Research and analysis

The predictors of police demand: A rapid evidence review

Published 16 July 2026

Applies to England and Wales

Acknowledgements

We would like to thank Ruby Forshaw and Jo Perkins for initiating the review and constructing the inclusion criteria, and Abbie Harrison for helping scope the literature and rate the papers. We would also like to thank our colleagues Luke Edwards, Paul Trenell, Charlotte Bryant, Alison Spence, Aniel Anand and Samuel Atkin for reviewing the work and providing comments.

Finally, we would like to thank the external reviewers, Professor Kate Bowers and Professor Tom Kirchmaier for their feedback on the paper.

Executive summary

This rapid evidence review explores previous research which has investigated the predictors of police demand. This review can be used to enhance insight and inform key decision making within crime and policing in the UK. The predictors of crimes and non-crime incidents are particularly important for explaining or predicting changes in police demand.

The review’s terms of reference include:

• searching for previous papers that investigate the predictors of UK crimes and non-crime incident policing demands
• providing a methodology which is in-depth, rigorous, and critical
• identifying police predictor and demand relationships
• understanding how predictors influence crime patterns

Demand on the police is recognised as the events which require police action and intervention. While policing demand is complex, it is important to understand both crimes and non-crime incident trends, particularly as non-crime incident demands take up a large proportion of police resource allocation (Hadjipavlou et al., 2018). While previous research has explored predictors of crime and non-crime incident demands, most of the research has only explored specific predictors for specific demands. This review has examined an extensive amount of research to identify predictors of demand, gaining a wide understanding to support future policing decisions. The purpose of this review was to establish the relationships between predictors and demands, not causation, as this cannot be established from data alone.

Methodology

The scope of the review was flexible enough to capture as many predictive factors as possible but also had a specific inclusion criteria, and method for critically appraising papers. The selection of papers included the following steps:

Inclusion criteria

1. Using Google Scholar, the search terms ‘Determinants’, ‘Predicting’, ‘Relationship’ and ‘Impact of’ were used in front of Home Office Offence Groups (for example, sexual offences, theft, violent crimes) and National Standard for Incident Recording counting rules (NSIR) (for example, Antisocial behaviour, Road traffic accidents) to produce relevant research outputs. 928 papers were identified from these searches.
2. Study characteristics used by Tarling and Dennis (2016) were adopted to select papers based on whether they were UK based, included a large sample (quantitative studies), conducted analysis of contemporary data, were conducted between 2011-2021, and used a regression analysis. Papers that met all factors were rated 3-stars (26 papers). Papers that were UK based but missed one of the factors were rated 2-stars (36 papers). A total of 62 papers were included in the review.
3. Each predictor-demand association identified from the regression analyses was recorded separately from the 62 papers. For each association, study information was extracted from the paper including author, date, scale of data (for example, Lower Super Output Area), predictor-demand associations, level of significance, effect sizes (if reported) and a short summary of findings.

Critical appraisal

There were 1,114 analyses conducted on predictor-demand associations from the 62 papers. Predictors were ranked based on the quantity of evidence. Predictors were ranked as good evidence with 7 or more predictor-demand association analyses (total of 18 predictors), fair evidence with 3 to 6 analyses (total of 13 predictors) and low evidence with 1 to 2 analyses (total of 18 predictors).

Key findings

There were 49 predictors of police demand which were associated with 27 crime and non-crime incident demands. Predictors were grouped into larger categories, which included socio-economic factors, population, traffic environment, demographic factors, geographic location, events, individual factors, and COVID-19.

Socio-economic factors: Higher rates of deprivation, education and income inequality were associated with increases in burglary, robbery, and theft, and increases in unemployment were associated with increases in violent crime. Sociodemographic factors (family structure) and socio-economic status (SES) were associated with teenage drug use and violent crimes (for example, weapon carrying/use).

Population: Generally, changes in population (total, churn, and density) were associated with increases in crime demand. Increases in total number of prisoners (prison population) in a police force area (PFA) was associated with decreases in crime demand.

Traffic environment: Increases in volume of traffic, total road length, number of junctions and road end points, and air pollution from traffic, were associated with increases in slight, serious, and fatal RTAs.

Demographic factors: Males, deviant youth behaviour and weak social bonds were associated with drug use, antisocial behaviour, and weapon carrying/use. While age was split into categories of 10 to 15 years, 15 to 24 years, 16 to 29 years, and ‘all ages’, the direction of association was mixed for crime types such as burglary, robbery, theft, fraud, vehicle crimes, violent crimes and weapon carrying/use.

Geographic factors: Vacant properties and public amenities were associated with increases in theft, and higher sanctioned detection rates (defined as the proportion of recorded crimes that have been assigned an outcome by the police) in an area were associated with decreases in many crime types. Other housing variables (average house price and housing mix) showed mixed associations with crime demand. Increases in urban land and severe weather were associated with increases in RTAs.

Events: Political events and terror attacks were associated with increases in hate crimes. Major sporting events were associated with increases in crimes such as criminal damage, theft, and violence against the person.

Individual factors: Alcohol dependency, mental health disorders, drug use and past victimisation were associated with increases in violent behaviours (for example, gang membership and weapon carrying/use).

COVID-19: Lockdowns due to COVID-19 were associated with decreases in crimes such as burglary and robbery but increases in cybercrime and domestic abuse. Lockdowns also caused an increase in previous predictors such as alcohol, poverty, and unemployment. Increases in unemployment were associated with increases in other crimes such as antisocial behaviour and drug offences.

Discussion/conclusion

Overall, the review explored many predictors which were significantly associated with police demand. While some predictors are generally associated with increases in crime and non-crime incident demands, others had mixed associations.

The current review explored a diverse range of research which highlighted predictors that either had limited or outdated literature to support them. While crime demands are generally well researched, some incidents such as missing persons, domestic violence, mental health, and RTAs need further research in future.

Due to COVID-19 lockdowns, crime and non-crime incident rates have been researched recently; at the time of this review, the long-term impacts of COVID-19 on police demand are unknown. Crime trends should continue to be monitored and explored further to understand the impacts of COVID-19 in relation to police resource allocation.

This review can be used to support and inform decisions in crime and policing across England and Wales. The predictors of crimes and non-crime incidents can be used as independent variables to explain or predict changes in demand.

1. Introduction

1.1 Background

In its simplest form, demand on the police can be recognised as events which require police action and intervention (Hadjipavlou et al., 2018). The College of Policing (2015) define police demand through 2 factors, public and protective demand. Public demand fits into the traditional view of the police as responding to service calls and incidents by request of the public and can include both crimes, and non-crime incidents. Protective demand can describe the actions taken to safeguard the public through various means, such as child protection and vulnerable adult referrals.

Policing demand is complex, challenging, and diverse. Observing and understanding changing crime trends increases our understanding of how crime is experienced today. Police recorded crime (PRC) and Crime Survey for England and Wales (CSEW), previously known as the British Crime Survey, are the 2 main data sources for reporting crime (ONS, 2021).

Non-crime incidents, which include antisocial behaviour (ASB), transport (such as road traffic accidents), and public safety and welfare (such as mental health incidents and missing persons), are estimated to take up 75% of the police’s total incident volume (Hadjipavlou et al., 2018). Non-crime incidents also account for 83% of Command-and-Control calls (College of Policing, 2015). While non-crime incidents are taking up more resource than they did a decade ago, this category covers a wide range of incidents which require a detailed breakdown to make sense of both its impact and how to tackle it (Hadjipavlou et al., 2018).

The types and numbers of crimes reported have changed over the years. There has been a rise in more harmful crimes involving violence and sexual offences (Hadjipavlou et al., 2018), with these crimes requiring in-depth investigations which consume police time and resource. Understanding the predictors of police demand can enhance insight and inform decision making in both reducing crime and monitoring police demand.

More recently, crime trends have also been impacted by the coronavirus (COVID-19) global pandemic due to the introduction of national lockdowns for the UK. When comparing to pre-COVID times for the year ending June 2019, there was a 12% increase in total crime, which was driven by a 43% rise in fraud and computer misuse cases for the year ending June 2021 (ONS, 2021). Changes in specific crime trends between April and June 2020 were evident through reductions in thefts and possession of weapons offences, and simultaneous rises in drug offences which reflect higher police activity on quieter streets (ONS, 2020). Still, annual reports ending in both June 2020 and June 2021 found that crimes stabilised at 5.8 million offences (ONS, 2020; 2021).

1.2 Aims for the review

There is an extensive body of literature which has explored predictors of both crimes, and non-crime incident demands. Despite this large body of evidence, previous reviews considered a limited number of predictors of police demand (for example, Jennings, 2013; Levi & Smith, 2021; Mawby, 2016) for specific crimes and non-crime incidents. Some predictors, such as the level of deprivation in geographical areas, have had a vast amount of research evidence to support this being a predictor of police demand (Brennan, 2019; Livingston et al., 2014; Mburu & Helbich, 2016; Tarling & Dennis, 2016; Quick et al., 2018). However, the current review sought to identify a wide number of predictors across the full range of crime and non-crime policing demands.

Conducting an extensive review of research on both crime and non-crime incident demands, and understanding the predictors of these demands, can be used to inform police decision making around demands.

The current review aims to:

• provide an in-depth search of UK crime and non-crime incident literature highlighting predictor and police demand relationships
• understand the variables which contribute to and change crime patterns
• provide an evidence base for use explaining and predicting changes in police demand

2. Methodology

This section describes the review’s method for scoping, selecting, and rating the quality, and quantity, of papers. A further brief discussion on extracting the data and presentation of evidence will be described.

2.1 Rapid evidence reviews

Rapid evidence reviews are used to synthesise evidence which can inform and guide future decision making by presenting data in a succinct and timely manner. These reviews are less exhaustive than a systematic review but far more rigorous than a literature review (GOV.UK, 2017).

For this rapid review, the question around what factors predict police demand was specified a-priori, and sources were found using Google Scholar. A rigorous and critical appraisal methodology was adopted however, results are descriptive and interpreted with caution (Khangura et al., 2012).

2.2 Search processes

Search terms were identified through conducting an initial scoping of literature using the terms ‘Demand’ and ‘Crime’. This was to determine which conjoining terms would produce the most results. Titles of papers that included the terms ‘Demand’ and ‘Crime’ commonly produced the conjoining terms ‘Determinants’, ‘Predicting’, ‘Relationship’, ‘Measuring’, ‘Impact of’ and ‘Modelling’. These were used as search terms, though, ‘Measuring’ and ‘Modelling’ were later removed due to there being few relevant research outputs.

