimate partner violence (IPV) (Katz et al., 2014; Morrow et al., 2016).

It is, however, important to bear in mind that unadjusted quasi-experimental studies are likely to contain a higher level of contamination bias (as officers in the BWC and non-BWC group were sometimes sent to the same incident) and may overestimate effects. Like most studies using BWC comparisons, the results are based on an officer level of analysis, comparing those officers with BWCs to those without, rather than activation of BWCs in specific incidents. The researchers also felt that some officers self-selected on whether or not to participate by transferring to a different precinct once they knew Maryvale was running the pilot.

The Katz et al. (2014) report presented aggregated crime data for arrests and found that BWC officers made significantly more arrests (17% increase compared to 9% in non-BWC). Aggregated crime data can be difficult to interpret as BWCs are likely to affect some crime types more than others. For effects beyond initial arrests, the analysis focuses on DV/IPV (Katz et al., 2014; Morrow et al., 2016). It has been suggested that BWCs may be particularly impactful in DV cases where victims may be reluctant to participate, and victim attrition is high. In IPV cases, the likelihood of making an arrest post-intervention fell for both BWC and non-BWC officers, but BWC officers were still significantly more likely to make an arrest (41% versus 34%) than officers without BWC.

BWC incidents were significantly more likely to lead to a charge (38% versus 26%), more likely to result in a guilty plea by the suspect (4.4% versus 3.1%), and more likely to result in a conviction at court (4.4% versus 0.9%). There may be some increases in demand downstream in terms of numbers going through the system, but the magnitude of these increases is small, and conviction rates for IPV remain low.

The Maryvale researchers also present BWC impacts on times to case disposition. While an increase in charges and convictions lead to increases in demand in terms of numbers going through the system, it has also been anticipated that BWCs may speed up the process. And increases in guilty pleas should also speed up time to disposal without the need to wait for time-consuming and costly trials. Overall, both BWC and non-BWC cases were processed significantly faster post-intervention, but this was largely due to the addition of a court liaison officer shortly after BWC implementation. And while all case processing times improved, these improvements were not as pronounced in the BWC cases. Officers reported frustrations with the difficulty of downloading and processing BWC material. Prosecutor staff also found BWC evidence difficult to process, causing some delays.

This suggests that there were increases in pressures on resources because of the BWC implementation, and this delay was not offset by the additional evidence gleaned from the BWC. The Maryvale pilot was replicated on a larger scale when BWC usage was expanded to the rest of the Phoenix Police Department (Huff et al., 2020). This is not included as a key study as they did not measure outcomes beyond the number of officer arrests.

Owens et al. (2014) applied an RCT on a BWC pilot in Essex (UK). Seventy police officers were randomly assigned to carry BWCs over a 4-month trial period, whereas the remaining 246 officers formed the control group. The focus of the analysis was on the impacts of BWC on DV incidents and their progression through the CJS. Although more robust than the Maryvale quasi-experimental study, this was a small-scale pilot over a relatively short duration. Furthermore, there were challenges in implementation, with officers failing to adhere to the test conditions. BWC officers were required to carry a camera, but activation of the camera was not mandatory, and officers did not record if the camera was switched on at the time of the DV incident. Only one in 6 test officers reported using the BWC in DV cases. The low usage of the cameras and small sample sizes, and the short duration of the study, meant that they could not track CJS outcomes beyond the initial police outcome.

There were no significant differences between BWC and non-BWC officers’ rates of sanction detection (that is, whether the suspect received a charge or a caution), but BWC incidents were statistically significantly more likely to be charged than cautioned (81% versus 72%). Implementing BWCs therefore made no difference in terms of the number of DV detections and arrests but were associated with a higher likelihood of charge. Other things being equal, this would have further downstream effects in terms of additional numbers of suspects going through the CJS, although what these effects may be and the magnitude of them was not tested further in this study.

A follow-up study involving the Metropolitan Police (London) supported by the College of Policing was conducted by Grossmith et al. (2015). The BWCs were implemented in 10 London boroughs over 12 months, with a focus on BWC usage at violent incidents. This was a cluster RCT as officers were randomly assigned to either test (649 officers) or control conditions (1,246 officers), with analysis able to account not only for differences between these officers but also variation across the different boroughs that formed distinct ‘clusters’. The focus was on whether BWC implementation affected police outcomes at violent incidents, including arrests and subsequent charges. The authors speculate that BWC evidence could lead to the progression of fewer, but stronger cases, although they did not test case progression beyond the charge stage.

There were no overall significant differences (across clusters) for the volumes of stop and searches, arrests, or charges. The only indication of a change in demand on the police was a slight reduction in the number of investigations of certain police complaints, but these are rare events and, unless these complaints are very serious, the direct impact is likely to be minimal. This study found that when controlling for areas/clusters, there were no significant main effects of BWC implementation on police arrests or charges.

Yokum et al. (2017) report the findings from a large-scale US pilot in Washington, DC over 11 months. This block RCT involved the random assignment of officers (1,035 officers) to the test conditions and control conditions (1,189 officers). They tested not only whether officers were in the experimental or control conditions but the geographical area (or block) they were operating in (to test for variation across areas of the city). Importantly, BWC officers were mandated to activate their BWCs at all incidents and police-initiated events they attended. While they still recognise the possibility that some officers may not comply with activation instructions, this was still an improvement over those studies that allowed officers discretion in BWC activation. Compliance was also measured and accounted for in the multivariate analysis. They measured police use of force, officer complaints, police activities and CJS outcomes, with case progression tracked until 10 months after the pilot ended. They did not, however, analyse the data by crime type.

Overall, Yokum et al. found very few treatment effects. No significant differences were found between BWC and non-BWC officers for arrests for disorderly conduct, prosecutions, not guilty or guilty verdicts or cases not pursued/ no further action (measures of use of force and citizen complaints were also not found to be significant). Time series analysis was used to check any change in impact over time – for example, whether initial effects of BWC implementation became more diluted over time – but this analysis failed to find any significant differences. Similar to the UK cluster RCT by Grossmith et al. (2015), accounting for between area variations identified no significant changes in terms of cases yielding an arrest or charge. Furthermore, they were able to test case progression through the CJS to the final case outcome and found no significant changes there either. They could also rule out any officer non-compliance and changes in effects over time.

The authors suggest that the results may not necessarily generalise to other locations in the US as policing in Washington, DC is often atypically ‘high stakes’ (that is, they police a much higher number of large-scale public events such as political events) compared to other cities in the US.

The next series of studies (Huff et al., 2021; Todak et al., 2022; White et al., 2021; White et al., 2018b) were all derived from a 6-month RCT on the implementation of BWCs in Tempe, Arizona in the US. Officers were randomly assigned to either the test (101 officers) or control condition (99 officers). Unlike other studies in the area that focus on either total crime or subsets of violent or DV incidents, these studies measured the impact of BWCs on misdemeanour or low-level volume crimes. Any change in downstream demands here is likely to have larger effects on the CJS due to the higher volume of these types of crimes. Consistent with most other work in the area, the effects of BWCs are based on the presence rather than activation of the BWC, but usage compliance rates were very high (95%).

The initial paper (White et al., 2018b) presents findings of all misdemeanour crimes combined, whereas subsequent papers focused on specific subtypes of misdemeanour crimes (Huff et al., 2021; White et al., 2021), or looked for changes in BWC effects over time (Todak et al., 2022).

The studies hypothesise that BWC deployment should enhance evidence quality, and therefore lead to an increase in guilty pleas and faster processing. Faster dispositions are assumed to result from more rapid decision-making – that is, it is easier to both exonerate innocent suspects and to make a stronger case against guilty ones, leading to an early acceptance of a guilty plea, with access to BWC material. Importantly, they did not necessarily disaggregate guilty outcomes into those related to a guilty plea versus guilty verdict at trial, each of which will have different impacts on downstream demand. White et al. (2018b) found an overall increase in guilty outcomes (351 per 1,000 cases compared to 332 per 1,000), and a decrease in case disposition time (on average 45.4 to 41.8 days), but neither of these changes were statistically significant.

Huff et al. (2021) examine the impact of both suspect race and BWC implementation on both dismissals and guilty plea convictions in traffic violations, comparing data from one-year pre-intervention for up to 19 months post-intervention. It was hoped that the use of BWCs might help reduce race-dependent differential case outcomes, arguing that BWC evidence should make decisions more impartial. They also expected to see increases in convictions, both in terms of guilty pleas and trial verdicts.

They found racial disparities existed regardless of BWC deployment. For example, White suspects were significantly more likely to have their case dismissed compared to both Black and Hispanic suspects. Both Black and White suspects were significantly less likely to plead guilty to a traffic violation when BWC evidence was involved (33.3% versus 36.7%; p < 0.01; 35.6% versus 36.7%; p < 0.05 respectively), with no significant effects of BWCs on cases involving Hispanic suspects. There were no significant effects of BWCs on case dismissals.

Contrary to expectations, BWCs did not appear to reduce demands by increasing either case dismissals or guilty pleas. Instead, BWC implementation is likely to have increased demand by lowering the likelihood of accepting early guilty pleas, leading to more costly and time-consuming court proceedings.

White et al. (2021) examined the impact of the BWC implementation on over 7,000 drug and alcohol cases, expecting they would increase the proportion of convictions but also shorten the time to disposition. The findings showed the presence of BWCs did not significantly impact on the likelihood of convictions, but they did lead to shorter disposition times. The average time to disposition was 6% lower in BWC cases, meaning on average a BWC case is resolved 5.5 days sooner than a non-BWC case. This demonstrates that BWC may lead to faster case processing, but the mechanism behind this is unclear. This finding is not associated with a higher rate of convictions, while breakdowns on the number of guilty pleas versus guilty verdicts, and case dismissals, are not presented.

Todak et al. (2022) analyse the use of BWC in over 50,000 misdemeanour cases, including offences against the person, property offences, traffic violations and drug/alcohol offences. To supplement their regression analysis, they also used Difference-in-Difference (DID) and Auto-Regressive Integrated Moving Average (ARIMA) time series analysis to test whether there was any change in case flow attributable to the BWC deployment. While this study applies a quasi-experimental design, it met our quality criteria and is useful as it can demonstrate the impact of BWC deployment on a large sample of cases from across a range of crime types. Furthermore, the use of time series analysis allows models to be tested against baseline data comparing the observed outcomes against predicted outcomes had no BWCs been deployed.

The study found no significant effects, suggesting that there were no changes in the number of cases filed post BWC intervention. In terms of aggregated crime, BWC cases were 10.3% less likely to lead to a conviction, but they were processed 4% faster on average than non-BWC cases. For both of these main findings, the results were statistically significant, but the findings were not consistent across all crime types. Disaggregation by crime type found that BWCs lowered the likelihood of conviction in both offences against the person (14% less likely) and traffic violations (12% less likely). It also revealed that both traffic violations and drug/alcohol offences were processed significantly faster in BWC cases (both of which witnessed a 6% reduction in time to disposition). Both findings suggest that BWC can lower downstream demands.

Clare et al. (2021) examined a 6-month RCT trial with the Western Australian Police Force. This study differed in that they randomly allocated each police shift to either the treatment (86 days) or control conditions (101 days), rather than individual officers. Therefore, there were days when officers were given a BWC to use and days when they were not. Officers were mandated to record all DV and violent incidents but were given greater discretion in terms of other incidents and events. They sought to address the issues recognised in previous studies of officers not complying with the treatment condition, so they monitored how well officers engaged with or ‘bought into’ the intervention, and categorised officers accordingly into high and low engagement groups. They also compared 2 different areas to test for variation within different geographical contexts – Perth (the state capital; metropolitan) and Bunbury (a regional town).

In terms of operational decisions and CJS outcomes, the evidence was mixed. On BWC days, officers issued a significantly greater number of ‘move-on’ notices for anti-social behaviour (ASB) (42 per 1,000 incidents compared to 38), and significantly more criminal charges were made (139 per 1,000 incidents compared to 126). High-use BWC engagement officers in particular had significantly higher sanction rates (for both ‘move-on’ orders and restraining orders) and charge rates on test days, compared to both their own control days, and the test days of low-use engagement officers. Cases involving BWC footage were no more likely to lead to a guilty plea, but when a guilty plea was accepted it occurred earlier in the process (47% of guilty pleas were accepted by the suspects by their first appearance in court, compared to 35% in the control cases).

There were no significant differences in terms of overall conviction rates between conditions; although they only tracked cases for 4 months after the intervention ended, so many of the cases had not been finalised. This is a common problem in BWC studies where the follow-up duration is not long enough to track cases to their final dispositions. Unlike previous studies, Clare et al. (2021) do not present comparisons of the 2 geographical areas.

The implications in terms of the impact on demand on the downstream are not straightforward. The higher rates of certain types of sanctions and charges will inevitably lead to an increase in downstream demand owing to more suspects going through the system, but there is considerable variation dependent upon the level of intervention engagement of individual officers. For example, most studies stipulate that officers are to be equipped with a BWC, but few measure whether the officer activates the BWC. This is a factor that has not been widely considered within the literature. The impact on CJS outcomes is limited given the short follow-up period, but there are indications that the increase in charges does not appear to lead to a greater number of convictions.

There are some reductions in demand in terms of the investigative time officers save through using BWC over more time-consuming handwritten reports on interviews. There may also be some time and cost savings in the courts if guilty pleas (although no more likely in the BWC cases) happen earlier in the process.

Petersen et al. (2021) present the findings of a 6-month cluster RCT in Miami Beach (Florida, US) which also involved the random assignment of shifts (rather than officers) to test and control conditions. They did so to minimise treatment contaminations, in particular to avoid the attendance of both test and control officers at the same incident. They also account for variations between areas by using cluster randomisation. They hypothesise that BWC implementation should have a differential effect dependent upon crime type, focusing here on DV, volume drug/alcohol offences, and assaults against the police, these being the offences most likely to be affected but in different ways. The presence of a BWC is measured with no indication of whether the BWC was activated in an incident or resulted in the capture of meaningful evidence. CJS outcomes are tracked from charged offences only, so this study does not indicate whether there are impacts on case initiation (or case flow). They measure outcomes in terms of the number of convictions (not disaggregating this into guilty plea versus trial convictions).

Once the analysis allowed for the effects of clustering (that is, the variation between areas) only BWC effects for assaults against the police were significant in terms of additional convictions: the likelihood of conviction/adjudication withheld was 93% more likely in BWC cases compared to non-BWC ones. These incidents are obviously quite rare, so while they create some downstream demand increases, these are not likely to have substantial impacts on the downstream. Once controlling for area, the use of BWCs in DV incidents had no significant effect on convictions.

The final key study by Pimley et al. (2022) differs from the others in that it is not an evaluation of a specific BWC intervention, but a longitudinal study examining changes in CJS outcomes over time - before, during, and after implementing BWCs in the area. They use data on 13,000 incidents from a small (un-named) police department in the Pacific Northwest of the US, from January 2004 to December 2018. BWCs were implemented in the force in April 2013. Interrupted Time Series Analysis (ITSA) is used to model the impact of the intervention by comparing the witnessed changes against the predicted changes had no intervention occurred (using the 9 years of pre-intervention/baseline data). It additionally accounts and controls for any seasonal variations and can explore not only pre- and post-intervention changes, but whether there are any delayed implementation effects.

There are some limitations, however, consistent with many studies in this field, and the authors are unable to disaggregate convictions into guilty pleas and court judgements. Furthermore, while they provide some comparison of crime types (for example, traffic violations, shoplifting, and simple assault) there was an insufficient variation of crime types in the sample to disaggregate in the main analysis for anything other than traffic violations (which was the most common type of offence).