The search terms for crime demands were defined using the highest level HO Offence Groups (for example, sexual offences, theft, violent crimes) under the HO Counting Rules (GOV.UK, 2021). For non-crime incidents, these were identified from the National Standard for Incident Recording counting rules (NSIR) (GOV.UK, 2011) which included road traffic accidents (RTAs), antisocial behaviour and missing person incidents.

Papers were sourced through Google Scholar using the identified search terms. For example, for the crime demand ‘arson’, the paired search terms were “arson and determinants”, “arson and predicting”, “arson and relationship” and “arson and impact of”. A date filter was applied from 2011 to 2021 to locate recent research. Papers published prior to this time frame, identified through systematic or literature reviews, were included in the review if recent research in the identified area was limited. For each pair of search terms, 10 pages of Google Scholar search results, containing 10 search outputs per page, were assessed for relevance. If a paper appeared to be relevant (that is, the item included research on police demands rather than, for example, a commentary piece) from reading the title and abstract, then the paper was recorded and read further to determine the rating.

Additional exploratory searches for papers were conducted using reference lists and forward citations from systematic or literature reviews. Details of papers that were initially considered to be of interest and possibly relevant for the review were recorded for further review. In total, 928 papers were identified through this search process.

2.3 Inclusion criteria: Selecting papers for further review

The procedure for rating papers in the current review was based on the method used by Tarling and Dennis (2016). A rating system was formulated and integrated into the research process to ensure papers were selected consistently.

The rating system characteristics adopted included selecting papers which were:

• UK based
• a large sample size
• conducted using analysis of contemporary data
• conducted and published between 2011 and December 2021
• regression analyses, specifically linking police demands with predictor variables

As studies required a large sample, qualitative studies were excluded due to small sample sizes. Examples of large databases used by included quantitative studies were Census data, Police recorded crime (PRC) data, and Crime Survey for England and Wales (CSEW) data.

As regression analyses’ primary use is to analyse statistical relationships between predictors and outcomes, it was important that studies were selected which conducted regression analyses, rather than using studies containing descriptive statistics. These tests can determine which variables are of relevance, matter the most, and which offer no influence on other variables and can be ignored (Sarstedt & Mooi, 2019). Searches for papers ceased in December 2021. Any research published beyond this date would not be included in this review.

Providing all the above requirements were met, a paper was rated as 3 stars. For 2-star papers, the studies were based in the UK but were missing one of the other requirements for inclusion. If further items were missing, papers were excluded from the review. Out of the 928 papers identified, 62 papers were rated as 3 or 2 stars and would be reviewed further. See table 1.1 for a breakdown of the papers that were included or excluded, and table 1.2 for a breakdown of exclusion reasons. Meta-analyses were excluded because they contain a synthesis of primary research. ‘Measures’ refers to papers which have focused on developing a conceptual tool to measure demand but have not conducted any original analyses.

Table 1.1: Inclusion and exclusion of papers

Papers included and excluded	Count	Per cent
Number of papers meeting the inclusion criteria	62	7%
Number of papers excluded	866	93%
Total number of papers identified	928

Table 1.2: Inclusion and exclusion of papers

Reasons for excluding papers	Count	Per cent
Not UK	334	39%
Beyond scope	171	20%
Irrelevant	160	18%
Evidence Reviews	87	10%
Older studies	41	5%
One-star	30	3%
Duplicates	16	2%
Measures	11	1%
Book	11	1%
Meta-analysis	3	0%
No access	1	0%
Unpublished research	1	0%

2.4 Extracting the data

Once papers had been rated, data was extracted from only those rated as 3 and 2 stars. Each predictor and demand association identified in the paper was recorded separately. To avoid duplicating variables due to different names, some variables were recoded to a larger grouping variable. For example, the variable ‘Residential Instability’ was recoded as ‘Population churn’. This process also ensured that it was clear where predictors had limited research.

Many papers included analyses for multiple predictors and/or demands. For each predictor-demand association, the corresponding study information was extracted:

• author and date
• scale of data, for example, Lower Super Output Area, local authority
• predictor and demand
• level of significance
• effect size, where reported
• a short summary of the findings

A total of 1,114 predictor-demand associations were identified from the 62 papers. The extraction of the predictor-demand associations was conducted independently by a minimum of 2 HO researchers.

2.5 Critical appraisal: Rating the predictor-demand association

Papers reviewed at this stage had previously passed the inclusion criteria and were of sufficient quality. Therefore, beyond this point papers were rated based on the quantity of analyses for each predictor-demand association.

For each predictor-demand association identified in each paper, the number of analyses conducted was recorded along with the significance levels. It was common for papers to report multiple analyses for each predictor-demand association, for example, a study explored the association of income as a predictor of violent crimes or drug offences. A paper which reported predictor variables with either one significant analysis, or 2 analyses (where one was significant and one non-significant) was rated as low. Two significant analyses were rated as fair, and 3 or more significant analyses were rated as good. This gave a clear indication of the number of papers that contained significant analyses to support each predictor-demand association.

A final rating of good, fair, or low evidence for each predictor was then made. Predictors with 7 or more analyses across all 3 groups (for example, 1 good, 2 fair, 4 low) were ranked as good evidence, those with 3 to 6 analyses were rated as fair evidence, and those with 1 or 2 analyses were rated as low evidence for the predictor-demand association.

Overall, there were 18 predictors rated as good evidence, 13 predictors as fair evidence and 18 predictors as low evidence. This totalled 49 different predictors.

3. Analysis and results

3.1 Demands recognised in the literature

For this review, 62 papers passed the inclusion criteria and from this, 49 predictors were found to be associated with 27 crime and non-crime incident demands. Below are the crime and non-crime demands recognised from the literature which were associated with one or more predictors:

3.2 Crime demands

• all crime
• arson
• burglary
• crime rates
• criminal damage
• cybercrime
• dark figure crimes (% of crimes not reported)
• domestic abuse
• drug offences
• fraud
• hate Crimes
• miscellaneous Crimes
• possession of Weapons
• property Crimes
• robbery
• sexual offences
• theft
• vehicle crimes
• violence against the person
• violent crimes

3.3 Non-crime incident demands

• antisocial behaviour
• gang membership
• missing persons
• road traffic accidents

3.4 COVID-19 demands

• alcohol (due to COVID-19)
• poverty (due to COVID-19)
• unemployment (due to COVID-19)

Alcohol, poverty, and unemployment have been explored as predictors of policing demand. For example, higher rates of unemployment were significantly associated with an increase in violent crimes. However, when focusing on COVID-19 as a predictor of policing demand, this was seen to be associated with alcohol, poverty, and unemployment, making these police demands. For example, COVID-19 was significantly associated with increases in unemployment.

3.5 Predictors of police demands

To present the results, the predictors were grouped. The higher-level groupings are socio-economic factors, population, traffic environment, demographic factors, geographic location, events, individual factors, and COVID-19. For example, socio-economic factors include predictors such as deprivation, unemployment, and education.

For ease of reporting, some predictors may also have been combined. For example, the socio-demographic (family structure) predictor also includes findings from the predictor, female lone parent. Therefore, although there were 49 predictors found, there are 41 predictor categories discussed in this section. See Appendix Table A1 for the higher-level group headings, and the 49 predictors that make up the 41 reported predictor categories. Appendix A2 shows the higher-level groupings and individual predictor names with the authors of the associated papers.

In addition, due to the large quantity of evidence found, predictors rated to have good supporting evidence will be presented first, in detail, and predictors rated to have fair or low supporting evidence will be briefly summarised. All results reported were significant at 0.05 level, unless stated otherwise.

Where studies have explored similar variables, results may be conflicting due to differences in granularity of the data, geographical areas, and times. As authors used different variables as a proxy for predictors, the first mention of these variables are in inverted commas (for example, ‘children not living with 2 parents ‘) to indicate that these words or phrases are used by the authors being referenced.

Tables 2 to 9 outline the higher-level grouping, the individual predictors and the predictors associated evidence rating (that is, good, fair, low). Appendix Tables A3 to A10 show summaries of the predictor-demand associations for each higher-level grouping.

3.6 Socio-economic factors

Findings from the research will be presented for predictors listed in Table 2 which all fall within the higher-level grouping of socio-economic factors.

Table 2: Socio-economic predictors that have been identified as affecting police demand, including the strength of evidence for each predictor

Predictor variable	Evidence rating
Deprivation	Good
Unemployment	Good
Education	Good
Income equality	Good
Socio-demographic (family structure)	Fair
Socio-economic status (SES)	Low
Affluent area	Low

3.6.1 Deprivation

The review identified 11 papers which explored deprivation and its relationship with police demands. ‘Impoverished neighbourhoods’ were associated with an increase in several crimes. These included burglaries (Quick et al., 2018), electoral fraud (Carl, 2017), property crimes (Livingston et al., 2014; Tarling & Dennis, 2016), robberies (Quick et al., 2018), violent crimes (Kawalerowicz & Biggs, 2015; Page, 2015; Quick et al., 2018; Tarling & Dennis, 2016), violence against the person (Livingston et al., 2014) and possession of weapons (Brennan, 2019).

‘Deprivation’ was a predictor of theft when measured in closer proximity (within 160 metres (m) and 320m) of street segments. However, at greater distances (within 480m and 640m), this was no longer significant (Mburu & Helbich, 2016). No significant association between vehicle crimes and deprivation has been found (Quick et al., 2018). Brennan (2019) further explored ‘area’s safety levels’ (that is, ‘fairly safe’, ‘fairly unsafe’, and ‘very unsafe’) but found no significant association with any area and possession of weapons.

Increased deprivation was associated with increased risk of RTAs (Haynes et al., 2007; Jones et al., 2008). Quddus (2008) used the ‘number of households without cars’ as a proxy for poverty; findings suggested that higher numbers of households without cars was associated with higher rates of both serious and fatal casualties, but not significant with slight casualties.

3.6.2 Unemployment

Unemployment was found to be a predictor of police demands in 17 papers.

When focusing on ‘rates of unemployment’, 3 studies found that higher rates of unemployment were associated with higher rates of violent crimes (Tarling & Dennis, 2016; Whitworth, 2012; 2013). One study reported that higher rates of unemployment was associated with a decrease in violent crimes (Alsharkas & Campaniello, 2019).

Findings were mixed for burglary, property crime, violence against the person, robbery, sexual offences, theft, and fraud. For burglary, while 4 studies concluded that higher rates of unemployment were associated with higher burglary rates (Bandyopadhyay et al., 2011; Davies & Johnson, 2014; Whitworth, 2012; 2013), 2 studies found associations with lower burglary rates (Gulma et al., 2018; Han et al., 2013).