Driving under the influence (DUI) and driving with a suspended licence (DWSL) saw significant decreases in time to disposition after BWC was implemented, indicating that some volume crimes cases were resolved faster. DWSL cases saw a significant decline in convictions. It is possible that some cases were dismissed quicker due to faster exoneration of suspects rather than increased incriminating evidence (and therefore more convictions).

There is evidence for some overall impacts on downstream demand, although as expected, often the effects were delayed until after the initial implementation period as “while policy may have an implementation date, the integration of the policy, or in our case BWCs, may take time to demonstrate an effect” (Pimley et al., 2022: 5). This is an advantage to longitudinal time series studies over previous pilot studies, which tend to have short follow-up durations (although the absence of controls is a weakness). In contrast, longitudinal studies consider not just short-term immediate implementation effects but are able to demonstrate long-term demand changes once the BWCs are fully embedded.

For all crimes, there was a delayed increase in prosecution referrals with an average of 0.35 more prosecutions per month post-intervention. Conviction rates had been in decline both before and during the implementation of BWCs, but post-intervention they significantly rose to an average of 0.37 more convictions per month. These both suggest some increased downstream demand in terms of numbers being dealt with following conviction, albeit after an initial delay while the intervention was being embedded. But these changes are also fairly modest in size.

The authors measured changes in disposition times in some detail and looked at a range of different types of dispositions. Before implementation, disposition times had been lengthening, and it was therefore anticipated that BWC would help speed up the processing of cases by providing more definitive evidence to make faster decisions either way. There is some support for this, as disposition times did eventually decrease by an average saving of 0.75 days per case per month, even though no significant changes were seen at the initial onset of BWC use. Comparing the patterns of change for different kinds of dispositions, convictions were initially subject to some considerable delays (on average increases of 31 days per case), but these delays were only short term, suggesting the CJS had to adapt to the changes.

Case dismissals, in contrast, were processed significantly faster after implementation. Their disposition times had been increasing prior to intervention and then fell dramatically (by on average 21 days per case). But this stabilised to a more modest level, on average a saving of 0.62 days per case. Overall, time to disposition did improve, and cases were processed faster. This longitudinal study was therefore important in demonstrating that BWC implementation, while it may take time to embed, may increase demands downstream in terms of additional convictions and more offenders being dealt with post-conviction but can also speed up the justice process. Cases were seen to progress faster, with weaker cases being dismissed earlier, and stronger cases moving more quickly through the CJS.

Petersen and Lu (2023) undertook a meta-analysis of a mix of 12 RCT and quasi-experimental studies which looked at the impact of BWCs on downstream outcomes. Because this review is very relevant to this literature review (that is, the impact of BWCs on the downstream), we have included it in this key study section. The authors found that, taken individually, 6 out of 8 studies reported an increase in the charge rates, but these effects were not significant when aggregated across all offence types in the meta-analysis. However, when the meta-analysis focused on DV charge rates, there was a significant and large effect for BWC use. Further, on guilty pleas at the aggregated level, the meta-analysis failed to find statistically significant differences between cases with and without BWC, but they did find a large statistically significant increase in guilty pleas for DV cases only.

Turning to overall convictions, Petersen and Lu (2023) found the same pattern of results – combining all crime types together failed to find any statistically significant increases but, taking the DV data in isolation, there were 3 to 4 times as many convictions in BWC cases. The authors caution that these studies tended to be quasi-experimental, which may have led to over-inflated effect sizes. BWCs appear to have a more pronounced impact on certain crimes, but little work has been done so far that explores non-DV crime types in sufficient detail. Petersen and Lu (2023) were therefore unable in their meta-analysis to explore anything other than all non-DV offences versus DV offences.

In summary, the relationship between BWC and downstream demand is complex. The evidence is also mixed with some studies indicating BWCs can increase the numbers of suspects going through the CJS through an increase in charges and convictions, while others have found demands to decrease as more offenders plead guilty. Other studies have found the BWC evidence can both speed up decision-making and slow it down because of the sheer weight of evidence to process. Others have found no significant demand changes at all. Research tracking the impacts on case progression to disposition is also still relatively scarce, and is also influenced by variations in research design, and differences in BWC implementation across forces. The selection of studies with more robust methodologies suggests, at most, some very small increases in terms of sanctions, charges, and convictions.

For example, evidence suggests that they can help reduce attrition in DV cases, although the impact on demand on the downstream is still likely to be modest. Some reductions may be achievable by processing and disposing of cases quicker but there are indications that this very much depends on the establishment and maintenance of a supporting infrastructure as the downloading and processing of BWC evidence is also associated with delays, at least in the short term. BWCs have a more pronounced impact on certain crimes, but little work has been done so far that explores non-DV crime types in sufficient detail. Evidence does suggest that there will be some increased downstream demands associated with BWCs in DV incidents, but the reliance on weaker methodologies in these studies warrants some caution.

The strongest evidence is from DV cases, but this tends to rely on quasi-experimental methods, sometimes with inadequate control comparisons (which may overestimate effects), and small sample sizes. Even the more robust DV focused RCT (Owens et al., 2014) had issues with officer compliance and only found a significant shift from out of court resolutions to charges. Few studies account for levels of officer compliance, and few measure anything other than whether a BWC was present at an incident, not whether it was activated, and whether any meaningful BWC footage was used as evidence (Petersen et al., 2023). There are huge differences between studies in terms of the implementation strategies and, in particular, mandates for officer activation of the BWC (that is, the level of discretion officers have).

Most studies fail to disaggregate specific crime types, despite the likelihood that the effects will be different. They also largely fail to account for potential variations between geographic areas. BWCs are theorised to work in several different ways, but often the mechanism of change is unclear due to the aggregation of guilty pleas and guilty verdicts at trial.

Even those studies with the more rigorous research designs, as selected here, suffer from several weaknesses, and there are some important gaps in the literature. Most studies allocate individual officers to either control or test conditions, but do not adequately control for officer compliance (confounding effects) or non-camera officers attending the same scene in experimental locations (contamination), even in many of the RCT studies.

Table 3.1 A summary of the findings on demand from the key and secondary studies

The impact of wearing a BWC….	Increase/decrease or neutral demand on CJS	Overall assessment of the impact	Evidence base
On arrests (All non-DV offence types)	Neutral	Difficult to assess impact due to mixed findings. Overall impact neutral.	Inconsistent findings. It depends on many factors, for example, activation of BWC, quality of BWV evidence, type of offence. UK (Grossmith et al., 2015) and US (Yokum et al., 2017) robust studies found no differences in arrests between BWC and non-BWC. One quasi-experimental study found BWC to increase arrests (Katz et al., 2014). Less robust studies found inconsistent findings.
On arrests (IPV and DV)	Increase	A likely increase in demand on the CJS.	One robust US study found an increase (Morrow et al., 2016). The small UK study found no differences (Owens et al. 2014). Other weaker studies found increases.
On charges (All non-DV offence types)	Increase	Mixed evidence, overall assessment is that there may be a small positive impact.	One robust Australian study found increases in charges (Clare et al., 2021). A US robust study found differences in charges could be explained by geographical variations (Yokum et al., 2017). One UK robust study found BWC cases were more likely to charge suspects than non-BWC (Owens et al., 2014). The second UK study, controlling for area, did not find differences for BWC/non-BWC on charges (Grossmith et al., 2015).
On charges (DV offence only)	Increase	On balance, likely to be an increase in charges for DV offences.	One UK RCT (Owens et al., 2014) and a US robust study (Morrow et al., 2016) found increased charges for DV cases – but it is small scale. Issue that there is inconsistent usage of BWC by officers. Other QE studies, reviewed as part of a meta-analysis, suggest an increase in charges for DV – but caution – likely inflation bias (Petersen and Lu, 2023).
On guilty pleas (Non-DV offences)	Neutral	Less or varied impact on non-DV offences.	Robust US studies find no statistically significant impact on guilty pleas for all other offences (non-DV) from BWC wearing (Clare et al., 2021; White et al., 2021; Yokum et al., 2017). One robust study (US) found BWC evidence to reduce guilty pleas (for a traffic violations) (Huff et al., 2021).
On guilty pleas (DV only)	Increase	Difficult to assess owing to study limitations. On balance, slight increase.	Most studies do not differentiate between guilty pleas before court or following court so difficult to disentangle demand on downstream. However, 2 robust US studies found guilty pleas increase for DV with BWC (Morrow et al., 2016, Katz et al., 2014).
On prosecutions and convictions (Non-DV offences)	Neutral	Likely impact offence specific. Difficult to assess owing to mixed findings.	Mixed evidence. Three robust US studies found no differences in prosecutions/convictions for BWC (Yokum et al., 2017; Clare et al., 2021; White et al., 2021). Another found a decrease in convictions (Todak et al., 2022) while the second found increases for assaults against the police (Petersen et al., 2021). One longitudinal study (US) found increases over time, but not for volume crimes (2 US studies) offences against the person, traffic offences (US x2) (Pimley et al., 2022).
Convictions – for DV offences	Increase	Overall increase in demand, but magnitude likely to be small.	One robust US RCT found no differences when geographic area controlled for (Petersen et al., 2021). Another quasi-experimental study found a modest increase (Morrow et al., 2016). Otherwise generally weak study methodologies. One meta-analysis review found 3 to 4 times as many convictions in BWC cases compared with non-BWC cases (Petersen and Lu, 2023). Direction suggests an increase in convictions for DV.
On timeliness	Neutral	On balance, BWC can both speed up and slow down an investigation. Impact on CJS is likely to be neutral.	Helps speed up time to disposition (Todak et al., 2022, White et al., 2021). One study found it slowed down the time to disposition – for IPV cases (Morrow et al., 2016) and caused more time-consuming court cases (Huff et al., 2021). But some studies also found it depended on the crime type (Pimley et al., 2022 and Todak et al., 2022).

3.3 Findings and implications

Despite narrowing the parameters of our review on higher quality design BWC studies with a focus (only) on charges and other cases outcomes, interpreting the evidence base in this area is challenging.

RCT research has considerable advantages over quasi-experimental designs, especially those with unadjusted control samples. They tend to reveal more muted but realistic impacts of BWC on the downstream CJS. But even RCT studies largely fail to account for contamination bias, with measurement of treatment effects often at an officer level, and the likelihood that officers from both test and control conditions attend some of the same incidents.

More studies are beginning to address the impacts of BWC on CJS outcomes, but for much of this work, little data is collected beyond the initial upstream impacts (for example, arrests). Furthermore, when cases are tracked to outcome finalisation, the follow-up periods tend to be too short, and end before many cases are closed. Few studies include data on times to case disposition.

There is variability in terms of force implementation strategies, and most studies only measure the presence (not the activation). Whether cameras were activated by officers attending scenes, and the footage is usable, is absolutely pivotal in understanding questions around the possible impact of BWC on downstream flows.

Most studies use aggregated crime data. Although there are some studies that look at specific crime types, not enough is known about a wider range of crime-type specific effects. DV appears to be one of the most affected crime types in terms of the impact of BWC on CJS outcomes. But the evidence here is still not the most robust. All we can say reliably is that BWC impacts appear to vary by crime type, with DV incidents looking most promising in terms of increased charges. The value of ‘all crime’ studies, even if methodologically strong, in assessing downstream impacts, is limited.

Many studies do not disaggregate convictions into those resulting from a guilty plea and those resulting from a full trial verdict. This is problematic, as it obscures fundamental differences in the impact on the downstream CJS. Increases in convictions may increase pressures on prison and probation populations. But an increase in early guilty pleas, due to the stronger evidence associated with BWCs, might actually reduce these demands by avoiding time-consuming and costly full trials and the application of sentencing discounts.

The relationship between BWC deployment and the likely impact on demands across the CJS is complex and difficult to disentangle, involving the potential to both increase and decrease downstream demands.

On arrests, there are inconsistent findings for non-DV offences. For DV offences, again the findings are inconsistent, but one key study and some secondary studies suggest arrests increase for BWC cases .

On charges, the evidence suggests BWC non-DV cases may end up with more charges – but studies which control for other factors, for example, location, find the effect to disappear. There appears slightly more robust evidence that BWC evidence for DV offences results in an increase in charges.

The evidence is mixed on prosecutions and convictions for non-DV and any impact is likely to be offence specific. For DV offences, the findings suggest an increase in convictions.

Evidence on whether BWCs impact on guilty pleas differs by crime type. Two US pilots found a slight increase in guilty pleas for DV. Others found no differences. While another US study found a slight reduction in guilty pleas for volume drug and alcohol offences but 2 other US studies found no statistically significant impact. Guilty pleas can, however, both increase and reduce downstream demands. For example, more guilty pleas can increase demand by contributing to an overall higher number of convictions. But they can also reduce demands by speeding up the disposition and avoiding the full cost of a trial.

Findings on the length of time to outcome is also mixed. There is evidence that BWC evidence may take longer to process and contribute to delays, therefore increasing demands (US studies). But further evidence suggests that while BWCs can cause initial delays in convicted cases, they can eventually improve disposition times, and lead to faster case dismissals.

None of the studies with the most rigorous methodologies (that is, cluster and block RCTs, that consider variations between geographical areas) found any significant differences between BWC and non-BWC except for a possible reduction in police complaints, (which is of less interest to us in this review).

Question 5

4.  Overall conclusions

Accepted Answer

This review has considered the research evidence on downstream impacts for 2 different but linked investigative tools. The selection of DNA and BWC as the focal points for this exercise was determined by the fact that, from our initial assessment of the wider research into factors that influence demand flow through the CJS, small bodies of research on BWCs and the use of trace DNA typically captured how these investigative tools had impacted on charges, convictions, sentencing, and case processing times. Within these small pools of research, there are only a handful of strong studies. Even those considered robust rarely cover the full range of potential downstream impacts (pleas, convictions, and sentencing). Despite these limitations, this review has identified some important insights in understanding the complexity of upstream interventions on downstream pressures.

The single robust study on the impact of retrieving trace DNA at property crime scenes suggests that, for this crime type at least, DNA has an additive effect on arrests and prosecutions. In those cases where traces of DNA are found at the scene, suspect leads and additional prosecutions (although the impact further downstream is not as well evidenced) are generated. These are cases which, without DNA, might previously have been closed with no further action (NFA). However, these impacts will only occur in that subset of crimes where DNA trace material is both present, efforts are made to collect and analyse it and only then can it contribute to the identification of an unknown suspect (or confirm a suspect was present or eliminate those with legitimate reasons for being at the crime scene).

And while these ‘DNA detections’ only constitute a small proportion of total recorded crime, they are likely to be otherwise very hard to detect without DNA (or other forensic techniques, such as fingerprints) where known offenders can be linked to the crime scene. The fact that the DNA detected property offences (such as burglary) seem to be linked to more prolific offenders suggests that the potential benefits of DNA may be more considerable than implied based on a simple assessment of its relationship with total crime.

The impact of DNA trace material on the outcome of more serious offences such as stranger rape and stranger homicide is not well covered in the research. But we do know DNA does successfully identify suspects in these hard-to-solve offences (Burrows et al., 2005a). And we might speculate that the impact of DNA on these serious offences is a little different to the impact on property crimes. This is because more serious offences, such as stranger rape, are not necessarily closed NFA if initial evidence gathering is not effective at identifying a suspect (whereas this will often be the case in a burglary), so collecting and analysing DNA traces may well end up resolving ‘open’ serious cases more rapidly than traditional investigative methods, as well as detecting cases which might have been undetectable without DNA (hence the important role that DNA evidence plays in cold cases (see Parnell and Russell, 2024)).