For property crime, for every 1% increase in the rate of unemployment, Brosnan (2019) found there was an associated 2.2% increase in property crime, whereas Alsharkas and Campaniello (2019) found there was an associated 0.21% decrease in property crime. Tarling and Dennis (2016) reported no significant association.

Higher rates of unemployment were associated with higher rates of violence against the person in low, middle, and high crime rate areas (Bandyopadhyay et al., 2011), but higher rates of unemployment were also associated with a general reduction in violence against the person (Brosnan, 2019) and a reduction in violent related injuries (Matthews et al., 2006). Han et al. (2013) reported no significant associations between unemployment and violence against the person.

For robbery, while Bandyopadhyay et al. (2011) and Whitworth (2012; 2013) found higher rates of unemployment were associated with higher rates of robbery, Han et al. (2013) reported an association with decreased rates of robbery. Wu and Wu (2012) found no significant associations.

Bandyopadhyay et al. (2011) found that higher rates of unemployment were associated with higher rates of sexual offences and theft but not fraud, whereas Han et al., (2013) and Wu and Wu (2012) found higher rates of unemployment to be associated with lower fraud rates, but not significantly associated with sexual offences or theft.

Other specific variables were used to measure unemployment when measuring its effect on drug offences, vehicle crimes, violent crimes, hate crimes, road traffic accidents (RTAs), arson and dark figure crimes. Dark figure crimes are crimes that are unknown and un-reported to the police, for example not reporting violence due to victims being afraid. These unreported crime statistics are collected and produced by the CSEW.

‘Male rates of unemployment’ were associated with an increase in drug offences and other crimes, but no significant association was found for vehicle crimes and violent crimes (Wu & Wu, 2012). Conversely, Whitworth (2012; 2013) found increases in male rates of unemployment was associated with increases in rates of vehicle crime. Page (2015) found that increases in rates of youth unemployment were associated with decreases in adult violence related injuries.

Buil-Gil et al. (2021b) found that as the ‘percentage of people in high and intermediate occupations’ increased, dark figure crimes decreased. However, when including people in low occupations, the results revealed no significant association.

While Williams et al. (2019) found that higher unemployment rates were associated with higher rates of racial or religious aggravated criminal damage, 2 other studies found no significant association (Whitworth, 2012; Wu & Wu, 2012).

Finally, Wang et al. (2009) and Quddus (2008) found a higher employment rate in an area was associated with a higher rate of RTAs, and Andrews (2011) found higher claimant rates of job seekers allowance in an area was associated with higher rates of arson.

3.6.3 Education

In the reviewed papers, education was measured using different levels of qualification attainment. Higher numbers of ‘individuals in an area without any qualifications’ were associated with an increase in dark figure crimes (Buil Gil et al., 2021b) and burglary rates (Gulma et al., 2018), but with a decrease in violence against the person crimes and racially or religiously aggravated harassment (Williams et al., 2019). There was no significant association between individuals without qualifications and racially or religiously aggravated criminal damage (Williams et al., 2019).

An increase in ‘basic education only’ in an area was associated with an increase in burglary, robbery, vehicle crimes (Whitworth, 2013), property crimes and violent crimes (Alsharkas & Campaniello, 2019). However, one study found this was associated with a decrease in violence (Whitworth, 2013). There was no significant association between ‘IQ scores’ and violence (Smith & Wynne-McHardy, 2019b).

An increase in ‘individuals attaining 5 GCSEs A* to C’ in an area were associated with a decrease in criminal damage and violence, though no significant associations were found for burglary, robbery, or vehicle crimes (Whitworth, 2012). One study found the smaller the ‘diversity in educational attainment’, the higher rates of burglary in an area (Gulma et al., 2018).

3.6.4 Income inequality

Studies have used a range of variables to measure income inequality. These include ‘the degree of income inequality’ in an area, and ‘male income inequality’.

For burglary, robbery, and theft, most of the research found that increases in ‘the degree of income inequality in an area’ was associated with higher rates of crime (Bandyopadhyay et al., 2011; Han et al., 2013; Whitworth, 2012; Wu & Wu, 2012). Only one study, which only focused on South Yorkshire and London data, had mixed results (Whitworth, 2013). They found no significant association between income inequality and burglary rates for both areas together, but for London data only it was found that as income inequality increased there was a decrease in robbery rates.

Similarly, most research found that increases in income inequality in an area was associated with an increase in vehicle crime (Wu & Wu, 2012; Whitworth, 2012). Whitworth (2013) found this direction of effect for London, but no association for South Yorkshire.

For violent crimes, 2 studies found that higher income inequality in an area was associated with higher rates of violent crime (Whitworth, 2012), specifically higher rates of adult violence related injury (Page, 2015). But one study found that an increase in income inequality was associated with a decrease in violent crime (Alsharkas & Campaniello, 2019). Two studies found there was no significant association (Whitworth, 2013; Wu & Wu, 2012).

For criminal damage, Whitworth (2012) found that higher income inequality in an area was associated with an increase in criminal damage, but Wu and Wu (2012) reported no significant association. For the broad category of property crimes, defined as crime where a victim’s property is stolen or destroyed without the use or threat of force against the victim, Brosnan (2019) found that higher rates of income inequality were associated with a reduction in property crime. Alsharkas and Campaniello (2019) found no significant association.

For fraud, geographical areas linked with higher rates of income (Han et al., 2013; Bandyopadhyay et al., 2011) and higher rates of male income inequality (Wu & Wu, 2012) were associated with higher rates of fraud. However, for total income inequality, there were no significant associations with fraud rates (Han et al., 2013; Bandyopadhyay et al., 2011).

When ‘income rates’ and ‘income inequality’ are measured separately, an increase in income was associated with an increase in sexual offences and violence against the person, but an increase in inequality was associated with a decrease in sexual offences and violence against the person (Bandyopadhyay et al., 2011; Han et al., 2013). Wu and Wu (2012) reported that an increase in income inequality was associated with an increase in sexual offences. Brosnan (2019) found no significant association between income and person crimes, which are commonly defined as violent crime that cause physical, emotional, or psychological harm to the victim.

For dark figure crime, it was found that increases in an area’s income was associated with higher rates of dark figure crimes (Buil-Gil et al., 2021b). Wu and Wu (2012) found that increases in male income inequality were associated with decreases in drug offences and ‘other/miscellaneous crimes’ (for example, handling stolen goods). For weapon carrying/use, there was no significant association with income inequality (Smith & Wynne-McHardy, 2019a).

3.6.5 Socio-demographics (family structure)

The association of family structure with drug use, violent behaviours (for example, weapon carrying/use) and violent crimes were explored. Aston (2015) explored the impact of ‘children not living with 2 parents’, which was associated with increased drug use at aged 16 years, but not at 13 years of age. ‘Low parental supervision’ and ‘viewing school as unimportant’ were also associated with increases in drug use at 16 years whereas ‘parent-child conflicts’ were associated with increases in drug use at 13 years.

One study evaluating 18-year-olds, found ‘a father not present in the family home’ was associated with increases in serious violence linked behaviours, which included weapon carrying/use, gang fighting and robbery (Smith & Wynne-McHardy, 2019b).

Livingston et al. (2014) found there was no association between ‘adult only’ or ‘lone parent households’ and all crimes, violence against the person, or property crimes. Similarly, Smith and Wynne-McHardy (2019a) found that, compared to’ living with both natural parents’, when a child ‘lived with their natural mother only’, ‘natural mother and a stepparent’, or ‘carer or adoptive parent’, there was no significant association with weapon carrying/use.

Smith and Wynne-McHardy (2019a) examined the ‘number of siblings in a household’ compared to the reference category of one sibling. Households with four or more siblings was associated with an increase of weapon carrying/use. Smith and Wynne-McHardy (2019b) also examined the ‘number of siblings in a household’ compared to the reference category of 2 children, ‘the mothers age at first birth’, and ‘being in foster care’ as predictors of violent crimes. There was no significant association between any of these predictors and violent crime.

3.6.6 Socio-economic status (SES)

Socio-economic status (SES) was researched as a predictor for both drug use and weapon carrying/use. A person’s SES is based on the type of work they do, or what they used to do if they are retired (GOV.UK, 2018). Aston (2015) found that an increase of ‘parents in manual jobs’ or ‘parents unemployed’ was associated with lower levels of drug use at the age of 16, but no significant association was found between socio-economic status and drug use at the age of 13 years.

Smith and Wynne-McHardy (2019a) found that when looking at ‘parent’s educational attainment’, individuals who had parents who did not attain any qualifications were associated with increases weapon carrying/use. No significant association between parents in manual jobs, or parents unemployed, and the likelihood of weapon carrying/use was found.

3.6.7 Affluent areas

One study explored ‘affluent areas’ and their association with theft (Mburu & Helbich, 2016). Affluence was measured through combining both the proportion of household members that held middle or senior management roles, and the proportion of homeowners. The study explored the level of risk associated with affluent areas using four proximities (within 160m, 320m, 480m and 640m of street segments). Affluence was associated with a 11% rise in the risk of theft within 640m of street segments (Mburu & Helbich, 2016), but not at closer proximities.

3.7 Population

Findings from the research will be presented for predictors listed in Table 3 which all fall within the higher-level grouping of population.

Table 3: Population predictors that have been identified as affecting police demand, including the strength of evidence for each predictor

Predictor variable	Evidence rating
Population density	Good
Population churn	Good
Total population	Good
Immigration	Good
Prison population	Fair
Population at risk	Low

3.7.1 Population density

A rise in ‘population density’ in an area was associated with a rise in property crime (Tarling & Dennis, 2016), drug offences, sexual offences, theft (Wu & Wu, 2012), and criminal damage (Whitworth, 2012). Though one study found population density was associated with a reduction in criminal damage (Wu & Wu, 2012).

For vehicle crimes, most research suggested an increase in population density was associated with a decrease in vehicle crime (Quick et al., 2018; Whitworth 2013). However, one study found a 1% rise in population density was associated with a 0.22% increase in vehicle crime (Whitworth, 2012) and another study reported no significant association between the 2 (Wu & Wu, 2012).

Research examining any associations between population density and crimes of burglary and robbery was mixed. While some studies found that as population density increased, there was an associated rise in burglary rates (Davies & Johnson, 2014; Whitworth, 2012) and robbery rates (Whitworth, 2012; Wu & Wu, 2012), but one study found burglary and robbery rates reduced (Quick et al., 2018). Similarly, for London data, a 1% rise in population density was associated with a 3% decrease in burglaries and a 2% decrease in robberies but, for South Yorkshire data, there was no significant association for either crime (Whitworth, 2013). Wu and Wu (2012) also reported no significant relationship between population density and burglary.

For violent crimes, some studies found that as population density increased, there was an associated increase in violent offences (Tarling & Dennis, 2016; Whitworth, 2012; Wu & Wu, 2012). However, in London and South Yorkshire population density was associated with a reduction in violence (Whitworth, 2013). Further, population density was associated with a reduction in violent crimes (Quick et al., 2018) and rioting (Kawalerowicz & Biggs, 2015).

There was no significant association between population density and dark figure crimes (Buil Gil et al., 2021b) or fraud (Wu & Wu, 2012).

3.7.2 Population churn

‘Population churn’ is the measure of the intensity of residential mobility and helps to quantify the stability of a population in an area. Most of the research found that increased population churn was associated with higher rates of burglaries (Whitworth, 2012; Quick et al., 2018) and robberies (Quick et al., 2018; Whitworth, 2012; 2013). Only one study reported no significant association between ‘population turnover’ and burglary (Whitworth, 2013).

The evidence for population churn as a predictor was mixed for vehicle crime, violent crime, and property crime. For vehicle crime, while one study concluded that higher rates of ‘residual instability’ was associated with higher rates of vehicle crime (Quick et al., 2018), another study found it was associated with lower rates of vehicle crime (Andrews, 2011), and 2 studies found no significant association between population churn and vehicle crime (Whitworth, 2012; 2013).

For violent crime, while 2 studies found that increases in ‘transient population’ (Tarling & Dennis, 2016), and residential instability (Quick et al., 2018) were associated with higher rates of violent crime, 2 studies found no significant association between population turnover and violent crime (Whitworth, 2012; 2013).

Finally, Whitworth (2012) reported that an increase in population turnover was associated with a decrease in criminal damage offences. Tarling and Dennis (2016) found that an increase in the transient population was associated with an increase in property crime offences.

3.7.3 Total population

Most research exploring ‘total population’ was consistent and suggested that this predictor was linked with an increase in crimes. An increase in the total population was associated with an increase in burglaries (Davies & Johnson, 2014; Whitworth, 2012), robberies, vehicle crimes (Whitworth, 2012), and arson offences (Andrews, 2011). One study which used data for England suggested an increase in the total population was associated with a decrease in violent crimes, and there was no significant association with criminal damage (Whitworth, 2012).

An increase in theft offences was associated with increases in ‘residential population’ and ‘workday population’ (Malleson & Andresen, 2016). Residential population was also associated with slight and serious RTAs but, no there was no significant association with fatal RTAs (Wang et al., 2009).

3.7.4 Immigration

Immigration was measured using several variables such as ‘A8 immigrants’ (individuals who arrived in the UK post 2004 from a European union country), ‘T2/work permit immigrants’ (skilled workers who wish to enter the UK to take up employment), ‘place of birth’, and ‘percentage of people born in the UK’.

Most of the research investigating immigration as a predictor relates to property crime. Bell and Machin (2011) found varied findings depending on the type of immigrant route being explored. They found that higher numbers of A8 immigrants and all T2/work permit immigrants were associated with lower rates of property crime. However, when breaking up these immigrant route further, higher numbers of ‘T2/work permit immigrants who are claiming benefits’ and the ‘number of younger A8 immigrants (16 to 24 years)’ were associated with higher rates of property crime. There was no significant association between T2/work permit immigrants (16 to 24 years) and property crime.

Bell et al. (2013) also found mixed findings. While an increase in ‘asylum population’ and ‘benefit claimant rate’ were associated with a rise in property crime, higher numbers of A8 immigrants and ‘young immigrants (16 to 24 years)’ were associated with lower rates of property crime. There was no significant association between A8 immigrants and property crime. Similarly, Jaitman and Machin (2013) also found no significant association between immigration and property crime.

For violent crimes, several papers reported no significant findings between immigration and violent crimes (Bell & Machin, 2011; Bell et al., 2013; Jaitman & Machin, 2013). While one study reported that higher numbers of T2/work permit immigrants (16 to 24 years) were associated with fewer violent crimes (Bell & Machin, 2011), another study reported that ‘place of birth frontier’ was associated with higher rates of violent crimes (Dean et al., 2019).

Place of birth (for non-UK born population) was also associated with higher rates of burglary, theft, vehicle crimes and crime rates (Dean et al., 2019). However, Jaitman and Machin (2013) found no significant association between immigration and crime rate.

Finally, Buil Gil et al. (2021b) found that as the percentage of people born in the UK in the geographical area increased, the amount of dark figure crimes decreased.

3.7.5 Prison population

One study examined the relationship between ‘prison population’ and 5 crime types. Prison population is defined as the potential total number of prisoners that an establishment can hold in each PFA, considering control and security of the prison. Findings suggested higher prison populations in an area were associated with lower rates of burglary, robbery, and sexual offences, but higher rates of violence against the person (Han et al., 2013). There was no significant association for fraud and forgery rates. No specific search was undertaken for crime in prison as this was beyond the scope of the review.

3.7.6 Population at risk

One study explored the ‘population at risk’ of slight, serious, and fatal RTAs. Haynes et al. (2007) proposed RTAs would be positively associated with the size of the resident population, but adjusted, to account for the national causality rates which are experienced at different age and sex groups. This was measured by multiplying each local authority district’s resident population in age and sex categories by crime type. Findings suggested that increases in population at risk was associated with an increase in slight, serious, and fatal RTAs.

3.8 Traffic environment

Findings from the research will be presented for predictors listed in Table 4 which all fall within the higher-level grouping of traffic environment.

Table 4: Traffic environment predictors that have been identified as affecting police demand, including the strength of evidence for each predictor

Predictor variable	Evidence rating
Volume of traffic	Fair
Road layout	Fair
Road length	Fair
Air pollution	Fair

For ease of reporting, all traffic environment predictors will be discussed together. Overall, volume of traffic was a frequent predictor linked to RTA’s. Increases in ‘average traffic activity’ were associated with more slight, serious, and fatal RTAs (Haynes et al., 2007; Jones et al., 2008; Quddus, 2008; Wang et al., 2009). ‘Individuals who own more cars’ were associated with being involved in slight, serious (Haynes et al., 2007; Jones et al., 2008) and fatal RTAs (Jones et al., 2008).

Similarly, increases in the ‘total length of roads’ in an area was associated with slight, serious, and fatal RTAs occurring (Haynes et al., 2007; Jones et al., 2008). The greater the length of A roads, B roads, minor roads, and motorways were all associated with higher rates of RTAs (Quddus, 2008; Wang et al., 2009). Other research showed that ‘areas with a higher percentage of minor roads’ were associated with fewer RTAs (Haynes et al., 2007; Jones et al., 2008).

In terms of road layout, more ‘nodes’ (that is, the number of junctions or end of road points) were associated with more fatal RTAs (Wang et al., 2009). However, there was no significant association between nodes and serious or slight RTAs. An increased ‘number of roundabouts’ were associated with more slight and serious RTAs (Wang et al., 2009). One study that explored road layout characteristics found an association that “streets with more potential usage” were found to be associated with higher burglary rates (Davies & Johnson, 2014).

A further study investigated the effect of air pollution on individual’s cognitive driving performance, such as reaction time or attention span (Sager, 2016). ‘Rising Nitrogen Dioxide (NO2) levels’ were associated with a rise in RTAs (Sager, 2016). Bondy et al. (2020) examined the daily ‘air quality index (AQI)’ of London over 2 years, as this major urban city has similar characteristics to other major cities around the world. It was found that an increase in 10 AQI units was associated with an increase in crime overall, types of violent crimes which include common assaults, harassment, or sexual offences, and both residential and non-residential criminal damage. There was no significant association between AQI and motor vehicle criminal damage, theft, burglary, robbery, or other types of violent crimes which include acute bodily harm, grievous bodily harm, murder, or rape (Bondy et al., 2020).

3.9 Demographic factors

Findings from the research will be presented for predictors listed in Table 5 which all fall within the higher-level grouping of demographic factors.

Table 5: Demographic factor predictors that have been identified as affecting police demand, including the strength of evidence for each predictor

Predictor variable	Evidence rating
Gender	Good
Young people (15 to 24 years)	Good
Young people (16 to 29 years)	Good
Ethnic composition	Good
Young people (10 to 15 years)	Fair
Other age groups	Fair
Deviant youth behaviour	Fair
Peer group	Low

3.9.1 Gender

Fitzsimons et al. (2018) found that when combining ASB (anti-social behaviour), including graffiti, public nuisance, and vandalism, gender was associated with ASB, males were 20% more likely to engage in ASB than women. However, when measuring the ASB factors separately, there was no significant association.

One study found no significant association between gender and gang membership or theft (Fitzsimons et al., 2018). However, another study found that being male was associated with an increase in violent crimes, which included gang fighting, weapon carrying/use, and robbery (Smith & Wynne-McHardy, 2019b). Two additional studies showed males were more likely to carry a weapon (Brennan, 2019; Smith & Wynne-McHardy, 2019a).

In addition, males were more likely to report that they had committed an assault, engaged in cybercrime (Fitzsimons et al., 2018), and were involved in drug use at both 13- and 16-years of age (Aston, 2015).

While one study found that males were associated with decreases in rates of property crime (Alsharkas & Campaniello, 2019), another found no significant association for property crime or violence against the person (Brosnan, 2019). Other studies also reported no significant association between males and dark figure crimes (Buil-Gil et al., 2021b) or violent crimes (Alsharkas & Campaniello, 2019).

3.9.10 Young people (15 to 24 years of age)

Studies have explored the population of various ages in an area as predictor. An increase in ‘people aged 15- to 24-years’ in an area was associated with an increase in burglary and theft, when measured in ‘all areas’ and ‘low crime areas’ (Bandyopadhyay et al., 2011), but there was no significant association within ‘high crime areas’. However, Han et al. (2013) found an increase in people aged 15- to 24-years was associated with less burglaries, and no significant association with the rates of theft.

For robbery, findings suggested an increase in 15- to 24-year-olds was associated with both an increase (Bandyopadhyay et al., 2011), and decrease (Han et al., 2013) in robbery offences.

In high crime areas only, increases in 15- to 24-year-olds were associated with lower rates of fraud (Bandyopadhyay et al., 2011). But Han et al. (2013) found no significant association between the 2, and there was no significant association in ‘low crime’ or ‘all areas’ (Bandyopadhyay et al. (2011).