The evidence around the downstream impacts of the introduction or expansion of DNA databases suggests that these interventions can have impacts which work simultaneously but in different directions. They can suppress demands on the CJS through targeted and general deterrence by reducing re-offending. But they can also increase demand in terms of additional CJS outcomes. The evidence on whether the net effect of database expansion on CJS demands is negative or positive is mixed and one might also raise questions about how long these various different effects continue post initial expansion. Nevertheless, for the policy makers, simply being alert to the potential for simultaneous effects when thinking about policy development seems an important principle to grasp.

Likewise, an important lesson from the current, albeit partial evidence base on BWCs and the relationship with CJS outcomes is that examining final court outcomes, without taking into consideration whether guilt has been determined by an early guilty plea or decision at court (and what that entails for sentencing), is a major limitation which masks important insights in to the actual impact on the CJS downstream.

The state of the evidence on DNA and BWC is quite different. Even though the field of DNA is complicated by covering both retrieval of the DNA trace effects and the effects from database expansion, the higher quality evidence base here is modest enough to be easily marshalled. That picture is not replicated in the still growing evidence base on BWC usage. Here we have a range of studies, many of which deal with other impacts of BWCs (police complaints, police behaviour, victim behaviour), reflecting the research community’s desire to explore a much broader canvas of effects from BWC wearing and usage. The core studies with a meaningful CJS outcome focus are relatively rare, and seemingly awash with conflicting findings. This lack of consistency partly reflects the importance of crime-type differences in determining BWC impact on the downstream of the CJS.

Tentative evidence suggests that BWCs may have the most impact – that is, more charges and prosecutions – on DA-flagged cases, while the impact on other offences appears to be more varied. But even this tentative summary needs caveating.

But we should be careful to not over interpret this very tentative finding as it is only relevant to DA-flagged cases. The principle that BWC impacts the downstream CJS in different ways depending on crime type is likely to extend to offences not covered by the current, partial evidence base. A good illustration of the potential for BWC evidence to identify suspects may emerge from its application in the disorder that took place in England and Northern Ireland in the summer of 2024 (House of Commons library, 2024). Robustly assessing the impact of BWCs on the CJS from spontaneous events like the summer disorder is also likely to be methodologically challenging.

Finally, we should acknowledge that the technological innovations and their impact on the downstream, where they happen, will do so alongside major tidal influences on CJS demand, such as officer experience and workload (for instance, see Cole et al., publication pending). Given the evidence presented in this review, DNA and BWCs have the potential to have important effects on downstream demand albeit within specific crime types. But their impact will be moderated or intensified by their interaction with a cocktail of other tidal factors active at the time.

References

Alderden, M., Cross, T. P., Vlajnic, M., and Siller, L. (2021). Prosecutors’ perspectives on biological evidence and injury evidence in sexual assault cases. Journal of Interpersonal Violence, 36(7 to 8), 3880 to 3902. Available online [Accessed 7 February 2025]

Amankwaa, A. O., and McCartney, C. (2021). The effectiveness of the current use of forensic DNA in criminal investigations in England and Wales. WIREs Forensic Sc., 3, e1414. Available online [Accessed 7 February 2025]

Anker, A. S. T., Doleac, J. L., and Landersø, R. (2021). The Effects of DNA Databases on the Deterrence and Detection of Offenders. American Economic Journal: Applied Economics, 13(4), 194 to 225. Available online [Accessed 7 February 2025]

Ariel, B. (2016a). Increasing Cooperation With the Police Using Body Worn Cameras. Police Quarterly, 19(3), 326 to 362. Available online [Accessed 7 February 2025]

Ariel, B. (2016b). Police Body Cameras in Large Police Department. Journal of Criminal Law and Criminology, 106(4), 729 to 768. Available online [Accessed 7 February 2025]

Ariel, B., Farrar, W. A., and Sutherland, A. (2014). The Effect of Police Body-Worn Cameras on Use of Force and Citizens’ Complaints Against the Police: A Randomized Controlled Trial. Journal of Quantitative Criminology, 31(3), 509 to 535. Available online [Accessed 7 February 2025]

Arnaud, C. H. (2017). Thirty years of DNA forensics: How DNA has revolutionized criminal investigations. Chemical and Engineering News, 95(37). Available online [Accessed 7 February 2025]

Baskin, D., and Sommers, I. (2010). The influence of forensic evidence on the case outcomes of homicide incidents. Journal of Criminal Justice, 38, 1141 to 1149. Available online [Accessed 7 February 2025]

Baskin, D., and Sommers, I. (2011). Solving Residential Burglaries in the United States: The Impact of Forensic Evidence on Case Outcomes. International Journal of Police Science & Management, 13(1), 70 to 86. Available online [Accessed 7 February 2025]

Baskin, D., and Sommers, I. (2012). Influence of Forensic Evidence on the Case Outcomes of Assault and Robbery Incidents. Criminal Justice Policy Review 23(2), 186 to 210. Available online [Accessed 7 February 2025]

Baskin, D. R., and Sommers, I. (2013). Commentary on: Peterson JL, Hickman MJ, Strom KJ, Johnson DJ. Effect of forensic evidence on criminal justice case processing. J Forensic Sci 2013 Jan;58 Suppl 1:S78 to 90. J Forensic Sci, 58(5), 1401-1402. Available online [Accessed 7 February 2025]

Bhati, A. (2011). Quantifying the Specific Deterrent Effects of DNA Databases. Washington, D.C.: Justice Policy Center, The Urban Institute Available online [Accessed 7 February 2025]

Bhati, A., and Roman, C. G. (2014). Evaluating and Quantifying the Specific Deterrent Effects of DNA Databases. Eval Rev, 38(1), 68 to 93. Available online [Accessed 7 February 2025]

Blaskovits, B., and Bennell, C. (2020). Exploring the Potential Impact of Body Worn Cameras on Memory in Officer-Involved Critical Incidents: a Literature Review. Journal of Police and Criminal Psychology 35(3), 251 to 262. Available online [Accessed 7 February 2025]

Boehme, H. M., and Schnell, C. (2022). Why did the use of force go up? Investigating the unexpected impact of a body-worn camera program on a use of force time series. Journal of Criminal Justice, 82. Available online [Accessed 7 February 2025]

Bradbury, S.-A., and Feist, A. (2005). The use of forensic science in volume crime investigations: a review of the research literature. (43/05). Home Office. Available online [Accessed 7 February 2025]

Braga, A. A., Barao, L. M., Zimmerman, G. M., Douglas, S., and Sheppard, K. (2019). Measuring the Direct and Spillover Effects of Body Worn Cameras on the Civility of Police–Citizen Encounters and Police Work Activities. Journal of Quantitative Criminology, 36(4), 851 to 876. Available online [Accessed 7 February 2025]

Braga, A. A., Sousa, W. H., Coldren, J. R., and Rodriguez, D. (2018). The Effects of Body-Worn Cameras on Police Activity and Police-Citizen Encounters: A Randomized Controlled Trial. Journal of Criminal Law and Criminology, 108(3), 511 to 538. Available online [Accessed 7 February 2025]

Briody, M. (2002). The Effects of DNA Evidence on Sexual Offence Cases in Court. Current Issues in Criminal Justice, 14(2), 159 to 181. Available online [Accessed 7 February 2025]

Briody, M. (2004). The effects of DNA evidence on homicide cases in court. Australian and New Zealand Journal of Criminology, 37(2), 231 to 252. Available online [Accessed 7 February 2025]

Briody, M. (2005). The Effects of DNA Evidence on the Criminal Justice Process. Griffith University. Available online [Accessed 7 February 2025]

Briody, M. (2006). The Effects of DNA Evidence on Property Offences in Court. Current Issues in Criminal Justice, 17(3), 380-396. Available online [Accessed 7 February 2025]

Burrows, J., Hopkins, M., Hubbard, R., Robinson, A., Speed, M., and Tilley, N. (2005a). Understanding the attrition process in volume crime investigations: London: Home Office. Available online [Accessed 7 February 2025]

Burrows, J., Tarling, R., Mackie, A., Poole, H., and Hodgson, B. (2005b). Forensic Science Pathfinder project: Evaluating increased forensic activity in 2 English police forces. Home Office Online Report 46/05. London: Home Office

Campbell, R., Bybee, D., Ford, J. K., and Patterson, D. (2008). Systems Change Analysis of SANE Programs: Identifying the Mediating Mechanisms of Criminal Justice System Impact: Project Summary. Available online [Accessed 7 February 2025]

Campbell, R., Bybee, D., Townsend, S. M., Shaw, J., Karim, N., and Markowitz, J. (2014a). The Impact of Sexual Assault Nurse Examiner Programs on Criminal Justice Case Outcomes: A Multisite Replication Study. Violence against women, 20(5), 607 to 625. Available online [Accessed 7 February 2025]

Campbell, R., Fehler-Cabral, G., Bybee, D., and Shaw, J. (2017). Forgotten Evidence: A Mixed Methods Study of Why Sexual Assault Kits (SAKs) Are Not Submitted for DNA Forensic Testing. Law and Human Behavior, 41(5), 454 to 67. Available online [Accessed 7 February 2025]

Campbell, R., Greeson, M., Bybee, D. I., Kennedy, A., and Patterson, D. (2011). Adolescent Sexual Assault Victims’ Experiences with SANE-SARTs and the Criminal Justice System. Available online [Accessed 7 February 2025]

Campbell, R., Patterson, D., and Bybee, D. (2012). Prosecution of Adult Sexual Assault Cases: A Longitudinal Analysis of the Impact of a Sexual Assault Nurse Examiner Program. Violence Against Women 18(2), 223 to 244 Available online [Accessed 7 February 2025]

Campbell, R., Patterson, D., Bybee, D., and Dworkin, E. R. (2009). Predicting Sexual Assault Prosecution Outcomes. Criminal Justice and Behavior, 36(7), 712 to 727. Available online [Accessed 7 February 2025]

Cayli, B., Hargreaves, C., and Hodgson, P. (2018). Body-worn cameras: determining the democratic habitus of policing. Safer Communities, 17(4), 213 to 223. Available online [Accessed 7 February 2025]

Challinor, J. M., DeTata, D. A., Pitts, K. M., and Burnier, C. (2021). Examination of Forensic Evidence. In K. D. a. W. Sam, T.P. (Ed.), Analytical Pyrolysis Handbook (3rd ed.). London: Routledge Online

Civil Service (no date) Rapid Evidence Assessment Toolkit. Available online [Accessed 7 February 2025]

Clare, J., Henstock, D., McComb, C., Newland, R., and Barnes, G. C. (2021). The results of a randomized controlled trial of police body-worn video in Australia. Journal of Experimental Criminology, 17(1), 43 to 54. Available online [Accessed 7 February 2025]

Cole, C., Richardson, V., Flannigan, L., Diver, M., Almond, S., and Feist, A., Identifying factors associated with changes in charge volumes: a statistical analysis.

Cross, T. P., Siller, L., Vlajnic, M., and Alderden, M. (2022). The Relationship of DNA Evidence to Prosecution Outcomes in Sexual Assault Cases. Violence Against Women. Available online [Accessed 7 February 2025]

Cubitt, T. I. C., Lesic, R., Myers, G. L., and Corry, R. (2017). Body-worn video: A systematic review of literature. Australian & New Zealand Journal of Criminology, 50(3), 379 to 396. Available online [Accessed 7 February 2025]

Doleac, J. L. (2017). The Effects of DNA Databases on Crime. American Economic Journal: Applied Economics, 9(1), 165 to 201. Available online [Accessed 7 February 2025]

Dunsmuir, W. T., Tran, C., and Weatherburn, D. (2008). Assessing the Impact of Mandatory DNA Testing of Prison Inmates in NSW on Clearance, Charge and Conviction Rates for Selected Crime Categories. New South Wales: Attorney General’s Department of New South Wales. Available online [Accessed 7 February 2025]

Ellis, T., Jenkins, C., and Smith, P. (2015). Evaluation of the introduction of personal issue body worn video cameras (Operation Hyperion) on the Isle of Wight: Final report to Hampshire constabulary. Portsmouth: Institute of Criminal Justice Studies, University of Portsmouth. Available online [Accessed 7 February 2025]

Goodall, M. (2007). Guidance for the police use of body-worn video devices. London: Home Office Available online [Accessed 7 February 2025]

Grossmith, L., Owens, C., Finn, W., Mann, D., Davies, T., and Baika, L. (2015). Police, Camera, Evidence: London’s cluster randomised controlled trial of Body Worn Video. Available online [Accessed 7 February 2025]

Hartman, J. L., and Belknap, J. (2003). Beyond the gatekeepers: Court Professionals’ Self-Reported Attitudes About and Experiences With Misdemeanor Domestic Violence Cases. Criminal Justice and Behavior, 30(3), 349 to 373. Available online [Accessed 7 February 2025]

Headley, A. M., Guerette, R. T., and Shariati, A. (2017). A field experiment of the impact of body-worn cameras (BWCs) on police officer behavior and perceptions. Journal of Criminal Justice, 53, 102 to 109. Available online [Accessed 7 February 2025]

Hedberg, E. C., Katz, C. M., and Choate, D. E. (2016). Body-Worn Cameras and Citizen Interactions with Police Officers: Estimating Plausible Effects Given Varying Compliance Levels. Justice Quarterly, 34(4), 627 to 651. Available online [Accessed 7 February 2025]

Henry, T. K. S., and Jurek, A. L. (2020). Identification, Corroboration, and Charging: Examining the Use of DNA Evidence by Prosecutors in Sexual Assault Cases. Feminist Criminology, 15(5), 634 to 658. Available online [Accessed 7 February 2025]

HMICFRS. (2019). The police response to domestic abuse: An update report. London: HMICFRS. Available online [Accessed 7 February 2025]

Home Affairs Committee. (2018). Policing for the future: Tenth Report of Session 2017 (19). London: House of Commons. Available online [Accessed 7 February 2025]

Hoffmann, R., Meakin, G., Morelato, M., and Roux, C. (2024). The utility of trace DNA within forensic science for investsigative and intelligence purposes. Wires Forensic Science. Available online [Accessed 7 February 2025]

House of Commons library (2024) Policing response to the summer 2024 riots. Available online [Accessed 7 February 2025]

Huff, J., Katz, C. M., and Hedberg, E. C. (2020). A randomized controlled trial of the impact of body-worn camera activation on the outcomes of individual incidents. Journal of Experimental Criminology, 18(2), 247 to 272. Available online [Accessed 7 February 2025]

Huff, J., White, M. D., and Padilla, K. E. (2021). The Influence of Defendant Race/Ethnicity and Police Body-Worn Cameras on Traffic Case Processing. Race and Justice. Available online [Accessed 7 February 2025]

Jameel, J., and Bunn, S. (2015). Body-Worn Video in UK Policing. Research Briefing London: UK Parliament Available online [Accessed 7 February 2025]

Jennings, W. G., Lynch, M. D., and Fridell, L. A. (2015). Evaluating the impact of police officer body-worn cameras (BWCs) on response-to-resistance and serious external complaints: Evidence from the Orlando police department (OPD) experience utilizing a randomized controlled experiment. Journal of Criminal Justice, 43(6), 480 to 486. Available online [Accessed 7 February 2025]

Johnson, D., Peterson, J., Sommers, I., and Baskin, D. (2012). Use of forensic science in investigating crimes of sexual violence: contrasting its theoretical potential with empirical realities. Violence Against Women, 18(2), 193 to 222 Accessed online [Accessed 7 February 2025]

Katz, C. M., Choate, D. E., Ready, J. R., Nuňo, L. S., Kurtenbach, M., and Johnson, K. (2014). Evaluating the Impact of Officer Worn Body Cameras in the Phoenix Police Department. Retrieved from Phoenix, AZ Available online [Accessed 7 February 2025]

Kelly, L., Lovett, J., and Regan, L. (2005). A gap or a chasm? Attrition in reported rape cases. Available online [Accessed 7 February 2025]

Kettrey, H. H., Davis, A. J., Liberman, J., and Seeman, A. (2022). Taking the Test: Participation in a Sexual Assault Medical Forensic Exam as an Indicator of Victim Cooperation with the Criminal Justice System and a Predictor of Suspect Arrest. Justice Quarterly, 40 (3) Available online [Accessed 7 February 2025]

Kjærulff, M. L. B. G., Bonde, U., and Astrup, B. S. (2019). The significance of the forensic clinical examination on the judicial assessment of rape complaints – developments and trends. Forensic Science International, 297, 90 to 99. Available online [Accessed 7 February 2025]

Lum, C., Koper, C. S., Wilson, D. B., Stoltz, M., Goodier, M., Eggins, E., Higginson, A., and Mazerolle, L. (2020). Body‐worn cameras’ effects on police officers and citizen behavior: A systematic review. Campbell Systematic Reviews, 16(3). Available online [Accessed 7 February 2025]

Lum, C., Stoltz, M., Koper, C. S., and Scherer, J. A. (2019). Research on body‐worn cameras. Criminology and Public Policy, 18(1), 93 to 118. Available online [Accessed 7 February 2025]

Mapes, A. A., Kloosterman, A. D., and de Poot, C. J. (2015). DNA in the Criminal Justice System: The DNA Success Story in Perspective. J Forensic Sci, 60(4), 851 to 856. Available online [Accessed 7 February 2025]

Margot, P. (2017). Traceology, the bedrock of forensic science and its associated semantics, in: Q. Rossy, D. Décary-Hétu, O. Delémont, M. Mulone (Eds.), The Routledge International Handbook of Forensic Intelligence and Criminology, Routledge, Abingdon (Oxon, UK), pp.30 to 39. Available online [Accessed 7 February 2025]

Maskaly, J., Donner, C., Jennings, W. G., Ariel, B., and Sutherland, A. (2017). The effects of body-worn cameras (BWCs) on police and citizen outcomes A state-of-the-art review. Policing: An International Journal of Police Strategies and Management, 40(4), 672 to 688. Available online [Accessed 7 February 2025]

McClure, D., La Vigne, N., Lynch, M., Golian, L., Lawrence, D., and Malm, A. (2017). How Body Cameras Affect Community Members’ Perceptions of Police: Results from a Randomized Controlled Trial of One Agency’s Pilot. Urban Institute. Available online [Accessed 7 February 2025]

McDonald, J., and Lehman, D. C. (2012). Forensic DNA Analysis. Clinical Laboratory Science, 25(2), 109 to 113. Available online [Accessed 7 February 2025]

McEwen, T., and Regoeczi, W. (2015). Forensic Evidence in Homicide Investigations and Prosecutions. Journal of Forensic Sciences, 60(5), 1188 to 1198. Available online [Accessed 7 February 2025]

Menaker, T. A., Campbell, B. A., and Wells, W. (2017). The Use of Forensic Evidence in Sexual Assault Investigations: Perceptions of Sex Crimes Investigators. Violence Against Women, 23(4), 399 to 425 Available online [Accessed 7 February 2025]

Morrow, W. J., Katz, C. M., and Choate, D. E. (2016). Assessing the Impact of Police Body-Worn Cameras on Arresting, Prosecuting, and Convicting Suspects of Intimate Partner Violence. Police Quarterly, 19(3), 303 to 325. Available online [Accessed 7 February 2025]

Nir, E., and Griffiths, E. (2018). Sentencing on the Evidence. Criminal Justice Policy Review, 29(4), 365 to 390 Available online [Accessed 7th February 2025]

NPCC. (2022). Body-Worn Video. Available online [Accessed 7th February 2025]

Owens, C., and Finn, W. (2018). Body-Worn Video through the Lens of a Cluster Randomized Controlled Trial in London: Implications for Future Research. Policing: A Journal of Policy and Practice, 12(1), 77 to 82. Available online [Accessed 7th February 2025]

Owens, C., Mann, D., and McKenna, R. (2014). The Essex Body Worn Video Trial The impact of Body Worn Video on criminal justice outcomes of domestic abuse incidents. College of Policing

Parnell, G., and Russell, R. (2024) The cold case unit solving crimes from the 1960s. BBC news. Retrieved from The South Yorkshire cold case team investigating crimes from the 1960s - BBC News [Accessed 7th February 2025]

Pelfrey, W. V., and Keener, S. (2016). Police body worn cameras: a mixed method approach assessing perceptions of efficacy. Policing: An International Journal of Police Strategies & Management, 39(3), 491 to 506. Available online [Accessed 7th February 2025]

Petersen, K., and Lu, Y.-F. (2023). The Downstream Effects of Body-worn Cameras: A Systematic Review and Meta-analysis. Justice Quarterly, 1 to 26. Available online [Accessed 7th February 2025]

Petersen, K., Mouro, A., Papy, D., Castillo, N., and Ariel, B. (2021). Seeing is believing: the impact of body-worn cameras on court outcomes, a cluster-randomized controlled trial in Miami Beach. Journal of Experimental Criminology. Available online [Accessed 7th February 2025]

Petersen, K., Papy, D., Mouro, A., and Ariel, B. (2023). The usage and utility of body‐worn camera footage in courts: A survey analysis of state prosecutors. Journal of Empirical Legal Studies, 20(3), 534 to 569. Available online [Accessed 7th February 2025]

Peterson, J., Sommers, I., Baskin, D., and Johnson, D. (2010). The Role and Impact of Forensic Evidence in the Criminal Justice Process. Available online [Accessed 7th February 2025]

Peterson, J. L., Hickman, M. J., Strom, K. J., and Johnson, D. J. (2013). Effect of Forensic Evidence on Criminal Justice Case Processing. Journal of Forensic Sciences, 58, 78-90. Available online [Accessed 7th February 2025]

Peterson, J. L., Mihajlovic, S., and Gilliland, M. (1984). Forensic Evidence and the Police: The Effects of Scientific Evidence on Criminal Investigations. Washington, D.C.: National Institute of Justice. Available online [Accessed 7th February 2025]

Peterson, J. L., Ryan, J. P., Houlden, P.J., and Mihajlovic, S. (1986). The uses and effects of scientific evidence in criminal case processing. . Washington, D.C.: National Institute of Justice Available online [Accessed 7th February 2025]

Pimley, N., Parks, M., and Makin, D. A. (2022). Time Saver or Time Sapper? An Examination of Body-Worn Camera Impact on Case Disposition Timelines. Criminal Justice Review. Available online [Accessed 7th February 2025]

Ready, J. T., and Young, J. T. N. (2015). The impact of on-officer video cameras on police–citizen contacts: findings from a controlled experiment in Mesa, AZ. Journal of Experimental Criminology, 11(3), 445 to 458. Available online [Accessed 7th February 2025]

Redlich, A. D., Bushway, S. D., and Norris, R. J. (2016). Plea decision-making by attorneys and judges. Journal of Experimental Criminology, 12(4), 537 to 561. Available online [Accessed 7th February 2025]

Rog, D., Chibnall, S., and Wells, W. (2021). The Evaluation of the Bureau of Justice Assistance Sexual Assault Kit Initiative: Final Report. Available online [Accessed 7th February 2025]

Roman, J. K., Reid, S., Reid, J., Chalfin, A., Adams, W., and Knight, C. (2008). The DNA field experiment: Cost-effectiveness analysis of the case of DNA in the investigation of high-volume crimes. Washington: Urban Institute Justice Policy Center. Available online [Accessed 7th February 2025]

Rowe, M., Pearson, G., and Turner, E. (2018). Body-Worn Cameras and the Law of Unintended Consequences: Some Questions Arising from Emergent Practices. Policing: A Journal of Policy and Practice, 12(1), 83 to 90. Available online [Accessed 7th February 2025]

Schroeder, D. (2007). DNA and Homicide Clearance: What’s Really Going On. Journal of the Institute of Justice & International Studies, 7, 279 to 298. Available online [Accessed 7th February 2025]

Schroeder, D. A., and Elink-Schuurman, L., Kristin (2017). The Impact of Forensic Evidence on Arrest and Prosecution. Available online [Accessed 7th February 2025]

Schroeder, D. A., and White, M. D. (2009). Exploring the Use of DNA Evidence in Homicide Investigations: Implications for Detective Work and Case Clearance. Police Quarterly, 12(3), 319 to 342. Available online [Accessed 7th February 2025]

Sherman, L. W., Gottfredson, D. C., MacKenzie, D. L., Eck, J., Reuter, P., and Bushway, S. D. (1998). Preventing Crime: What Works, What Doesn’t, What’s Promising. National Institute of Justice: Research in Brief

Spencer, D., and Cheshire, R. (2018). Ten Years of Body Worn Video in Northamptonshire Police. Policing: A Journal of Policy and Practice, 12(1), 116 to 119. Available online [Accessed 7th February 2025]

Steele, M. J. (2020). Effects of Different Types of Forensic Evidence on Arrest Probability: Toward a New Typology of Evidence. Criminology, Criminal Justice, Law & Society, 21(3), 17 to 38. Available online [Accessed 7th February 2025]

Struyf, P., De Moor, S., Vandeviver, C., Renard, B., and Vander Beken, T. (2019). The effectiveness of DNA databases in relation to their purpose and content: A systematic review. Forensic Sci Int, 301, 371 to 381. Available online [Accessed 7th February 2025]

Tiry, E., Zweig, J., Walsh, K., Farrell, L., and Yu, L. (2022). Beyond Forensic Evidence: Examining Sexual Assault Medical Forensic Exam Mechanisms that Influence Sexual Assault Case Outcomes. Journal of Interpersonal Violence, 37(7/8), NP5693 to NP5727. Available online [Accessed 7th February 2025]

Todak, N., Gaub, J. E., and White, M. D. (2022). Testing the Evidentiary Value of Police Body-Worn Cameras in Misdemeanor Court. Crime & Delinquency. Available online [Accessed 7th February 2025]

Tully, L. (1998). Examination of the Use of Forensic DNA Typing from Two Perspectives. (PhD.). The University of Maryland, Maryland. Available online [Accessed 7th February 2025]

Vakhitova, Z., Iliadis, M., Harris, B., Tyson, D., and Flynn, A. (2022). The merits and risks of body-worn camera footage in domestic and family violence incidents and legal proceedings: a study of police perceptions and experiences. Policing and Society, 1 to 17. Available online [Accessed 7th February 2025]

Wallace, D., White, M. D., Gaub, J. E., and Todak, N. (2018). Body‐Worn Cameras as a Potential Source of Depolicing: Testing for Camera‐Induced Passivity. Criminology, 56(3), 481 to 509. Available online [Accessed 7th February 2025]

White, M. D., Gaub, J. E., Malm, A., and Padilla, K. E. (2021). Implicate or Exonerate? The Impact of Police Body-Worn Cameras on the Adjudication of Drug and Alcohol Cases. Policing: A Journal of Policy and Practice, 15(2), 759 to 769. Available online [Accessed 7th February 2025]

White, M. D., Gaub, J. E., and Todak, N. (2018a). Exploring the Potential for Body-Worn Cameras to Reduce Violence in Police–Citizen Encounters. Policing, 12(1), 66 to 76. Available online [Accessed 7th February 2025]

White, M. D., Todak, N., and Gaub, J. E. (2017). Assessing citizen perceptions of body-worn cameras after encounters with police. Policing: An International Journal of Police Strategies & Management, 40(4), 689 to 703. Available online[Accessed 7th February 2025]

White, M. D., Todak, N., and Gaub, J. E. (2018b). Examining Body-Worn Camera Integration and Acceptance Among Police Officers, Citizens, and External Stakeholders. Criminology & Public Policy, 17(3), 649 to 677. Available online [Accessed 7th February 2025]

Wilson-Kovacs, D. (2021). Digital media investigators: challenges and opportunities in the use of digital forensics in police investigations in England and Wales. Policing-an International Journal of Police Strategies & Management, 44(4), 669 to 682. Available online [Accessed 7th February 2025]

Wilson, D. B., McClure, D., and Weisburd, D. (2010). Does Forensic DNA Help to Solve Crime? The Benefit of Sophisticated Answers to Naive Questions. Journal of Contemporary Criminal Justice, 26(4), 458 to 469. Available online [Accessed 7th February 2025]

Wilson, D. B., Weisburd, D., and McClure, D. (2011). Use of DNA testing in police investigative work for increasing offender identification, arrest, conviction and case clearance. Campbell Systematic Reviews, 7. Available online [Accessed 7th February]

Woodman, P., Spiranovic, C., Julian, R., Ballantyne, K. and Kelty, S. (2020) The impact of chemical trace evidence on justice outcomes: exploring the additive value of forensic science disciplines Forensic Science International Vol 307.Available online [Accessed 7th February 2025]

Yokum, D., Ravishankar, A., and Coppock, A. (2017). Evaluating the Effects of Police Body-Worn Cameras: A Randomized Controlled Trial. Washington, DC.: The Lab @ DC. Available online [Accessed 7th February 2025]

Question 6

Annex A: Keywords and search terms

Accepted Answer

Increased use of new technology: body-worn cameras (BWCs)

‘body worn camera’ OR ‘BWC’ OR ‘bodycam’ OR ‘body worn video’ OR ‘BWV’

AND

‘Police’ OR ‘CPS’ OR ‘Prosecution’ OR ‘Courts’ OR ‘Prison’ OR ‘Probation’ OR ‘Criminal Justice System’ OR ‘CJS’

AND

‘Case Progression’ OR ‘Attrition’ OR ‘Prosecute’ OR ‘Charge’ OR ‘Convict’ OR ‘detect’ OR ‘clearance’ OR ‘arrest’ OR ‘outcome’ OR ‘impact’ OR ‘resources’ OR ‘costs’ OR ‘delay’ OR ‘backlog’ OR ‘guilty plea’

Use of forensic evidence by the police and prosecutors: forensics and impacts on case progression

‘Police Officer’ OR ‘Prosecute’ OR ‘SOCO’

AND

‘DNA’ OR ‘Forensic’

AND

‘Case Progression’ OR ‘attrition’ OR ‘Prosecution’ OR ‘Conviction’ OR ‘Delay’ OR ‘Sentence’ OR ‘Resource’ OR ‘charge’ OR ‘detect’ OR ‘clearance’ OR ‘arrests’ OR ‘criminalistics’ OR ‘proceeded against’ OR ‘case outcome’ OR ‘offences brought to justice OR ‘criminal justice system’ OR ‘CJS’

Question 7

Annex B: Quality assessment criteria

Accepted Answer

Outlined below are the criteria for selecting primary studies for inclusion based on relevance and quality.

Primary/key studies - studies must be (a) directly relevant and (b) high quality:

1. Identifies changing demands within the CJS because of BWCs or DNA when:

an intervention is made
a change in law/policy/practice occurs
there is a demonstrable change in context (that is, changes in resources)
a change in crime patterns (rates, volumes, and/or complexity)

2. Includes the following CJS outcomes: CJS upstream – charges, detections, clearance, arrests – but only if charging or downstream outcomes were also covered in the analysis, and cases accepted by the prosecution. CJS downstream – suspects proceeded against at court, cases going to trial, convictions, prison places or sentence length. Plus – guilty pleas, timeliness to intermediate or final outcome.

3. Must be of sufficient interest/applicability to the current UK CJS.

4. High methodological rigour (sampling, quasi-experimental methods, direct measures of variables, variables well operationalised, analysis that controls for the effects of other variables) with a high level of confidence that there is a causal relationship between an intervention, change in policy, practice, context, or demand on the one hand, and CJS outcomes on the other.