An increase in the number of ‘people aged 16- to 24-years’ in an area was associated with an increase in racially or religiously aggravated harassment, violence against the person and criminal damage (Williams et al., 2019). In ‘low crime areas’ only, an increase in the number of 15- to 24-year-olds in an area was associated with an increase in violence but not ‘high crime’ or ‘all’ areas (Bandyopadhyay et al., 2011).

Tarling and Dennis (2016) suggested an 1% increase in 15- to 24-year-olds was associated with a 0.66% rise in property crime but found no significant association with violent crimes. Higher numbers of 16- to 24-year-old was associated with an increase in rioting (Kawalerowicz & Biggs, 2015). There was no significant association between 15- to 24-year-olds and sexual offences (Bandyopadhyay et al., 2011; Han et al., 2013).

3.9.11 Young people (16 to 29 years of age)

For burglary, a 1% rise in ‘people aged 16 to 29 years’ was associated with a 0.19% decrease in burglary rates (Whitworth, 2012). When examining data for South Yorkshire and London areas, a 1% rise in people aged 16 to 29 years was associated with a 2% increase in burglary rates in both areas (Whitworth, 2013).

Further, a rise in the number of 16- to 29-year-olds was associated with an increase in violence (Whitworth, 2012; 2013), criminal damage (Whitworth, 2012), and when using London data only, an increase in robbery (Whitworth, 2013). For South Yorkshire data only, there were no significant associations between 16- to 29-year-olds and robbery (Whitworth, 2012; 2013).

A rise in 16- to 29-year-olds was associated with an increase in vehicle crime using London data, but no significant association using South Yorkshire data (Whitworth, 2013) or for all of England (Whitworth, 2012). There was no significant association between the proportion of young people aged 16 to 29 years in an area and ‘all crimes’, violence against the person or property crimes (Livingston et al., 2014).

3.9.12 Ethnic composition

There is a large evidence base which has investigated the effect of ethnic composition (for example, ‘the ethnic diversity in an area’) on several demands including violent crime, burglary, robbery, vehicle crime, theft, criminal damage, violence against the person, antisocial behaviour, all crime, and unreported crimes. However, the findings are inconsistent.

The most evidenced demand is violent crime, but findings are mixed. Some studies found that a decrease in ‘number of people who identify as white as a proportion of the population’ (Tarling & Dennis, 2016), a sharp change in ‘ethnic composition’ (Dean et al., 2019), or an increase in ‘ethnic heterogeneity’ (Quick et al., 2018) were associated with increases in violent crime. However, when using London data, a proportional increase in the population of people that identity as non-white was associated with a decrease in violence but, no significant association was found for South Yorkshire (Whitworth, 2013) or at middle layer super output areas (MSOA) level (Whitworth, 2012).

Areas with higher proportions of ‘ethnic minorities’ were associated with higher rates of violence against the person (Livingston et al., 2014; Williams et al., 2019). Similarly, as ‘ethnic minority density’ in an area increases, so does violence related injury rates (Matthews et al., 2006). There was no significant association between ‘non-white population’ and weapon carrying/use (Brennan, 2019; Smith & Wynne-McHardy, 2019a).

Kawalerowicz and Biggs (2015) found that when controlling for deprivation, an area with increased ‘numbers of people who identified as African’ and ‘identifed as Caribbean’ were associated with increases in rioting, when compared to ‘people who identify as being White British’. Further, ‘individuals with a Pakistani origin’ were more associated with increases in rioting compared to those ‘individuals with an Indian origin’.

For burglary, more ‘ethnically diverse areas’ (Gulma et al., 2018), sharp changes in ‘ethnic composition’ (Dean et al., 2019), and ‘ethnic heterogeneity’ (Davies & Johnson, 2014) were all associated with increased rates of burglary when using MSOA data (Whitworth, 2012). However, when looking at South Yorkshire and London data separately, both areas found that an increase in the percentage of non-white population was associated with a decrease in burglary (Whitworth, 2013). A further study found no significant association between ethnic heterogeneity and burglary rates (Quick et al., 2018).

While one study found that a higher percentage of ethnic minorities in an area was associated with higher rates of property crimes (Livingston et al., 2014), another study found no significant association (Tarling & Dennis, 2016).

Two studies found that increases in the percentage of people who identify as non-white in the population (Whitworth, 2012) and ethnic heterogeneity (Quick et al., 2018) were associated with increases in robbery rates. However, data for South Yorkshire and London found no significant association with non-white population and robbery rates (Whitworth, 2013).

Dean et al. (2019) looked at the effect of a ‘sharp change in social/ethnic characteristics between neighbouring communities’ on crimes in Sheffield, compared to neighbouring communities which do not have statistical differences in ethnic populations. Dean et el. (2019) found that this sharp change was associated with higher rates of vehicle crime. Data for London showed a 1% rise in the percentage of people who identify as white non-white population was associated with a 1% decrease in vehicle crimes, but data for South Yorkshire found no significant association (Whitworth, 2013). Finally, there was no significant association between ethnic heterogeneity and vehicle crime (Quick et al., 2018).

For theft, a sharp change in social/ethnic characteristics between neighbouring communities was associated with an increase in shoplifting (Dean et al., 2019) and an increase in ethnic heterogenicity was associated with increases in the risk of bike theft (Mburu & Helbich, 2016).

Two studies explored criminal damage; one found a 1% increase in the percentage of people who identify as non-white population was associated with a 0.1% increase in criminal damage offences (Whitworth, 2012) however, another study exploring the ‘proportion of people who identify as Black, Asian and Minority Ethnic (BAME)’ in an area found there was no significant association with criminal damage or aggravated harassment (Williams et al., 2019).

For ASB, a self-report survey from 14-year-olds found that ‘people who identify as Black Caribbean’ and ‘Black African’ were more likely to state that they engaged in ASB than ‘people who identify as White’. However, there was no significant difference in self-reported rates of ASB for ‘people who identify as Indian, Pakistani, Bangladeshi, mixed race or other ethnic groups’ when compared to people who identify as White (Fitzsimons et al., 2018).

For fraud, Carl (2017) found that an increase in ‘Bangladeshi or Pakistani population’ was associated with an increase in the number of cases of alleged electoral fraud.

Two studies found that a higher percentage of ethnic minorities (Livingston et al., 2014) and a sharp change in ethnic composition (Dean et al., 2019) was associated with higher rates of all crime.

Finally, for dark figure crimes, Buil-Gil et al. (2021b) found a higher ‘percentage of people who identify as Asian’ in an area was associated with lower rates of dark figure crimes however, no significant association was found for the percentage of people who identify as White in an area and dark figure crimes.

3.9.13 Young people (10 to 15 years)

A higher ‘proportion of individuals aged 13 to 15 years’ in an area was associated with fewer person crimes (for example, aggravated assault) (Livingston et al., 2014), however the higher the ‘proportion of individuals aged 10 to 15 years’ in an area was associated with an increase in burglaries (Davies & Johnson, 2014).

When compared to being ‘13 years old’, being ‘15 years of age’ was associated with weapon carrying/use but being ‘14 years old’ was not significantly associated with weapon carrying/use (Smith & Wynne-McHardy, 2019a). There was no significant association between ‘young people’ and antisocial behaviour (Fitzsimons et al., 2018), property crimes, or a measure of all crimes (Livingston et al., 2014).

3.9.14 Other age groups

One study reported that ‘children aged 12 or over’ were more likely to become a missing person more than once, than ‘children aged 11 and under’ (Hutchings et al., 2019). Quddus (2008), examined age using 2 groups (‘residents aged 59 and under’; ‘residents aged 60 and over’) but only residents aged 59 and under were associated with an increased risk of slight RTA casualties occurring.

Increases in the ‘age range in an area’ were associated with increased burglary rates (Gulma et al., 2018). There was no significant association between age and weapon carrying (Brennan, 2019), antisocial behaviour, criminal damage, or sexual offences (O’Brien et al., 2018).

3.9.15 Deviant youth behaviour

Deviant lifestyle behaviours have been defined by Aston (2015) to include behaviours such as offending, drug use, weekly smoking, and drinking. Several predictors were grouped as deviant youth behaviour from the evidence including ‘previous offending’, ‘weekly smoking and drinking’, ‘best friend’s volume of offending and drug use’, ‘weak social bonds’ (with parents and teachers), ‘more than a few peers in trouble with police’, few peers in trouble with police’, ‘truanting’, ‘gambling’, and ‘youth conceptions’ (unplanned pregnancies).

Aston (2015) found that previously offending (1 to 3 times), and an individual’s best friend’s drug usage, were associated with the individuals own drug use at 16 years of age but not at 13 years of age. Previous offending between 4 and 10, 11 and 21, and 22 or more times, weekly smoking and drinking and weak social bonds were all associated with increased drug use at both ages. The individual’s best friend’s volume of offending and ‘hanging around outside where the individual lived in the evenings and at weekends’ were not significant associated with drug use at both ages (Aston, 2015).

One study evaluated individual’s peer groups. Brennan (2019) found people who had ‘more than a few peers’ or ‘few peers’ in trouble with the police were associated with an increased likelihood of weapon carrying than those without any peers in contact with the police.

Smith and Wynne-McHardy (2019a) found that for 14-year-olds, truanting, gambling, ‘previous public disorder offences’, ‘cybercrime’, and ‘underage drinking (between 10 and 19 times across 12 months)’ were associated with increased weapon carrying/use. Individuals (14-year-olds) ‘drinking between 1 and 9 times’ or ‘drinking 20 times or more’ were not significantly associated with weapon carrying/use. A later study by Smith and Wynne-McHardy (2019b) found that for 18-year-olds, public disorder offences were not significantly associated with weapon carrying/use, gang fighting and robbery.

Finally, Whitworth (2012) reported that an increase in youth conceptions was associated with an increase in violence and criminal damage, though there were no significant associations for vehicle crime, robbery, or burglary.

3.10 Geographical location

Findings from the research will be presented for predictors listed in Table 6 which all fall within the higher-level grouping of geographical location.

Table 6: Geographic location predictors that have been identified as affecting police demand, including the strength of evidence for each predictor

Predictor variable	Evidence rating
Average house price	Good
Police force factors	Good
Land use	Good
Housing mix	Fair
Seasonal effects	Fair
Vacant properties	Low
Public amenities	Low

3.10.1 Average house price

While Whitworth (2012) found, when using MSOA level data for England, that increased ‘average house prices’ were associated with a reduction in burglary, robbery, vehicle crime and criminal damage rates. When using data only for London and South Yorkshire, it showed no significant associations between average house price and these crime types (Whitworth, 2013). There was no significant association between average house price and property crimes (Tarling & Dennis, 2016).

In terms of violent crimes, while 2 studies found that increased ‘house prices’ (Matthews et al., 2006) and average house prices (Whitworth, 2012) were associated with a decrease in violence related injuries and violent crimes respectively, 2 later studies reported no significant associations between the average house price and violent crimes (Whitworth, 2013; Tarling & Dennis, 2016).

Further, higher house prices were associated with a lower percentage of dark figure crimes (Buil-Gil et al., 2021b).

3.10.2 Police force factors

‘The sanctioned detection rate’ is defined as the proportion of recorded crimes that have been assigned an outcome by the police (GOV.UK, 2013b). Most of the research examined suggested that higher clearance rates for crimes in a police force area (PFA) were associated with a reduction in reported crimes.

An increase in the sanctioned detection rate was associated with a decrease in robbery (Bandyopadhyay et al., 2011; Whitworth, 2012; Wu & Wu, 2012), violent crimes (Whitworth, 2012; Wu & Wu, 2012), fraud (Bandyopadhyay et al., 2011; Wu & Wu, 2012), violence against the person (Bandyopadhyay et al., 2011), theft, and sexual offences (Bandyopadhyay et al., 2011). Though another study found no significant association between the sanctioned detection rate and sexual offences (Wu & Wu, 2012).

While one study found that higher sanction detection rates were associated with a decrease in burglary rates (Bandyopadhyay et al., 2011), 2 studies found there was no significant association (Whitworth, 2012; Wu & Wu, 2012). Similarly, Wu and Wu (2012) found that higher sanction detection rates were associated with a decrease in criminal damage offences, but Whitworth (2012) found no significant association.

An increase in the sanctioned detection rate was associated with an increase in drug offences (Wu & Wu, 2012). There was no significant association between detection rate and vehicle crime (Whitworth, 2012), or other offences (Wu & Wu, 2012).

One study explored the ‘number of officers working in a police force area (PFA)’ (Whitworth, 2012). Higher numbers of officers in a PFA were associated with a decrease in criminal damage. There were no significant associations with burglary, robbery, vehicle crimes or violent crimes.

3.10.3 Land use

A few studies explored various uses of areas of land, (for example, urban land, retail outlets and bars) and the association with police demands. Haynes et al. (2007) found that areas with a higher ‘percentage of urban land’ were associated with higher rates of slight and serious RTAs, but decreased rates of fatal RTAs.

Higher proportions of urban land were associated with increases in the percentage of dark figure crimes (Buil Gil et al., 2021b). Additionally, ‘suburban neighbourhoods’ that are located outside of the main conurbations were associated with increases in dark figure crimes. There was no significant association between the level of suburban area and dark figure crimes (Buil Gil et al., 2021b).

For theft, Bowers (2014) reported that higher ‘numbers of retail shops’ were associated with increases in thefts from the person inside a building. Another study reported that increased ‘opportunities for looting’ was associated with increased rioting (Kawalerowicz & Biggs, 2015). Further, Bowers (2014) found that higher ‘numbers of facilities with a history of four or more thefts’ were associated with increases of thefts from inside the building and decreases of thefts outside the building. The ‘number of recreational land parcels’, ‘service land parcels’, or ‘other commercial land parcels’ were not significantly associated to theft either inside or outside the buildings.

‘Alcohol outlets in an area’ were associated with an increase in person crimes, property crimes and all crimes (Livingston et al., 2014). ‘Bars’ were associated with increases in bike theft when the bar was measured within 640m of the theft; there were no significant associations for the bar measured within 160m, 320m or 480m (Mburu & Helbich, 2016).

Two studies found that rises in ‘on and off trade alcohol prices’ (Page, 2015) and ‘real beer price’ (Matthews et al., 2006) were associated with lower rates of violent related injuries.

3.10.4 Housing mix

Housing clusters refer to a development in which the houses are arranged in relatively close groups. Clusters include ‘owner occupied and private renting’, ‘owner occupied and social renting’, ‘social renting and owner occupied’, ‘social renting’, ‘other flats’ and ‘high flats. The first label is the most common housing in the given area.

Using crime data from 2008, owner occupied, and private renting areas were associated with an increase in all crimes and person crimes but not property crimes (Livingston et al., 2014).

Owner occupied and social renting areas were associated with an increase in person crimes but not all crimes or property crimes. Social renting and owner-occupied areas, or only social renting areas were associated with an increase in all crimes and person crimes but not property crimes.

When exploring high flats, and other flats (Livingston et al., 2014) there were no significant associations between high flats and all crime, person crimes or property crimes. However, other flats were associated with an increase in all crimes and property crimes but not person crimes.

One study examined areas which were rated to be ‘not safe at all’ or ‘not very safe’ with a reference category of ‘very safe or safe’ areas; findings suggested living in unsafe areas was associated was higher rates of weapon carrying/use (Smith & Wynne-McHardy, 2019a).

For student households specifically, one study found that an increase in ‘student housing’ in one area was associated with higher rates of burglaries (Davies & Johnson, 2014).

3.10.5 Seasonal effects

Two studies found that violence related injuries (Matthews et al., 2006) and violent crimes (Page, 2015) reduced in ‘autumn’, ‘winter’ and ‘spring’ when compared to ‘summer’.

Sager (2016) explored different weather conditions and their relationship with RTAs. Increases in ‘ground temperature’, ‘humidity’, ‘rainfall’, and ‘windspeed’ were all associated with an increased risk of RTAs, whereas an increase in ‘cloud coverage’ was associated with a reduction in risk of RTAs.

3.10.6 Vacant properties

Two studies explored vacant houses, that is, houses that were unoccupied. Mburu and Helbich (2016) found that ‘the presence of vacant houses’ was associated with an increased risk of bicycle theft, and Davies and Johnson (2014) reported no significant associations between the ‘percentage of vacant houses’ and burglary rates.

3.10.7 Public amenities

Two studies evaluated various public amenities and their associated risk with theft. Mburu and Helbich (2016) found that ‘the presence of bicycle rentals’, ‘pawnshops’, ‘universities’, ‘cycle stands’, and ‘train stations’ were associated with an increased risk of bicycle theft however, there was no association with ‘police stations’ and ‘cycle repair shops’. ‘Tree coverage’ reduced the risk of bicycle theft at shorter proximities (160m, 320m) but not at larger proximities (480m, 640m) (Mburu & Helbich, 2016).

Another study explored the features of train stations (Newton et al., 2014). Internal factors which included ‘staff levels’, ‘below surface platforms (that is, underground)’, ‘surface platforms (overland)’ and ‘shop rentals’ were all associated with reduced risk of theft. However, external factors which included ‘roads and paths around the station’, ‘tourists at the station’, and ‘terminus stations’ were all associated with an increase in the risk of theft.

3.11 Events

Findings from the research will be presented for predictors listed in Table 7 which all fall within the higher-level grouping of events.

Table 7: Event predictors that have been identified as affecting police demand, including the strength of evidence for each predictor.

Predictor variable	Evidence rating
Political events	Fair
Terror attacks	Fair
Major sporting events	Fair

3.11.1 Political events

Hate crimes are recognised as any criminal offence perceived to be motivated by prejudice or hostility towards a person’s race, religion, disability, sexual orientation, or transgender identity.

Devine (2018) collected data using HO official statistics and reported that the ‘UK’s EU referendum’ was associated with an increase in hate crimes. In addition, Carr et al. (2020) compared the numbers of racial and religious hate crimes before and during the EU Referendum period. While there was no significant association before the referendum, during the lead up to the vote, were was an associated increase in racial and religious hate crimes.

When focusing on daily data around the EU Referendum, the day of the vote was associated with a reduction in racially and religiously aggravated offences however, across the following 15 days after the vote, this was associated with an increase in hate crimes (Piatkowska & Stults, 2021).

3.11.2 Terror attacks

Commonly, terror attacks are known to be associated with increases in hate crimes. Terror attacks, such as the ‘2017 Finsbury Park Mosque’, ‘Westminster’ (Devine, 2018; Piatkowska & Stults, 2021), ‘London bridge’ (Devine, 2018) and ‘Manchester Arena attack’ (Piatkowska & Stults, 2021), were associated with sharp increases in hate crimes in the aftermath of the event, but these crimes reduce much faster than the EU referendum, where the aftereffects are more prolonged. Another study reported an immediate increase in Islamophobic hate crimes directly after a terror attack occurs (Ivandic et al., 2019).

3.11.3 Major sporting events

One study was conducted exploring the ‘ambient population surrounding a stadium’ on match and non-match days and its association with 4 crime types (Ristea et al., 2018). Increases in ambient population were associated with increases in criminal damage, theft, violence against the person and all crimes on match days and non-match days. Most associations were higher for these crimes on match days compared to non-match days.

Although not recent, a further study examined ‘national major sporting events’ (between 1996-1999) were associated with an increase in violence related injuries (Matthews et al., 2006).

3.12 Individual factors

Findings will be discussed from predictors listed in Table 8 which all fall within individual events.

Table 8: Individual factor predictors that have been identified as affecting police demand, including the strength of evidence for each predictor

Predictor variable	Evidence rating
Alcohol	Fair
Mental health	Fair
Drug use	Low
Past victims	Low

3.12.1 Alcohol

One study reported that ‘men who have an alcohol dependence’ were associated with committing a violent offence or being a member of a gang (Coid et al., 2013). Another study explored AUDIT-C, which is an alcohol screening measure to identify individuals who are hazardous drinkers or have present alcohol use disorders. O’Brien et al. (2018) found that both ‘men and women with higher AUDIT-C scores’ were more likely to have engaged in violent crime, criminal damage, and sexual offences.

3.12.2 Mental health

One study found that ‘men who had experience with mental health, such as symptoms of psychosis, anxiety, depression, antisocial personality disorder, suicidal attempts, and psychiatric admissions’ were associated with violent offences or gang membership (Coid et al., 2013).

Smith and Wynne-McHardy (2019b) found ‘being diagnosed at 10 years of age by a health professional for a learning, behaviour, development, or mental health problem’, was associated with engaging in violent behaviours. However, when compared to having no symptoms, there were no significant associations between having ‘1, 2, or 3 or more symptoms of antisocial personality disorder’ and violent behaviours which were defined as weapon carrying/use, gang fighting and robbery (Smith & Wynne-McHardy, 2019b).

Smith and Wynne-McHardy (2019a) found that ‘self-harming behaviours’ were associated with weapon carrying/use, but ‘parents suffering with their mental health’ was not significantly associated with a child engaging in weapon carrying/use.