Secondary studies – more indirect relevance and/or lower quality:

more descriptive studies where changes are identified based on the characteristics of the case or area (in cross-sectional studies) but without robust cause and effect evidence
indirect/implicit measures of CJS outcomes
less focus on changing demand
poor applicability to current UK CJS
less robust in terms of methods, samples
also includes those studies which might be high in methodological rigour but of low relevance to research questions

Question 8

Annex C: Tables of literature reviews, key studies and secondary studies for forensic/trace DNA studies

Accepted Answer

Use of forensics/DNA tables

Table C.1: Summary of some of the key literature reviews in the area

Author(s) and review date	Type of review (method and focus)	Type(s) of forensic evidence	Number of identified eligible studies	Main conclusions
Bradbury and Feist (2005)	Literature review; volume crime	All forensics	50	- Estimate one-quarter of volume crimes detected using forensics. - As well as corroboration, DNA can be used to identify unknown suspects (7/10 cases). - Provide suspect linkage to other cases. - Role in detection of hard-to-solve crimes. - CJS outcomes overall not significantly different until split into specific crime types. - Forensics increase likelihood of conviction and longer sentence in burglaries, rapes, and homicides.
Wilson et al. (2010, 2011)	Campbell systematic review	DNA	5	- Highlight the methodological weaknesses of most work in the field. - Only one RCT (Roman et al., 2008; US; burglaries). - Modest to large improvements in suspects identified, arrested, and prosecuted in burglaries across the US. - Inclusion of 4 QE studies exploring wider range of crime types but note inability to control for other confounding factors. - Some crossover of improvements to other crime types, such as serious violent crime, but evidence is weaker due to methodology.
Ludwig and Fraser (2014)	Systematic review of ‘grey’ literature on forensics in volume crime	All forensics	36 reports	- Thematic analysis of government literature mostly from the UK. - Little systematic evaluation of forensic impacts. - Focuses on factors influencing ineffective implementation of forensic science.
Struyf et al. (2019)	Systematic review of effectiveness of DNA databases	DNA	18	- Examines their effects in terms of detections and clearance rates as well as the broader effects on deterrence. - Screens literature by quality of research design includes RCT, QE, and cross-sectional studies. - Direct impacts of DNA on suspect ID matches but more difficult to establish direct impact on further case progression stages as other extraneous factors cannot be controlled for. - DNA appears to lead to more detections and charges. - Reductions in both re-offending and general crime rates associated with effects of DNA databases but differs by crime type.

Table C.2: Secondary studies and summaries for forensics/DNA

Author(s) and date	Notes
Briody (2002; 2004; 2005; 2006)	Thesis and publications from this. Quasi-experimental. Australia. Compares outcomes of cases with DNA with those without. Focuses on sexual offences, property crimes, and homicide cases. Attempts to find reasonable comparisons/controls in terms of some case matching criteria (for example, case seriousness), and use of logistic regression to control for some other identified confounding variables. But likely to be other differences between DNA and non-DNA cases. DNA presence greatly increased prosecutions, convictions and impacted on sentences, but varies by crime type.
Burrows et al. (2005)	Intervention evaluation of expansion of DNA collection in the UK. Although a quasi-experimental study comparing pre- and post-DNA expansion data, they do not control for other factors which may have influenced case outcomes. Findings reveal some slight improvements in detection rates and convictions for pilot areas compared to non-pilot areas.
Kelly et al. (2005)	This evaluation provides an examination of the effectiveness of SARC interventions in 3 pilot areas. They compare case outcomes for sexual assaults in those areas with control areas and found that in 2 of the areas SARC involvement led to significantly higher conviction rates. Conviction rates did, however, remain low, and it is not possible to separate out the specific effects of DNA collection as this was only one aspect of a broader intervention package.
Schroeder (2007) Schroeder and White (2009)	The first reference is the full government report, and the second the published journal article. They compare the outcomes of cases with DNA with those without in homicide cases in the US. No attempt to match cases in both DNA tested and not tested groups. Small sample sizes. Higher clearance rates for those without DNA evidence, but probable selection bias in that DNA only resorted to in hard-to-solve cases, those without were stronger cases with a high success rate.
Dunsmuir et al. (2008)	Thesis. Time series analysis of expansion of DNA database in New South Wales, Australia. Quasi-experimental. They could not test for other variables that might have affected patterns of case outcomes (DNA versus non-DNA tested) over time – such as other changes in police procedures. Changes in charge rates over time, but dependent on crime types.
Campbell et al. (2008; 2009; 2011; 2012; 2014)	Body of work assessing the impact of SANE and/or SART interventions across sites and over time in the US. Some of these studies include robust case progression outcomes, such as whether referred to trial, prosecuted, guilty plea, or conviction. Intervention, however, is a broader package of support for victim including a forensic examination, but also other forms of victim support (medical, legal, therapeutic).
Peterson et al. (2010) Johnson et al. (2012)	These studies use the same cross-sectional dataset across 5 areas of the US in 2003. Dataset includes aggravated assaults, rapes, homicide, robberies, and burglaries. The first reference is to the larger government report, while the second reference is a journal article focusing on rape cases. They found positive significant effects of collection of physical crime scene evidence on arrests, prosecution referrals, charges, and convictions across crime types. Although there were some variations between crime types, and other case factors also significantly predicted outcomes. Some of the reported effect sizes were large. But they did not look for DNA evidence specifically which they found was rarely utilised except in homicides. Johnson et al. (2012) reported that forensic evidence significantly impacts on case progression in rape cases but not as much as other factors, such as victim participation. DNA evidence was also rarely used in rape cases.
Baskin and Sommers (2010; 2011; 2012)	Cross-sectional study across 5 jurisdictions in the US, re-analysis of Peterson et al. (2010) data claiming weaknesses of original analysis. Unlike the original analysis, they pool data across sites, but also control for selection bias. First paper examines homicide data, the second study examines residential burglaries, and the third assaults and robberies. They found no significant effect for the presence of DNA evidence on case outcomes. Instead, they highlight the significance of victim/witness participation as a predictor of arrests, prosecution referrals and charges in all studied crime types.
Peterson et al. (2013)	This study is a re-analysis of the Peterson et al. (2010) dataset after the contradictory findings of Baskin and Sommers (2010, 2011, 2012). In contrast to the previous analyses, they aggregate the data across crime types (although do grade them in terms of crime seriousness) and jurisdictions. They find that forensic evidence was 3 times more likely to lead to arrests. If analysed, the evidence was twice as likely to lead to prosecution referral, 4 times more likely to be charged, and 2 to 3 times more likely to be convicted. Baskin and Sommers (2013) further disputed their findings and criticised their use of aggregate crime data, and lack of controls for selection bias.
McEwen and Regeoczi (2015)	Uses 294 homicide cases in Cleveland, Ohio, US, to examine the extent to which different types of forensic evidence predict case closure. While some other forms of forensic evidence did significantly aid in solving the case, DNA evidence was negatively related suggesting that for cases that used DNA evidence an arrest was significantly less likely to be made. They also looked at not just the presence of DNA, but whether a match was made. The closure rate was slightly higher (but not significant). However, in most cases (128 cases compared to 23), DNA results were returned only after the case was closed.
Schroeder and Elink-Schuurman (2017)	Similar in design to Peterson et al. (2010), they provide a cross-sectional analysis of cases using DNA evidence compared to those without across crime types (that is, assault, burglary, robbery, rape, and homicide) from only one US state, over 3 years. Significantly they measure whether evidence is used rather than gathered. They use logistic regression to measure impacts on arrests, convictions, and sentence lengths. Their findings were inconsistent with Peterson et al. (2010) – they largely found no significant effects of forensic evidence on arrests, convictions, and sentence lengths.
Nir and Griffiths (2018)	Re-analyses cross-sectional data from Peterson et al. (2010, 2013). They add in the effects of type and quantity of forensic evidence on sentence length and found that for violent crime, the use of forensic evidence led to longer sentences.
Steele (2020)	Re-analyses case files from Peterson et al. (2010, 2013) to examine the impact of forensics on arrest probability across 5 crime types in 5 US cities. Highlights the effect of situational variables on arrest probability rather than forensic evidence. Concludes that forensic evidence is not particularly useful in making arrest decisions, except in cases where they lack other investigative leads.
Woodman et al. (2020)	Explores the impact of various biological traces, including chemical traces and biometric evidence (DNA) on CJS outcomes. Looked at 238 cases with forensics collated with data from police investigation and court processes. Offences were crimes against the person, property, and other (mainly weapons). Found DNA traces were a statistically significant predictor of court outcomes. Chemical trace evidence in combination with ballistics or tool marks was also statistically significantly likely to predict court outcomes. These were not a random sample of cases – cases were selected only if they included chemical trace evidence.
Cross et al. (2022)	US cross-sectional. Compares DNA and non-DNA sexual assault case progression. Findings suggest that DNA cases were more likely to progress to a prosecution and more likely to be convicted. For example, half of DNA cases led to conviction compared to less than one fifth of non-DNA cases.
Kettrey et al. (2022)	Similar to the work of Campbell and colleagues, this work tests the impacts of SANE intervention in the US on the likelihood of the arrest of the perpetrator. However, the focus is on how the use of Sexual Assault Medical Forensic Examinations (SAMFEs) impact on victim engagement, which then impacts on case outcomes. They conclude there is no direct effect of DNA analysis but rather an indirect effect via victim co-operation in the case.
Tiry et al. (2022)	As above, this work explores whether the presence of a SAMFE in the US predicts the likelihood of achieving criminal case processing outcomes and whether the relationship of the SAMFE with criminal case processing outcomes is moderated by other legal and extra-legal case characteristics. They conclude that participation in the forensic exam did not predict suspect ID, arrest, or conviction.

Table C.3: Key studies and summaries for forensics/DNA

Authors and date	sample	method
Roman et al. (2008)	US, case file data of 500 crime scenes in each of five police sites across the US.	Randomised Control Trial (RCT) of effects of DNA testing on property crime case outcomes, across five sites.

Relevance and quality - intervention expanding use of DNA testing to property crimes, and more timely analysis of DNA evidence. Measure outcomes in terms of whether suspect identified, arrest made, and referral for prosecution. They were not able to track cases to final dispositions.

Findings - those property crime cases allocated to the DNA analysis condition (as opposed to just a traditional investigation) were more than twice as likely to identify a suspect (in 31% of cases versus 12%) and result in an arrest (16% of cases versus 8%). They were twice as likely to be prosecuted (19% versus 8%). DNA analysis was significantly more likely than fingerprint analysis to lead to a suspect match and an arrest. The profiles of suspects identified via DNA matches were also different – they were significantly more likely to have been arrested and convicted– representing more serious prolific property offenders. Although there were considerable variations between sites, all five sites showed greater levels of case progression in test conditions compared to the control data.

Implications - DNA analysis in cases such as property crime (which has low clearance rates) can reduce case attrition. An increase in arrests and referrals for prosecution will inevitably lead to an increase in demand downstream, with more suspects entering the system and facing criminal charges. The impact is larger given the higher proportion of property crimes. We do not know from this study, however, how this translates into terms of final case dispositions, that is, whether DNA analysis of property crimes also leads to more convictions and/or longer sentences.

Authors and date	sample	method
Bhati (2011) Bhati and Roman (2014)	US, re-offending data from violent, property and drug offenders in Florida from 1996 to 2004.	QE design comparing re-offending data from released offenders included in the DNA database with those not included. Data is adjusted/ weighted for differences between test and control samples. Longitudinal data is used to examine staggered changes in DNA database expansion over time.

Relevance and quality - Measure impacts on re-offending of an expansion of the DNA database, include a wider range of crime types. They control for expected experiential re-offending changes. They also weight the data to account for some other differences between test and control groups.

Findings - This study separates out the specific deterrent effects of DNA databases (that inclusion will reduce an individual offender’s re-offending due to fear of capture) from probative effects (that inclusion will make the offender easier to capture and convict if they re-offend). They found that there was a small but significant deterrent effect for both robbery and burglary (2 to 3% reduction in re-offending). Probative effects of DNA were however much higher – ranging from 20 to 30% increases in rearrests and (to a lesser extent) reconvictions across all crime types.

Implications - Inclusion in the DNA database was found to have a slight deterrent effect on burglary and robbery offenders, but the probative effect of DNA was much stronger. Therefore, re-arrest and reconvictions increased across all crime types as offenders on the database were subsequently easier to apprehend and convict. So, while there are some small decreases in demand associated with deterrence of re-offending, the probative effects of DNA are more likely to lead to larger increases in demand especially in terms of making more arrests for repeat offenders already entered in the database.

Authors and date	sample	method
Doleac (2017)	US, re-offending, and crime rate data from seven US states from 2000 to 2010.	QE design. Pre- and post-DNA database expansion. Focus on felony/serious offences. Longitudinal data over several states. Follow-up re-offending period of five years.

Relevance and quality - QE study that measures both specific and aggregate deterrence using re-offending data and data on state crime rates over time. They demonstrate that both test and control groups look similar in terms of their criminal histories so some attempt to match data groups.

Findings - there were significant reductions in re-offending rates. This was most marked for serious violent crime where there was a 21% reduction. Serious property crime re-offending decreased by 3% but was not significant. This suggests some deterrent effects, particularly for violent offenders. As more offenders’ profiles were included in the DNA database over time there were significant reductions in crime rates. The addition of one DNA profile per 100,000 citizens led to a reduction of 0.05 property offences, and 0.32 violent offences.

Implications - the expansion of DNA databases led to some significant reductions in demand both in terms of a lower recidivism rate, but also a reduction in crime rates. These reductions were more marked for violent offences than property offences. These specific deterrence effects are likely to lead to reductions in downstream demands as fewer offenders are re-entering the CJS over the next five years. Furthermore, the aggregate deterrence in terms of slightly lower crime rates should also impact on lowered downstream demands.

Authors and date	sample	method
Anker et al. (2021)	Denmark, re-offending data comparing offenders on the database with those not included. Large national sample of 39,674 offenders over nearly 67,000 observations.	Pre- and post-DNA database expansion. Unlike the US studies, this involved a rapid reform to the DNA database at one point in time. Follow-up re-offending rates are measured for the next three years. Analysis used DID analysis.

Relevance and quality - this quasi-experimental study examines the deterrence effects of an intervention (rapid expansion of the NDNAD) on recidivism. They match and balance the samples to control for other extraneous variables. Unlike in the previous study, in the US they separate out the effects of deterrence from increased risk of detection (given the probative value of DNA databases).

Findings - there were significant reductions in re-offending for offenders who were included in the expanded DNA database. These were large and significant – overall a 43% reduction in re-offending, and a 42% reduction in reconvictions, in the first-year post-expansion. As with their previous study, the deterrent effects were larger on serious violent crimes than property crime. First-time offenders, those with children, and those whose first offence was violent were more likely to be deterred from re-offending. They were also more likely to engage in positive life changes post registration. Offenders were also more likely to be detected post-expansion, due to the probative effects of the database. But the deterrent effects were much stronger - they estimate that for every 1% increase in likelihood of detection, crime is reduced by 2.7%. There was also some evidence of aggregate deterrence effects.

Implications - expansion of the DNA database to include all serious offenders in Denmark led to some large reductions in demand. Offenders were much less likely to be re-arrested and reconvicted of serious offences, especially those involving violence. This would inevitably lead to downstream demand reductions, with fewer offenders re-entering the CJS. These effects persisted for the first three years after the expansion. Any increases in downstream demand caused by the probative value of DNA – that is, increased detections of DNA profiled offenders – were more than mitigated by the reductions in recidivism.