3.12.3 Drug use

While Coid et al. (2013) found that ‘men who are drug dependent’ were more likely to be a member of a gang, the same study also found no significant association between drug dependence and violent crimes. Another study found no significant association between ‘parental substance abuse’ and violent crimes (Smith & Wynne-McHardy, 2019b).

While one study found that ‘individuals who had used drugs in the past year’ were more likely to carry a weapon (Brennan, 2019), another found no significant association between ‘individuals who had ever tried drugs’, or ‘whether their parents engaged in drug use’, and weapon carrying/use (Smith & Wynne-McHardy, 2019a).

3.12.4 Past victims

Several papers found that ‘people who had experienced violence’ (Brennan, 2019) or ‘had been a victim of violence’ (Brennan, 2019; Smith & Wynne-McHardy, 2019a), were more likely to be carrying a weapon than those who had not. However, there was no significant association between ‘individuals who had been threatened with violence’ and weapon carrying (Brennan, 2019).

3.13 COVID-19

Findings will be discussed from predictors listed in Table 9 which all fall within COVID-19.

Table 9: COVID-19 predictors that have been identified as affecting police demand, including the strength of evidence for each predictor

Predictor variable	Evidence rating
COVID-19 effect on demands	Good
COVID-19 effect on predictors	Good
Unemployment due to COVID-19	Good

The current review was conducted during the global pandemic, COVID-19, which saw governments across the world commence national lockdowns in the UK. These occurred in March 2020, October 2020, and January 2021. The effects of the pandemic were identified as not only impacting economies and unemployment rates, but it also impacted the demands on police and how crime was experienced.

3.13.1 COVID-19 effect on demands

COVID-19 directly impacted and changed crime and non-crime incident demands. Neanidis and Rana (2021) found that ‘COVID-19 related lockdowns in 2020’ were associated with increases in antisocial behaviour rates, drug offences and public order offences compared to pre-lockdown.

For burglary offences, most research suggests that lockdown measures were associated with reduced burglary rates (Nivette et al., 2021; Sun et al., 2021; Neanidis & Rana, 2021; Halford et al., 2020). Similarly, lockdown measures were associated with reduced robberies (Nivette et al., 2021; Neanidis & Rana, 2021), criminal damage/arson, violence, sexual offences, and possession of weapons (Neanidis & Rana, 2021) compared to pre-lockdown. However, the ‘COVID-19 infection rate’ had no significant effect on robbery rates in March 2020, April 2020, or May 2020 (Sun et al., 2021).

One study that explored COVID-19 effects on cybercrime found there was an increase in personal hacking, social media and email hacking, online fraud and all cybercrimes in May 2020 (during lockdown) compared to May 2019 (pre-lockdown), but a decrease in computer viruses and hacking combined with extortion (Buil-Gil et al., 2021a). There was no significant difference in server hacking in May 2019 and May 2020. For fraud, there was an increase in romance fraud, for both men and women aged 20 to 59 years in 2020 compared to 2019 (Buil Gil et al., 2021c).

For domestic abuse, lockdown was associated with increases in domestic abuse from current partners and family/guardians compared to pre-lockdown, however, there was a decrease in domestic abuse from ex-partners (Ivandic et al., 2020).

Although there was no significant relationship between the COVID-19 infection rate and theft in March 2020 (Sun et al., 2021), a higher COVID-19 infection rate in both April and May 2020 was associated with lower rates of theft (Sun et al., 2021). Similarly, 3 studies found that both theft and vehicle theft were lower in lockdown compared to pre-lockdown (Halford et al., 2020; Neanidis & Rana, 2021; Nivette et al., 2021). For bicycle theft, shoplifting and theft from the person, there were no significant changes between pre-lockdown and lockdown (Neanidis & Rana, 2021).

Lockdown measures were associated with reduced assaults (Nivette et al., 2021; Halford et al., 2020) and while there were no significant changes, homicide rates were lower during periods with lockdown measures (Nivette et al., 2021). The probability of being a victim of hate crime increased due to the pandemic if you are Chinese compared to other ethnic groups across the same time frame. (Gray & Hansen, 2021).

3.13.2 COVID-19 effect on predictors

COVID-19 directly impacted variables that have been previously seen to be predictors of crime demand. These include alcohol, poverty, and unemployment.

Daly and Robinson (2021) reported that 2020 lockdowns were associated with increases in individuals’ drinking habits compared to 2017 to 2019 levels. This was for all age groups, both females and males, white and non-white groups, and different income households. In addition, findings suggested a higher proportion of the population displayed high risk drinking (5 or more drinks per day) after lockdown, compared to pre-lockdown (Jackson et al., 2021).

While AUDIT-C scores did not significantly alter over lockdown (Rao et al., 2021); there was an increase in the percentage of older adults scoring between 0 to 3 points which indicates low-risk drinking, but a decrease in older adults scoring more than 5 points which indicates high-risk drinking.

Alcohol consumption increased during 2020 compared to 2019 in adults who had children at home (Ingram et al., 2020). There were no significant results when exploring differences in alcohol consumption in students, work status changes, contracting COVID-19 in the individual, having a COVID-19 case in the household, when shielding, in vulnerable adults or in adults self-isolating (Ingram et al., 2020). However, one study found that there was a reduction in university student’s alcohol use (Evans et al., 2021).

Brewer and Tasseva (2020) explored poverty lines for the entire population. When measuring using a fixed poverty line, poverty increased throughout the pandemic for the entire population. However, when measuring using a floating poverty line, which looks at how individuals in the lowest income bracket are affected, poverty decreased for the entire population (Brewer & Tasseva, 2020).

Su et al. (2021) reported that the UK unemployment rate in 2020 increased compared to the 2019 unemployment rate, but there was no association between COVID-19 related deaths and unemployment rates in 2020 compared with 2019 (Su et al., 2021).

3.13.3 Unemployment due to COVID-19

Unemployment due to COVID-19 also was associated with crime demands. Increases in ‘numbers of furlough claimants’ were associated with reduced burglary rates, public order, violence, and sexual offences. However, a higher rate of claimants was associated with higher rates of antisocial behaviour, drug offences, bicycle thefts and vehicle crimes. There was no significant association between higher rates of claimants and robberies or shoplifting (Kirchmaier & Villa-Llera, 2020).

4. Discussion

4.1 Key findings

4.1.1 Socio-economic factors

The combination of deprivation, unemployment, education, income inequality, sociodemographic (family structure) and socio-economic status (SES) have created a large higher-level grouping of socio-economic factors. Each predictor has been shown to be associated with police demand and many of the predictors overlap, influence, and drive each other. For example, deprivation can be influenced by unemployment rates, lower educational attainment, long-term illnesses, and living in shared housing (Mburu & Helbich, 2016).

Deprived areas were consistently associated with an increase in a range of crime demands including burglaries, fraud, property crime, robberies, violent crimes, and RTAs. A possible way of interpreting this finding is that the level of deprivation in an area can influence crimes, as there can be a greater need to commit crimes, to ensure individual’s needs are met (Livingston et al., 2014). The nature of deprivation and its association with higher frequency of RTAs occurring could be explained by many factors including environmental factors, which may present themselves as more hazardous in more deprived areas, or a culture of poorer driving existing in these areas (Jones et al., 2008).

The reviewed literature tended to use different variations of income inequality, at times dividing income and inequality and analysing separately, or measuring male income inequality specifically. However, most of the research for income inequality consistently suggested that an increase in income inequality in an area was associated with increases in burglary, fraud, robbery, and theft, but findings for other crimes, such as, criminal damage, and sexual offences were mixed. Han et al. (2013) suggested that while higher unemployment motivates potential offenders to commit crime, it decreases their opportunity to do so due to not having a workplace. Conversely, those in employment have increased opportunity to commit crimes. This increase in opportunity could increase certain crime rates (for example, burglary, theft, and fraud). Therefore, the net effect of unemployment or income rate will depend on which effect is stronger, motivation or opportunity (Han et al., 2013).

Unemployment rates were generally associated with rises in violent crimes, but findings for many other crime types such as burglary, robbery, property crime, and fraud were mixed. Unemployment could be the driving force which motivates offenders to commit crimes. Conversely, there could be fewer opportunities available to the offenders as their networks and interactions are reduced (Han et al., 2013).

Generally, lower levels of educational attainment were associated with increases in certain crimes, such as burglary and robbery, however, there was a general decrease in violent crimes. While research tended to focus on lower level of attainment, Gulma et al. (2018) highlighted more sophisticated crimes such as, cybercrimes, can often be conducted by highly educated individuals.

4.1.2 Population

Findings generally show that increases in density, churn and total population were associated with increases in crime. High residential instability can reduce feelings of security and familiarity within the neighbourhood which can, in turn, reduce an individual’s social connectivity and thus, reduce the chances of intervening in times of need (Shaw & McKay, 1942). For population density, there were mixed findings for burglary and robbery, which could suggest in more populated areas, though there is a higher number of targets, there is equally an increased risk of getting caught (Cohen & Felson, 1979; Quick et al., 2018).

Han et al. (2013) suggested that higher prison population could be expected to reduce crime in the surrounding area due to increased levels of neighbourhood security. Additionally, higher rates of offenders in incarceration and so fewer crimes being committed, could be due to a reduction in the overall pool of criminals available to commit the crime (Han et al., 2013).

For immigration, most of the research focusing on property crime was inconsistent. However, this inconsistency has previously been explained in terms of labour market attachment (Bell et al., 2013), with immigrants who have come to the UK with an interest to work (A8 and T2/work permit immigrants) being less likely to commit property crime compared to those with poor labour market opportunities (low paid workers, asylum seekers).

4.1.3 Traffic environment

Increases in volume of traffic, total road length, nodes (for example, number of junctions or road end points) and air pollution were associated with increases in slight, serious, and fatal RTAs. It is reasonable to assume that higher road density (number of cars) creates more opportunity for RTAs to happen (Wang et el., 2009). Similarly, length of roads increases RTA exposure (Zeng & Huang, 2014). While junctions and road endpoints were associated with fatal RTAs, roundabouts were associated with slight or serious RTAs. This supports previous research which has found that traditional junctions can cause more fatal RTAs than roundabouts (Hels & Orozova-Bekkevold, 2007), particularly for vulnerable road users such as cyclists and motorcyclists (Shen et al., 2020).