Question 9

Review of the secondary studies of forensic and trace DNA and the relationship with CJS outcomes

Accepted Answer

DNA retrieval and arrests

With the advent of DNA, it soon became apparent that the DNA traces at crime scenes meant crimes could be linked and linkages between suspects to crime scenes could be made, the expectation being that it would lead to unsolved crimes being solved. There is some evidence that this has happened. Burrows et al. (2005a) concluded that without forensic evidence establishing a ‘first link’ between a crime scene and a suspect, 24% of volume crimes would be difficult or even impossible to detect. In Burrows et al.’s (2005b) evaluation of the Pathfinder pilot in the UK, forensic material (and, in particular, DNA) provided the first link between the crime scene and suspect in 45% of cases where crime scene DNA matched with profiles on the NDNAD.

In a small-scale study in the US, Schroeder (2007) identified that out of 40 homicide cases, trace DNA helped identify the suspect in 16 cases. Mapes et al. (2015) found that in a very small proportion of cases in The Netherlands, trace DNA yielded ‘cold hits’ allowing an unknown suspect to be identified and linked to the crime. Although this can prove useful in cases lacking other leads, it happened rarely in only 3% of serious crimes and 1% of volume crimes. Bradbury and Feist (2005) concluded that trace DNA had a more significant role in first linkage than previously suspected. It aided in the improved detection of crime because between 51 to 75% of DNA matches led to detection.

There is, therefore, an expectation that trace DNA will enable a breakthrough in the investigation and lead to an increase in suspect arrests and a subsequent increase in downstream demand as more cases go through the courts. Peterson et al. (1984) (US) found that cases involving ‘scientific analyses’ were approximately 3 times more likely to be cleared by the police, and that these tended to be in cases where other evidence was weak. It is important to note, however, that they did not break this down into the more specific effects of DNA evidence. Later studies have found similar results when looking at all crime types combined. For example, Peterson et al. (2013) (US) found that arrests were made in 47% of cases where forensics were collected (compared with 32% where none were taken).

The findings are less consistent when data is disaggregated into specific crime types. Several studies in the US have found that forensic evidence predicted a significantly higher likelihood of arrest in violent (Johnson et al., 2012; Peterson et al., 2010) and sexual offences (Johnson et al., 2012; Peterson et al., 2010), robberies and burglaries (Peterson et al., 2010). Looking at trace DNA alone, research has also found higher arrest rates in sexual offences, robberies, and burglaries in Australia (Dunsmuir et al., 2008). But these findings have been contested, and other US studies have failed to find any significant effects of forensics on arrests across a range of crime types (Baskin and Sommers, 2011, 2012; Schroeder and Elink-Schuurman, 2017; Tiry et al., 2022). Steele (2020) split forensic evidence into different types concluding that there were important differences between them and found that DNA evidence had no effects across crime types.

With regard to US homicide, cases with forensics (Baskin and Sommers, 2010; Peterson et al., 2010), and trace DNA more specifically (Schroeder, 2007), are no more likely to result in an arrest. In fact, some US evidence suggests both DNA traces (Schroeder and White, 2009), and biological traces more broadly (McEwen and Regoeczi, 2015), may even be associated with a lower arrest likelihood. This may be because DNA is only analysed in homicide cases that already have weak evidence (Wilson et al., 2010). Much of this research relies on cross-sectional designs which compare case progression for those cases with trace DNA available, with those without, but do not adequately account for the other ways in which these cases are likely to differ.

For example, in their review of police use of forensic evidence, Bradbury and Feist (2005) additionally found large variations in terms of how UK police forces converted suspect matches into arrests. In terms of volume crime, 27% of matches did not lead to detection (Burrows et al., 2005b) for a variety of reasons (that is, elimination of the suspect due to them having legitimate access to the site, insufficient evidence). This outlines why simply recovering the DNA evidence is not an important metric when looking at the effectiveness of DNA. The presence of DNA may simply identify the fact the suspect has legitimate access to the crime scene or confirms the identity of the victim – background noise (Margot, 2007). What is important is the utility of the DNA such that new information is gleaned from the DNA traces about the crime and the offender (Hoffmann et al., 2024).

DNA and the impact on case progression

Struyf et al. (2019) examined the effectiveness of DNA databases and conclude that while it is easy to examine the number of matches made between crime scenes and suspects, it is much less easy to find a direct link between these matches and subsequent arrests. While it was hoped that it would speed up the investigation and clearance of crimes, traces of DNA take time to process, and results are often returned too late in an investigation to lead to a direct detection. DNA traces are more likely to be analysed from ‘hard-to-solve’ cases where a lead suspect is not readily identified, but typically cases are cleared through other means before DNA results are readily available (Schroeder and Elink-Schuurman, 2017). In a review of 600 homicide cases in the US, Schroeder and White (2009) found that detectives did not tend to rely on DNA traces in their investigations, only submitting it for analysis in the absence of other lines of enquiry or collecting it in the anticipation that it would strengthen a prosecution at trial.

In the US, McEwen and Regoeczi (2015) found that only 17% of homicide cases had a DNA match back prior to making an arrest, but while DNA traces are unlikely in these cases to directly lead to an arrest, they have a wider usage in cases where a suspect has already been identified, by corroborating that the suspect was present at the crime scene and adding weight to prosecution evidence. DNA traces have also proved important in linking different crime scenes together. In the UK Burrows et al. (2005b) found that 17% of forensic matches connected multiple crimes to the same offender, which often led to separate charges being made against them.

Although these literature reviews indicate that trace DNA can be useful at an investigative level, the impact of trace DNA on outcomes at the downstream end of the CJS is quite complex. Furthermore, the impacts are very dependent upon the crime type. As DNA trace analysis becomes more commonplace as part of a police investigation, increases in demand may not match police capacity to process, causing potential backlogs in the system. This has already been recorded in the US in terms of unprocessed forensic material from sexual assault cases, for example Campbell et al. (2017) and Rog et al. (2021). It is therefore important to ascertain the nature and extent of potential changes in CJS downstream demand associated with DNA usage, as there may be unanticipated effects in terms of delays and bottlenecks in the system.

DNA was felt to be useful not only in terms of identifying unknown suspects in ‘hard-to-solve’ cases, but also in enhancing the probative value of evidence. Research suggests that both police and prosecutors find DNA useful even when it does not lead to an arrest (Alderden et al., 2021; Henry and Jurek, 2020; Kettrey et al., 2022; Kjærulff et al., 2019; Menaker et al., 2017). Studies have therefore sought to examine whether traces of DNA in cases furthers case progression and leads to more convictions. Although the use of DNA evidence may not directly lead to arrests and case closure, the results are theorised to be additive. The corroborative probative value of DNA evidence is proposed to enhance the strength and quality of evidence, which should have knock-on effects in terms of case progression.

With stronger evidence, it is expected that more criminal charges and prosecution referrals will be made, and ultimately more convictions gained. This would inevitably lead to increases in downstream demand, although it might also be expected that with higher evidentiary weight in a case the suspect would also be more likely to accept a guilty plea, so some savings can be potentially made if the case avoids a full trial.

There is now a considerable body of literature exploring the impacts of forensics/ traces of DNA on case progression and outcomes, although again they rely heavily on cross-sectional research or quasi-experimental designs without adequate matched control groups (Wilson et al., 2010; Wilson et al., 2011). There also exists a body of literature, mainly from the US, researching the impacts of interventions to improve the timely collection of high-quality forensic evidence in sexual assault cases. We will draw from this literature in the synopsis to follow but will also discuss their limitations. Table C.3 in the Annex C summarises the findings from our secondary studies. Only one experimental study exists to date (Roman et al., 2008), which is not surprising given the ethical implications of randomly assigning DNA analysis to some cases and not others (Wilson et al., 2011).

It is also worth noting, however, that as a case progresses through the CJS, it becomes more and more difficult to disentangle the impacts of DNA evidence from other factors (Struyf et al., 2019).

DNA and the impact on charge rates and prosecutions

Several studies, both cross-sectional and quasi-experimental, find a significant increase in charge rates and prosecutions from forensic evidence cases, although unfortunately much of the literature here does not disaggregate the effects to DNA evidence more specifically. Peterson et al. (2013) concluded that forensic evidence significantly helped in decisions to prosecute and charge across all crime types. Cases were charged in 32% of cases with forensic evidence (compared with 14% without). For burglary, robbery, assault, rape, and homicide forensic cases were (at least) 4 times more likely to be charged (Johnson et al., 2012; Peterson et al., 2010; Peterson et al., 2013). Other research has found that the results differ dependent upon crime type.

In quasi-experimental studies testing cases before and after a DNA database expansion in Australia, it was found that DNA evidence improved the charge rates with more cases progressing to prosecution in property crimes (burglaries and vehicles offences) (Briody, 2005, 2006), robberies (Dunsmuir et al., 2008), homicides (Briody, 2004, 2005), and sexual offences (Briody, 2002, 2005; Dunsmuir et al., 2008). Cross et al. (2022) also found that for sexual offences cases in the US, DNA cases were statistically significantly more likely to go to court. Other US studies have failed to find any significant effects of forensics on charge rates and prosecutions (Baskin and Sommers, 2010, 2011, 2012; Schroeder and Elink-Schuurman, 2017), which makes any clear conclusions difficult in this area.

The effects of forensic evidence on the downstream CJS are therefore contentious and have led to some controversy. For example, Baskin and Sommers were originally researchers on the Peterson et al. (2010) study exploring the impact of forensics on case outcomes. They argued that the findings presented in that work were flawed as the study failed to disaggregate different crime types (Baskin and Sommers, 2013).

In their re-analysis of Peterson’s data (splitting offences by crime type) they failed to replicate the original findings (Baskin and Sommers, 2010, 2011, 2012). Peterson et al. (2010) responded with their own re-analysis of the data, but without accounting for differences between crime types. Although Baskin and Sommers’ research (2010, 2011, 2012) seem more reliable as analysis is by crime type, unfortunately they also potentially conflate variation between different types of forensic evidence as they are unable to eliminate the effects of DNA evidence in isolation from other forms of biological forensics.

While there is some support that forensics and/or DNA cases are more likely to receive a criminal charge and be prosecuted, the evidence for a subsequent impact on conviction rates is more limited. For example, in their cross-sectional research in the US, Peterson et al. (2010, 2013) concluded that across crime types, forensic cases were twice as likely to be convicted. Other researchers using the same dataset, however, found no such effects once types of crimes and types of forensic evidence were disaggregated, and once other significant predictors were considered (Baskin and Sommers, 2010, 2011, 2012).

Examining DNA more specifically, quasi-experimental work in Australia found that DNA was significantly more likely to lead to conviction in homicides (Briody, 2004, 2005; Tully, 1998), serious assaults (Briody, 2005), and sexual offences (Briody, 2005). Both DNA and wider forensic studies suggest that other factors, such as victim and witness participation, were much more reliable predictors of conviction. Similarly, an earlier study by Peterson et al. (1986), showed that although conviction rates were higher for cases involving forensic evidence (in 5 out the 6 jurisdictions they studied), other extra-legal factors were more predictive of convictions. It is therefore apparent that the relative value of DNA on case outcomes depends upon both crime type and crime context.

For example, DNA evidence may have an important contribution to aiding the investigation of burglaries but prove less effective in robberies. Within sexual violence cases DNA’s effectiveness is likely to depend on the relationship between the victim and offender – that is, it will prove more useful in cases where the victim and offender are strangers. Woodman et al (2020) also point out that many of the above studies do not consider whether the DNA has been analysed, only whether DNA traces were collected in the first place. Once the utility of the DNA trace was factored in, DNA was found to be a predictor of court outcomes.

The downstream demands are likely, however, to be different dependent on how a conviction is reached. Therefore, it is important to disaggregate guilty plea verdicts from trial convictions. If DNA evidence does, as proposed, improve the quality of evidence, it is expected that more guilty pleas will be expected from suspects given the evidentiary weight of the prosecution case against them. The research evidence is less extensive as many studies do not distinguish between types of convictions. There is some evidence that cases with DNA evidence were significantly more likely to result in a guilty plea for sexual offences – in the US (Tully, 1998) – and property crime – in Australia (Briody, 2005, 2006).

Impact on case progression from sexual assault interventions

Useful evidence can also be drawn from quasi-experimental research evaluating various interventions for victims of sexual assault in the US, and to a lesser extent, the UK. Research aimed at reducing attrition in sexual assault cases in the US have looked extensively at the impact of SANEs (Campbell et al., 2014a; 2012; 2009), SARTs (Campbell et al., 2011), and SAMFE programmes (Kettrey et al., 2022; Tiry et al., 2022) on case progression outcomes. In the UK, Kelly et al. (2005) evaluated the impact of SARCs. Although there are variations between interventions, SARCs tend to work collaboratively with the police and other CJS professionals, with an aim to support the victim throughout the CJS process. This includes supporting them during the investigative aspects such as initial interviews with police and collection of forensic evidence, as well as support (both legal and psychological) through trial proceedings, and providing access to psychotherapeutic services.

A key aim of such interventions is to facilitate more timely recovery of high-quality DNA material. Evaluations of these intervention studies tend to track case progression to CJS outcomes, so they are a good indicator of whether improvements in the quality of evidence have demonstrable impacts on the downstream CJS. But it is important to highlight that as these intervention programmes are multifaceted it is impossible to determine the effects of DNA evidence alone. Overall, evaluations tend to show that involvement in these interventions lead to higher charge volumes. For example, Campbell et al. (2008) evaluated the SANE programme’s effectiveness in one Midwestern state comparing pre- and post-intervention data over a series of longitudinal studies (comprising 12 years of data).

The SANE programmes were found to significantly impact on case outcomes, after controlling for other variables. Cases from SANE were more likely to be referred for prosecution, were more likely for the defendant to accept a guilty plea and were more likely to result in a conviction. Overall, SANE cases were 70% more likely to attain greater case progression compared with those prior to the intervention.

Campbell et al. (2014a) went on to evaluate SANE programmes across 6 sites, comprising almost 1,700 sexual assault cases, and concluded that cases with SANE intervention were statistically significantly more likely to be prosecuted. Although the prosecution rates remained very low, and they failed to replicate the more dramatic effects seen in the previous 2008 evaluation. In the UK, rape cases were more likely to be discontinued during the investigative level if victims were not given SARC intervention, and 2 of the 3 SARC areas had higher conviction rates compared to the control areas (Kelly et al., 2005).

As these interventions do not just improve the quality of DNA evidence, but have a much wider remit, it is it is difficult to establish which aspects of these broad interventions most successfully impact on case progression. It is likely that the timely collection of forensic evidence contributed to an increased evidentiary burden of proof, therefore having some indirect effects on case outcomes, but unfortunately this research did not examine the precise effects of the different components of the intervention (that is, the effects of the obtained DNA evidence alone). This is supported by more recent research in the US which looked at the impacts of just the SAMFE aspect rather than the full SANE intervention (Kettrey et al., 2022; Tiry et al., 2022). They concluded that there were no direct effects of DNA on case outcomes.

It is therefore most likely that DNA may have indirect effects on case progression as the evidential weight of DNA may make the police and prosecutors more likely to pursue the case (Kettrey et al., 2022), and those who participate in the interventions are both more likely to receive a forensic examination and be supported throughout the prosecution of the case.

Impact on sentencing length

While forensic evidence and traces of DNA are theorised to impact on increases in demand in terms of more arrests, prosecutions, and convictions, the literature also suggests some unanticipated impacts on sentencing. Early cross-sectional research in the US suggested that cases involving forensic evidence were associated with higher rates of incarceration, and longer sentences (Peterson et al., 1986). In a later study (Peterson et al., 2013) also found that sentences were significantly longer when crime scene evidence was collected (224 months compared to 33 months) and analysed (300 months compared to 56 months).

The impact on sentencing may also differ dependent upon the crime type. For example, Peterson et al. (2010) found that forensic evidence cases were significantly associated with longer sentences in assault cases, but it was the suspect’s criminal history which was the most significant predictor of a custodial sentence. But this research has been criticised for failing to control for other influencing factors, and to disaggregate both crime types and specific types of forensic evidence (Baskin and Sommers, 2013). Nir and Griffiths (2018) examined a range of evidential, offender, and case factors’ impacts on sentencing of violent crimes, and found that cases with a wide variety of evidentiary types, especially forensic evidence, were associated with longer sentences. But again, they did not look at DNA in isolation to other forms of biological forensic evidence.

There is comparatively little evidence regarding the impact on sentencing from DNA traces more specifically. One US quasi-experimental study by Tully (1998) compared pre-DNA database and post-DNA database cases (although in some cases they have control groups of post-database cases lacking in DNA evidence). The study found a positive relationship between cases with DNA traces and outcomes and found that sexual offences with DNA traces led to significantly longer sentences than those without. Further evidence in Australia found that cases post-DNA database expansion were more likely to be given a custodial sentence for property crimes (Briody, 2002, 2006), but shorter sentences (on average 5 months shorter) for homicide and manslaughter (Briody, 2004).

It is difficult to ascertain what causes these sentencing disparities, especially as much of the work is cross-sectional or quasi-experimental, and cases with and without DNA traces, or pre- and post-expansion comparisons, are likely to differ in other ways that the analysis cannot adequately control for. Briody (2004) suggests that the shorter sentences in homicide and manslaughter cases may be due to the longer processing times of DNA, so that by the time the offender is convicted they have spent more time in remand custody. Whereas the longer sentences shown for property crimes in DNA cases may potentially be because the offender profiles differ between DNA and non-DNA cases. Prolific offenders are much more likely to have their DNA profile stored in a DNA database than less prolific offenders, so it may well be that the higher sentencing tariff is the result of the offender’s criminal history rather than the mode of evidence used to support a conviction.

Importantly, Nir and Griffiths (2018) suggest that the increased evidential weight of DNA cases may have an undue influence on a judge post-conviction. Arguably without the collection of DNA, and subsequent matching to a profile already in the DNA database, a suspect may fail to be identified at all.

Deterrent effect

The final way that DNA is expected to impact on downstream demands is in terms of a possible deterrent effect on future crimes (Struyf et al., 2019). This works in 2 ways – there may be a specific deterrent effect on individual offenders (that is, previously convicted individual offenders will know they are on a DNA database making them easier to detect, therefore they might be less likely to commit further offences), as well as a broader deterrent effect on a societal level impacting on reduced crime rates rather than just recidivism rates.

The direct impacts of DNA on deterrence are, however, very difficult to unpack from its probative effects. While DNA databases might reduce re-offending as offenders fear easier detection, if they re-offend, they are more likely to be identified, arrested, and convicted by being on the database. There are several quasi-experimental studies in this area that measure impacts on deterrence, using longitudinal data, these have been reviewed in the key studies section. Struyf et al. (2019) provide a recent review of the material on deterrence and conclude that there are likely to be impacts of DNA databases on both recidivism and crime rates. They also caution that research needs to explore the long-term changes over time, while controlling for any other changes over this time period and acknowledge that the impacts are likely to differ by crime type.

Question 10

Annex D: Tables of Literature Reviews, key studies and secondary studies of body-worn camera evidence

Accepted Answer

Table D.1: Summary of some key BWC literature reviews in the area

Authors and review date	Type of review (method and focus)	Number of identified eligible studies	Main conclusions
Cubitt et al. (2017)	Systematic literature review; efficacy of BWC.	11	- Reduce certain crimes. - Reduce complaints against police officers. - Reduce court costs. - Lack of methodologically strong research. - Lack of peer reviewed published material.
Maskaly et al. (2017)	Narrative review; police and public perceptions and behaviour.	21	- Police view BWC positively. - Positive impacts on police-community relations. - More arrests and downstream demands. - More as well as fewer assaults on police. - Strong methodologies (that is, RCT) in emerging research.
Lum et al. (2019)	Narrative review; police and citizen behaviours and attitudes; investigative and organisational impacts.	70	- Findings not always consistent and across literature very wide range of outcomes. - Generally, less use of force and officer complaints. - No clear pattern on arrests – some show increase, some decrease, and some no changes. - Even when impacts seen changes are very modest. - Weakness in studies as only tend to look at the presence of BWC not activation.
Gaub and White (2020)	Review of existing literature reviews.	5 review papers	- The early BWC studies showed clearer BWC benefits but suffered from some methodological weaknesses. - Early studies showed less use of force and fewer police complaints. - Problems with treatment contamination even in better research designs such as RCT. - Mixed findings may relate to differences in force implementation of BWC, in unit of analysis (officers versus shift level), and wide range of outcome variables (not always easy to compare).
Lum et al. (2020)	Systematic Campbell Review; meta-analysis; police and citizen behaviours and attitudes.	30	- Select RCT (20 studies) and quasi-experimental (with adequate matched control group) (10 studies) designs. - 13/30 studies reviewed include some data on police activities (including arrests) but do not review any further downstream demands. - 3.9% reduction in arrests for BWC officers (not statistically significant).
Petersen and Lu (2023)	Meta-analysis; downstream demand impacts.	12	- Some analysis of design but not as stringent on quality as Lum et al. (2020). - Mix of RCT and quasi-experimental (QE) including unadjusted QE. - No significant aggregate crime effect on downstream demands. - For domestic violence, there were significant and large effect sizes on all outcome variables. - DV 3-4 as many convictions (guilty pleas and trial verdicts combined) but designs tended to be QE. - QE studies had inflated effects compared to RCTs, and therefore overestimate downstream demands.

Table D.2: Secondary studies and summaries for BWC

Author(s), date and location of the study	Notes
Goodall (2007) Plymouth UK	Broader guidance document produced by the Home Office and Association of Chief Police Officers (ACPO). There is the inclusion of an early QE BWC pilot in Plymouth in 2005. The report presents very little detail on the design, data and analysis. Lacks methodological rigour and results not generalisable.
ODS Consulting (2011) Scotland, UK	Consultant report for 2 small-scale QE pilot evaluations in Scotland. QE with unadjusted controls, the report gives little detail of design and analysis on the report. Including only simple pre- and post-intervention statistics without considering other confounding variables. Lacks methodological rigour and results not generalisable.
Ariel et al. (2014) Rialto, CA, US	12-month RCT (randomly assigning shifts to either test or control conditions rather than officers) measuring police use of force and complaints only. The use of force was twice as likely in control conditions. There were significantly fewer complaints made about BWC cases.
Ellis et al. (2015) Isle of Wight, UK	A QE (unadjusted) 12-month pilot which compares crime data from the year before implementation to the year after. They report reductions in some crimes post-intervention (8.8% reduction in ASB and assaults), but the study is methodologically flawed.
Ariel (2016) Denver, Colorado, US	A QE 6-month pilot comparing one district where BWC was deployed to other control districts. Focuses on wider outcomes but includes some data on police activities such as arrests made comparing those with and without BWCs. They found no deterrence effect of BWC with no reductions in crime in the test area. BWC cases had a lower likelihood of arrest compared to non-BWC suggesting some de-escalation in BWC cases.
Wallace et al. (2018) Spokane, Washington, US	RCT trial concerned with the impacts of BWC on police activities (in particular ‘de-policing’ effects) rather than CJS outcomes and demands. The study does, however, report data on arrest numbers and they found some significant effects.
Owens and Finn (2018) London, UK	This article is based on the data discussed in Grossmith et al. (2015) College of Policing report but in this article they only present limited detail and focus mostly self-report data rather than CJS outcomes.
Braga et al. (2018) Las Vegas, CA, US	Large-scale RCT pilot focusing on police activity data. BWC officers used less force and received fewer complaints. They issued more citations and made more arrests.
Braga et al. (2019) Boston, Massachusetts, US	One-year pilot involved both a RCT and a matched controls QE aspect (the latter to test for any spill-over effects into other districts). They similarly found that BWC cases received fewer officer complaints and less use of force. But they did not find any significant differences in terms of police activities including the number of arrests made.
Huff et al. (2020) Phoenix, Arizona, US	Building upon earlier Maryvale pilot (included as a key study), this larger-scale RCT study expands to the other 6 precincts in Phoenix, over 18 months. Less focus on further downstream impacts in this study but inclusion of relevant data on arrest rates and officer complaints. Violent offences where BWCs were present were significantly more likely to lead to an arrest (34% versus 27%), and less likely to lead to a complaint. They caution that while the findings are significant the effects are small.

Table D.3: BWC – Key studies

Authors and dates	Sample	Methods
Katz et al. (2014) Morrow et al. (2016)	Maryvale, Phoenix, Arizona, US 56 cameras in one precinct of the Phoenix Police Department from April 2013 to July 2014.	QE design. Pre- and post-intervention comparisons as well as comparing area with, and area without, BWC. Small-scale implementation of 56 cameras over 15 months. Morrow et al. paper focused on the data on Intimate Partner Violence (IPV) only.

Relevance and quality - this study has a range of good CJS outcome measures – aggregate crime and DV arrests, complaints against officers, and DV case progression/outcome variables (for example, charges, prosecutions, guilty pleas, convictions, and time to case disposition).
As an unadjusted QE study, this is weaker in design compared to QE work with better matched control groups, and particularly RCT work.

Main findings - BWC officers made significantly more arrests for aggregate crimes. The BWC group saw a 17% increase over time compared to 9% in the control group, this translated to a 42.6% higher daily arrest rate per officer per day (on average 0.04 more arrests per day). For IPV, while both groups post-intervention were less likely to make an arrest compared to pre-intervention, BWC officers were still significantly more likely to make an arrest than non-BWC officers (41% versus 34%). BWC officer IPV cases were significantly more likely to be charged (37.7% versus 26%), more likely to lead to a guilty plea (4.4% versus 3.1%), and to receive a conviction in court (4.4% versus 0.9%). Post-intervention case disposition times were significantly shorter than pre-intervention due to improvements in the court, but the improvements were less marked in the BWC group. BWC officers also received significantly fewer complaints against them (non-BWC had a 2.55 times higher chance of receiving a complaint).

Implications - increased demands due to more suspects being arrested and charged, and more convictions with BWCs in operation. This applies to both aggregate crime and IPV incidents. There may be some reduction in demand due to the higher number of guilty pleas saving time and money in the courtroom, and lower level of police complaints. As time to disposition was longer for BWC cases (although both improved over time) there is a suggestion that BWC evidence may be causing some delays. The magnitude is small.

Authors and dates	Sample	Methods
Owens et al. (2014)	Essex, UK – 70 officers randomly assigned to condition, compared to control group of non-BWC officers. Incident and officer level of analysis.	RCT. Crime and CJS outcome data, officer surveys and interviews. Measured DV incidents over 4-month trial period.

Relevance and quality - quality of data – some problems meant not enough data to track further downstream effects beyond charge. Low usage of BWC by officers assigned to test condition. Small scale and ended up with very limited CJS outcome data.

Main findings - there were no significant differences in terms of the rate of sanction detection and arrests. Significantly more criminal charges were made in cases from the test condition (in 81% of cases versus 72% in the control group). Although they intended to track cases further in CJS, low usage meant the numbers were too low to track any further outcomes.

Implications - show no increase in demand in terms of arrests (that is, suspects entering the system), but some increased demands as more suspects were charged. Further downstream demands were untested.

Authors and dates	Sample	Methods
Grossmith et al. (2015)	Metropolitan police (10 boroughs) London, UK – 814 officers assigned to wear BWC; 649 wore them during course of trial (1,246 control) and 500 cameras to emergency response teams.	Cluster RCT (across different boroughs). Violent incidents over 12-month trial. Uses police activity data (stop and search), official police complaint data, and crime data. Also included officer surveys and interviews, victim self-report and community surveys.

Relevance and quality - clustering allows testing of variation between boroughs and accounting for this in the multivariate analysis so is an advantage over simple RCTs. Use a good range of measures of police activity and some CJS outcomes, but cases are not tracked beyond charges made. Larger scale than previous pilot in Essex (above).

Main findings - some reductions in police complaints for BWC officers in 6/10 boroughs; but in aggregate, these were not statistically significant. BWC officers did, however, receive significantly fewer police-public interaction complaints and had significantly lower average complaints per officer than control officers. Non-BWC officers were 2.55 times more likely to receive complaints for excessive use of force.
There were no significant differences in arrest rates; 5/10 boroughs had lower average charge rates for BWC officers, this was statistically significant in 2 boroughs.

Implications - in this methodologically more sophisticated study, which accounts for variation between areas, there were very little changes in demand in terms of arrests and charges. There is some evidence that BWC officers are less likely to receive a complaint, showing very minimal demand reductions as these are rare events.

Authors and dates	Sample	Methods
Yokum et al. (2017)	Washington, District of Columbia, US. Officer level analysis.	US government funded block RCT – 11 months of data 2015 to 2016 with approx. Half of officers randomly assigned to test (1,189) and half to control (1,035) (total 2,224 officers). Use crime data, self-reported police activities, official complaint data, and limited judicial outcomes data.

Relevance and quality - good design with block randomisation rather than simple RCT. Large scale over almost a year, with careful testing of covariates, such as any confounding effects of the non-compliance of officers. Wider range of CJS outcomes beyond initial charge. Use of time series analysis to explore effects over time. Track case outcomes to March 2017.

Main findings - the multivariate analysis found very minimal and non-significant treatment effects on all police activity and CJS outcome variables. So, there were no significant differences between BWC and non-BWC officers in terms of use of force, citizen complaints, arrests for disorderly conduct, prosecution, guilty or not guilty verdicts (including guilty pleas), and cases dismissed. They checked to see if effects of BWC changed over time and again found no significant differences.

Implications - no increase or decrease in demand associated with BWC implementation. They conclude that we need to rethink the widespread implementation and reliance on BWC given the lack of evidence from those most methodologically sound studies.

Authors and dates	Sample	Methods
White et al. (2018b)	Tempe, Arizona, US – all patrol response officers randomly assigned to test (n=101) and control (n=99) conditions. Officer level.	Low-level volume crime focus. Data on police activities, and case outcome data. Also included officer surveys, citizen, and stakeholder interviews. Aggregate crime data presented in the first article, and then this is broken down into areas of specific interest in later papers. All the following papers (Huff et al., White et al., Todak et al.) use the same 6-month RCT.

Relevance and quality - measure police activities, time to case disposition data and CJS outcomes. Unlike previous work, these studies focus on only low-level/misdemeanour crime. Measure whether BWC is present in the case – not whether it is activated.
This first paper in the series focused on aggregate volume crime but with more of a focus on police and citizen behaviour and attitudes with limited outcome data.

Main findings - there was a 6% increase in guilty outcomes and an 8% decrease in the time to disposition post BWC implementation, but neither of these effects is statistically significant.

Implications - unlike other studies, the focus is on misdemeanour crime, where increases and decreases in demand may have more downstream impacts given the higher volume of these offences.

Authors and dates	Sample	Methods
Huff et al. (2021)	Tempe, Arizona, US. Same sample and method as White et al. (2018b) above.

Relevance and quality - focused on traffic violations looking at whether BWC has impacted on case dismissal or guilty plea, as well as any possible interactions with suspect race. Logistic regression analysis.

Main findings - counter to expectations, BWCs slightly decreased the likelihood of a suspect accepting a guilty plea. Defendants were significantly less likely to plead guilty to a traffic violation when BWC involved (in White defendants 35.6% versus 36.7%; in Black defendants 33.3% versus 36.7%; no significant difference for Hispanic defendants). There were no significant effects on case dismissal.

Implications - small change in demand in that guilty pleas were less likely for BWC cases – this could lead to a small increase in demand in that cases may be more likely to proceed to trial, or a small decrease in demand if this impacts on less overall convictions. But the effects are so minimal that any changes in demand are likely to be very small.

Authors and dates	Sample	Methods
White et al. (2021)	Tempe, Arizona, US. Same sample and method as White et al. (2018b) above.

Relevance and quality - focused on drug and alcohol offences analysing over 7,000 misdemeanour cases during RCT.

Main findings - logistic regression analysis showed no significant effect of BWC on guilty pleas. There was a general 16% reduction in time to adjudication, with cases being processed faster overall (both BWC and non-BWC). Binomial regression found that the presence of BWC in cases led to significantly reduced times to disposition. The average time to disposition in BWC cases was 4% less than in non-BWC cases, resulting in an average saving of 5.5 days. But this faster processing is not the result of more guilty plea verdicts.

Implications - for drug and alcohol misdemeanours, BWC cases took less time to disposition, leading to some reduction in demands.

Authors and dates	Sample	Methods
Todak et al. (2022)	Tempe, Arizona, US. Same sample and method as White et al. (2018b) above.

Relevance and quality - looks at all misdemeanour crimes over 50,000. Uses DID and ARIMA time series analysis looking for changes due to BWC over time. Outcomes are changes in case flow, case processing, case outcomes and time to disposition.

Main findings - aggregate analysis found that BWC cases were 10.3% less likely to result in a conviction, but that they were processed on average 4% faster than non-BWC cases. The effects were not consistent across all crime types. Further analysis revealed that the lower likelihood of conviction was significant in both offences against the person (14% less likely) and traffic offences (12% less likely). While the faster processing of cases was significant in both traffic and drug/alcohol offences (which both witnessed a 6% reduction).

Implications - results, and therefore any changes in demand, are not consistent across crime type. Significant, but small, reductions in demand as BWC led to fewer convictions and faster case processing. The exact nature of the demand reduction is difficult to interpret as they did not differentiate between convictions due to a guilty plea and those obtained through a guilty verdict at trial.

Authors and dates	Sample	Methods
Clare et al. (2021)	Western Australia – 498 officers over 187 days (86 test and 101 control).	Six-month RCT pilot with a range of police and citizen behaviours as well as range of good CJS measures. Unit of analysis is shifts rather than officers, so they could additionally test changes within the same officers’ behaviour and compare high versus low BWC engagement officers. Use aggregate crime data (not split into crime types).

Relevance and quality - simple RCT with random assignment of shifts to control or test condition. Some good outcome measures using crime data, police activity data, official complaints data and court records. Compared to different areas (metropolitan versus regional) to explore context differences. They only looked at case outcomes 4 months after the trial, so outcome data was limited as most cases had not been resolved.

Main findings - charge rates were significantly more likely in the test condition (139 charges per 1,000 compared to 125.5). The biggest differences came from high BWC engagement officers (averaged 64.3 charges per 1,000 on control days versus 104.5 charges on test days).
There were no significant differences in either the number of guilty pleas from test to control (65% versus 64%), or convictions (75% for both). There is some evidence that guilty pleas, while not more likely, happened a little earlier in the case in the test condition.
Complaints were significantly more likely (although still very rare) in the test condition (0.7 per 1000 incidents versus 0.4). But there were no significant differences in assaults against the police (0.8 per 1000 cases versus 0.7).

Implications - increased demands in terms of more sanctions, including charges resulting in more cases entering the CJS, but there was little evidence about any further downstream demands. This is difficult to ascertain here as the follow-up time of 4 months after the pilot was too short to properly monitor case outcomes.

Authors and dates	Sample	Methods
Petersen et al. (2021)	Miami, FL, US – Spatial temporal units – 22 treatment versus 17 control (103 officers versus 97); 1,452 charged offences in the test group and 1,153 in the control group.	Cluster RCT – as looking at 22 police units in treatment compared 17 control units. Range of crime types (DV, assaults on officers, and drug/alcohol charges) Twelve months follow-up data on cases through State Attorney Office files.

Relevance and quality - good design in terms of cluster rather than simple RCT to avoid treatment contamination effects. Looked at charged offences and their prosecutorial and court-related outcomes (declined prosecutions, adjudication withheld and convictions). They looked at aggregate crime, and DV, volume drug/alcohol offences, and crimes against police officers more specifically.

Main findings - simple bivariate analysis found that BWCs significantly increased the number of successful prosecutions for crimes against the police (44.1% in the test condition versus 29.8% in the control), and significantly increased the number of DV convictions (14.6% in the test condition versus 2.1% in the control). Multi-level regression analysis (accounting for the cluster variation) only found significant increases in prosecutions for crimes against the police where the odds of conviction (or adjudication withheld) were 93% higher in charged cases involving BWC. The DV effect was attributable to between cluster variation, with the initial significant finding in part caused by a small sample size of DV convictions.

Implications - the results suggest little change in demand overall. There are some increases in demand in terms of more successful prosecutions of assaults against police officers (including both convictions and OOCRs) but these are relatively rare events.

Authors and dates	Sample	Methods
Pimley et al. (2022)	Pacific Northwest, US – 13,000 incidents from one small police department.	Longitudinal data from January 2004 to December 2018, testing the impact of intervention (deployment of BWC) starting in April 2013. Uses ITSA. Looks at not only the immediate effects of the intervention but delayed impacts.

Relevance and quality - although not experimental, this study uses pre- and post-intervention longitudinal data and sophisticated time series analysis to examine changes over time on range of outcomes (conviction rates, dismissals, and times to disposition). Some consideration of variation by crime type but mostly aggregate crime. Did not distinguish between guilty pleas and guilty trial verdicts.

Main findings - in terms of volume crime DUI and DWSL traffic offences were dealt with significantly faster after BWC implementation. But DWSL saw significant declines in guilty verdicts.
For aggregate crime there was a (delayed) increase in the referral of incidents for prosecution (which were in decline pre-intervention) – with 0.35 more prosecutions per month post-intervention.
Case dismissals had been declining but saw a significant spike of increase at the onset of the BWC implementation, but this was not sustained as time went on.
For guilty outcomes, these had been decreasing pre and during the immediate intervention period, then increased slightly post-intervention (0.37 more convictions per month).
In terms of time to disposition (and they measured this in several different ways depending on the case outcome), before the adoption of BWCs, disposition times had been gradually increasing. At the initial implementation, this did not significantly change, but then cases were disposed of significantly faster (for example, a 0.75 reduction of days per case per month until final disposition). There is some variation here dependent on the type of outcome; so, for example, case closure by conviction saw some initial large significant delays when BWCs were first adopted (31 days increase on average) but these delays were short lived in contrast to cases rejected by the prosecution, although the processing time had been increasing pre-intervention, once BWC were first introduced this fell considerably (by on average 24 days), and then stabilised with a saving of 0.62 days per case.

Implications - some increases in demand early in the CJS as significantly more prosecutions. But the pattern of changes in demand is very complex.
There was some immediate demand reduction in terms of more rejections of cases for prosecution in the short term, and then some delayed increased demand in terms of more convictions.
Demand reductions are also seen in terms of faster case processing times, although this varies in terms of the impact on specific types of disposition. It seems that convictions saw some considerable short-term delays, perhaps as the CJS was adjusting to the changes. While rejection of cases initially happened much faster.

Question 11

Review of secondary studies of BWC evidence and the relationship with outcomes

Accepted Answer

Case building effects: arrests

Police can be viewed as gatekeepers to the CJS as their actions dictate the influx of suspects (Hartman and Belknap, 2003). Any changes in terms of how many detections are made and how many sanctions are given, are likely to impact on the downstream CJS if there are more suspects being arrested and charged. Once a case is opened, resources will be needed to investigate and process that case, even if no charges are made and the case is later NFA. Speculation about how the use of BWCs might impact the number of sanctions the police issue is not very clear (Lum et al., 2020). It may be that the implementation of BWCs would influence police officers to act more legalistically and exercise less discretion. As they are likely to perceive that their actions are being monitored, they may be inclined to increase the use of sanctions (including arrests) to establish that they are performing their job well and in accordance with what is legally expected.

They may also feel more confident about making an arrest as the recording of evidence may be seen as substantiating their actions. On the other hand, if BWCs are felt to have a civilising effect on the police and the public, it may be that certain situations can be de-escalated, avoiding the need for an arrest to be made. They may also feel more scrutinised and therefore less likely to make an arrest if they think their action may be criticised later as being heavy-handed or are concerned the recorded evidence might not support their decision. This obviously applies to some incidents and crime types more than others – that is, those where the police are present as the incident is taking place – for example, public order offences. The officer’s behaviour is also, however, likely to vary dependent upon the level of discretion they are given in terms of the activation of their BWC.

In some evaluated trials, officers were merely mandated to carry a BWC and given considerable discretion as to when it was used. In contrast, those in other studies were given strict instructions that the BWC must be used in specified circumstances, such as all violent or domestic violence incidents, in all contact with the public.

There is a growing body of evidence in terms of how BWCs impact on police activities, including the number of arrests made. The results are, however, rather inconsistent across studies. Several North American studies have found that officers carrying BWCs made significantly more arrests – in the US (Braga et al., 2018; Ready and Young, 2015) and Canada (Toronto Police Service, 2016), made significantly fewer arrests (Ariel, 2016b; Headley et al., 2017; McClure et al., 2017), or made neither more nor fewer arrests than non-BWC officers (Hedberg et al., 2016; Wallace et al., 2018; White et al., 2018a). With a lack of clear consensus, it is therefore difficult to ascertain how BWCs may change police and citizen behaviour, and how these impact on arrest volumes and further downstream demands.

Rowe et al. (2018) suggest that these differences in arrest volumes between studies may be related to the degree of discretion officers are given about their use of BWCs. Huff et al. (2020) (Phoenix, Arizona, US) looked specifically at whether the BWC was activated and found that when activated, they were significantly more likely to lead to an arrest (35% versus 27%). This gives some indication that more arrests may be associated with BWC activation. It is, however, difficult to draw conclusions from studies with such varying methodologies, crime types, and implementation strategies (Lum et al., 2020; Lum et al., 2019).

Case building impact: police investigations

Regardless of whether more suspects enter the CJS via greater numbers of arrests, it has also been argued that BWCs can aid police in their execution of their investigations, and lead to the collection of higher quality evidence. Research has provided some support that BWCs may enhance police investigations. For example, it may improve the accuracy of memory, particularly in incidents that are complex and/or induce stress (Blaskovits and Bennell, 2020). One US study also found that the use of BWCs improved police productivity, efficacy, and evidence collection (Cayli et al., 2018).

Pelfrey and Keener (2016) (US) found that it aided officers in the collection of information for case file preparation. The footage obtained from BWCs can be useful in domestic violence investigations to identify primary aggressors, as well as reducing the pressure on victims to give testimony (Vakhitova et al., 2022). As discussed above in the case of DNA traces, if the evidential strength of the case is increased, this should be demonstrated by changes in ‘hard’ CJS outcomes, such as faster case progression, more charges, prosecutions, and convictions (Redlich et al., 2016). To date, however, there are only a limited number of studies that robustly test the impacts of BWCs on these outcomes.

Any increase in the number of crimes charged and/or individuals prosecuted will inevitably lead to increases in demand in terms of more suspects flowing through the CJS. There is some evidence from early QE studies in the UK that BWCs led to significantly higher charge rates – see Goodall (2007) and (Ellis et al., 2015) in England and ODS Consulting (2011) in Scotland. For example, Goodall (2007) found that BWC conditions resulted in a significantly higher conversion of incidents to recorded crimes, increased sanction detections, increased the use of penalty notices, and increased the charge/summons rate for violent crimes.

Lum et al. (2019, 2020) caution against relying too heavily on QE studies with no matched control comparisons, suggesting that they increase the risk of bias, and are likely to over-inflate the significance and size of the findings. In a meta-analysis by Petersen and Lu (2023), 6 out of the 8 studies that presented data on charge rates reported an increase but, overall, the effects were not significant for aggregated crime. Meta-analysis of DV charge rates did reveal a significant and large effect for BWC use, but these studies tended to be QE rather than the higher quality RCTs (so there is some caution again about potential sources of bias and over-inflated effect sizes). Only one RCT focused on DA and this study only found a move towards charging and away from other sanctions (cautions) (Owens et al., 2014).

Obtaining better-quality evidence from BWC footage was expected to lead to more guilty plea verdicts. As the strength of the evidence increases, it may be more likely for suspects to offer a guilty plea, given the weight of the evidence against them. There would also be the expectation that a suspect would plead guilty earlier in the case proceedings. Greater evidential strength could also encourage the victim to continue their participation (as witnessed in the use of DNA evidence).

Many studies, however, only look at the rate of convictions without distinguishing between guilty pleas and the finding of guilt following a full trial. ODS Consulting (2011) found some evidence of this in Scotland, claiming that BWC cases were 80% less likely to go to trial, but the sample size is very small. Petersen and Lu (2023) again failed to find any significant effects for the use of BWC on guilty pleas for all crimes combined, finding a large statistically significant increase in guilty pleas for DV cases only. Turning to overall convictions, Petersen and Lu (2023) found the same pattern of results – combining all crime types together failed to find any statistically significant increases but, taking the DV data in isolation, there were 3 to 4 times as many convictions in BWC cases.

Clearly, BWCs have a more pronounced impact on certain crimes, but little work has been done so far that explores non-DV crime types in sufficient detail. Petersen and Lu (2023) were therefore unable in their meta-analysis to explore anything other than aggregate crimes versus DV offences. Evidence does suggest that there will be some increased downstream demands associated with BWCs in DV incidents, but again, the reliance on weaker methodologies in these studies warrants some caution.

Case building: timeliness

The impact on downstream demands is potentially quite complex, as an increase in guilty verdicts may lead to an increased demand in terms of more prison places or probation supervision, but a higher proportion of guilty pleas can also potentially reduce demand if cases can be resolved without a lengthy and costly court process. Few studies have examined the impact of BWCs on case timelines – that is, the time from the occurrence of a crime to its final disposition. Petersen and Lu (2023) identified only 4 studies that measured processing time – of which 2 studies showed BWC cases were processed faster, and 2 were associated with slower progression.

Overall, this review found a statistically significant increase in processing times. This increase in time might be expected immediately following the force’s implementation of BWCs when the system is adjusting to recent changes. There are also reports of delays due to the demands of reviewing and processing BWC material (Lum et al., 2019; Maskaly et al., 2017). There is some anecdotal evidence, however, that some of these issues can be overcome by establishing and maintaining a supporting infrastructure (Spencer and Cheshire, 2018).

ACPO	Association of Chief Police Officers
ARIMA	Auto-Regressive Integrated Moving Average
BWC	body-worn cameras
CJS	criminal justice system
CPS	Crown Prosecution Service
DID	Difference-in-Difference
DNA	deoxyribonucleic acid
DUI	driving under the influence
DV/DA	domestic violence/ domestic abuse
DWSL	driving with a suspended licence
IPV	intimate partner violence
ITSA	Interrupted Time Series Analysis
NCJRS	National Criminal Justice Reference Service
NDNAD	National DNA Database
NFA	no further action/actioned
NPCC	National Police Chiefs’ Council
OOCR	out-of-court resolutions
QE	quasi-experimental
RCT	randomised control trial
REA	Rapid Evidence Assessment
SAMFE	sexual assault medical and forensic examination
SANE	Sexual Assault Nurse Examiners
SARC	Sexual Assault Referral Centre
SART	Sexual Assault Referral Team
UK	United Kingdom
US	United States of America

Cookies on GOV.UK

Applies to England and Wales

Authors

School of Criminology, University of Leicester

Acknowledgements

List of Abbreviations