Increased air pollution was associated with increases in overall crime and RTAs. For the latter, the authors conclude that this effect is not due to changes in traffic volume, from analysis with traffic counts, but rather associated with an increase in RTA rate on polluted days. The authors speculate that safer driving performance is reduced through either impaired cognitive performance (for example, reaction time), or rendering them to be more aggressive or impatient drivers (Sager, 2016). The authors acknowledged that air pollution could also affect drivers’ performance in other ways such as distraction due to respiratory issues. However, a previous review has found that studies have reported reductions in pollutants due to electric cars (Requia et al., 2018), which may make results like this less problematic.

4.1.4 Demographic factors

Deviant youth behaviour and weak social bonds (for example, with parents/teachers and peers in trouble) were associated with drug use and weapon carrying/use. Males were also associated with increases in violent behaviours including weapon carrying/use and antisocial behaviour, as well as crimes such as assault and cybercrime. Such findings suggest that males engage in higher levels of criminal behaviours (McArthur et al. 2012), with the influence of peers known to increase this risk taking (Steinburg, 2008). These findings align with the 2019 crime statistics which show that of all the convictions, 73% are males and 27% are females (GOV.UK, 2020).

While age was split into granular categories (including 10 to 15 years, 15 to 24 years, 16 to 29 years, and all ages), most of the research found the direction of associations for crime types such as burglary, robbery, theft, fraud, vehicle crimes, violent crimes and weapon carrying/use was mixed. The only exception was that for burglary, an increase of 10 to 15-year-olds in an area, and an increase in age in an area, were associated with increases in burglary. It is likely that while the former is largely associated with the peak age of offending (Farrington, 1983), then latter refers to an increase in older victims, which presents more opportunities for offenders to potentially steal (Gulma et al., 2018).

Findings for ethnic composition were mixed for burglary, robbery, vehicle crime and antisocial behaviour. However, increases in ethnic composition were associated with offences such as violence against the person. It is thought that social isolation or dissimilar neighbourhoods contribute to escalated violence (Dean et al., 2019).

4.1.5 Geographic factors

Most findings revealed that higher average house prices were associated with a lower rate of crimes such as, burglary, robbery, vehicle crime, criminal damage, violent crimes, and unreported crimes.

The association between housing structures and crime rates varied. While ‘other flats’ were associated with a rise in all crimes and property crimes, ‘high flats’ were not associated with any crime measured. Environmental characteristics such as security or surveillance measures which may be found in high flats could be a deterrent to offenders (Livingston et al., 2014). Student housing, which can tend to be cheaper accommodation and less frequently resided in (Tilley et al., 1999), was found to be associated with an increase in burglaries. As students tend to reside for shorter tenancies, the household’s anonymity increases and thus, any social connectivity decreases which loses a natural guardianship for the house (Tilley et al., 1999). Vacant properties were associated with bicycle theft but not burglary, and public amenities such as universities, cycle stands and train stations, were associated with bicycle theft.

Generally, higher sanction detection rates were associated with fewer crimes such as, robberies, violent crimes, thefts, and an increase in drug offences. However, the level of crime in an area should be considered, as PFAs with less crime could arguably have more resource to maintain a higher sanction detection rate. Further, the finding that higher numbers of police officers in an area is associated with a reduction in criminal damage could be explained by the presence of police officers acting as a deterrent for crimes, but this was not the case for burglary, robbery, vehicle, or violent crimes.

Further, internal characteristics (for example, lifts, waiting rooms, fewer platforms) on the London underground tube were associated with a reduction in theft. This was suggested to be from an increase in guardianship through higher staff levels, nearby shops, and the natural dispersion of passengers who spread out across the station to avoid congestion which reduce areas of anonymity (Newton et al., 2014). Following this, external features around tube stations, known as crime attractors, such as, tourists and higher theft rates in the area, were increasing the risk of theft (Newton et al., 2014).

Two studies found generally there were more violent crimes occurring in the summer compared to other seasons. Although outdated and therefore beyond the scope of the review, Shephard (1990) supports a link between increased hours of daylight and an increase in city centre assault, suggesting a prolonged seasonal effect on crime.

4.1.6 Events

Research into political events and terror attacks predominantly focused on the association with higher rates of racial and religious hate crimes. However, for research that investigated hate crime, one consideration is that the data available was of reported hate crimes only, which leaves the possibility that events such as the EU Referendum did not influence a rise in hate crimes, but more so a rise in the reporting of them. Media coverage could also have encouraged people to report hate crimes when previously they would have stayed silent (Devine, 2018).

Major sporting events were associated with increases in violence related injuries. While the causative route remains ambiguous, it is likely to be due to an association between alcohol consumption at such events, and violence (Matthews et al., 2006).

4.1.7 Individual factors

Individual factors were found to contribute to and predict crimes occurring. The review found alcohol, mental health, drug use and past victimisation were all associated with police demand.

Generally, alcohol outlets and having an alcohol dependency were associated with an increase in person crimes, property crimes, violent offences, and gang membership. Further, as the price of alcohol increased, there were fewer rates of violence related injuries associated. The author concluded that increases in opportunity to consume alcohol could be causing the rise in violence but rises in the price of beer has the opposite influence (Matthews et al., 2006).

Some studies which explored mental health found it was associated with violent offences and behaviours such as gang membership and weapon carrying/use. Drug dependency was also associated with gang membership, but not violent crimes. Coid et al. (2013) described the finding as unsurprising, with consideration to the large number of gang members engaging with drug economy. Past victimisation was a predictor of weapon carrying compared to people who had not been the victim of violence before. Brennan (2019) indicated 2 fundamental explanations for why someone might carry a weapon which were the expectation of being either a victim, or a perpetrator, of violence. Therefore, it is plausible that an individual who is a previous victim of violence carries a weapon for concerns about safety.

4.1.8 COVID-19

A vast amount of academic research was conducted throughout the pandemic.

Lockdowns, due to COVID-19, have shown to be associated with a reduction in crime types such as burglary, robbery, and theft, but increases in other crime types such as cybercrime, fraud, and domestic abuse, as well as increases in individual’s drinking habits.

It is likely that decreases in crimes such as burglary, theft and robbery reflect reduced movement outside the home during lockdown. As there was a stay-at-home order, homes were more likely to be occupied, which could act as a deterrent to burglars (Kirchmaier & Villa-Llera, 2020). In addition, decreased interaction with people could have reduced opportunities for theft. However, this reduction in mobility resulted in increased interaction with close family members which could have created greater opportunities for domestic abuse (Dixon et al., 2022), as well as a greater opportunity to drink alcohol (Jackson et al., 2021).

Conversely, it is likely that increases in crimes such as cybercrime and fraud reflect the increase in virtual mobility, with criminals taking advantage of the vulnerabilities of home working and schooling (Dixon et al., 2022).

In addition, lockdowns may have caused an increase in poverty and unemployment due to business closure and firms hampered to pay their employees (Su et al., 2021). COVID-19 has impacted the world of work negatively regarding reduced working hours and employment losses (Su et al., 2021).

COVID-19 caused an increase in unemployment, which led to a subsequent rise in other crimes such as antisocial behaviour and drug offences. Kirchmaier & Villa-Llera (2020) interpreted these findings by suggesting that the way criminals engage in crime, and the way police respond, has changed during the pandemic. As individuals are not resuming their engagement with legal activities because of the lockdowns, it could increase the economic pressure which drives people to commit crimes. While crimes such as burglary are more difficult to commit, this has subsequently shown increases in antisocial behaviour and drug offences. However, due to increased street visibility during the pandemic, this could explain the high number of arrests in drug-related crimes.

4.2 Evidence informed recommendations

4.2.1 Gaps in literature

The current rapid evidence review covered a vast range of research topics. However, despite many studies investigating police demand, due to the inclusion criteria, they may have been rated as beyond scope for this review. During the search for some demands, it was clear that there were evidence gaps in the literature, with limited or not up-to-date research.

For crime demands, most of the research exploring possession of weapons used United States (US) data. This limited the number of papers that met inclusion criteria in the current review. It would be difficult to generalise any US data to the UK due to different laws on weapon carrying and individual lifestyles. Therefore, to understand the predictors of this demand, further UK exploration would be beneficial.

Many gaps in the literature refer to non-crime incident demands, particularly regarding public safety and welfare and RTAs. For example, research focusing on missing persons was limited, with only one study meeting the inclusion criteria. Some research discussed using DNA to identify missing persons (Budowle et al., 2011), and the behavioural tendencies of adult’s when missing (Bonny et al., 2016), but there was no research which passed the inclusion criteria to directly highlight predictors, if any, of missing person demand on the police.

There was also limited research investigating the predictors of police demand for domestic violence, with only COVID-19 highlighted as a predictor. Due to the research identified relying on reported rates, it is likely that research into domestic violence is limited due to it being a ‘hidden crime’ that is often not reported to police. However, recent increases in new domestic violence cases, due to the pandemic, have resulted in individuals being more likely to report it (Butt, 2020). These increases could allow for more in-depth research to be conducted, exploring possible predictors. Further research is also needed to see if these trends persist beyond COVID-19.

Mental health could be considered both a predictor of police demand, in terms of predicting the number of crimes occurring, and a police demand by itself, in terms of the police demand in dealing with incidents involving concerns for an individual’s mental health. However, there was limited evidence which passed the inclusion criteria for this review which discussed mental health as a predictor. Whilst it’s known through police incident data, that mental health concerns can be demanding on police time through ‘call outs’ and wellbeing checks, the existing literature was not of sufficient relevance for the purpose of this review, which explored the predictors of police demand and not the predictors of mental health.

For example, previous UK research has recognised deprivation (Wickham et al., 2014), antenatal anxiety (Prady et al., 2013) and economic inactivity (Fone et al., 2007) as factors which were found to link to causing mental health distress in the future. Whilst indirectly there is a link through these factors, there is little evidence to support a direct link of mental health predicting demands on the police.

Predictors of RTAs were explored; however, most of the included literature was not conducted between 2011 to 2021. This research was included in the current review because it is assumed that road characteristics and structures will not have altered the demands on police significantly in more recent years, and therefore the evidence was considered to be relevant. However, there is a need for studies investigating RTA demand to be updated.

To summarise, research in the current review clusters around exploring predictors of crime demand but is limited in exploring the predictors of non-crime incident demand (College of Policing, 2015). This is despite non-crime incidents dominating police demand and resource (Hadjipavlou et al., 2018). Therefore, it is recommended that the predictors of non-crime incident demand, such as missing persons, mental health, and RTAs, are included in the focus of future research.

4.2.2 Long term effects of COVID-19

Although trends in crime rates due to the COVID-19 lockdowns have been extensively researched, the longevity of these changes are not known, and may not be known for some time. The only current long-term suggestions are based on theory and evidence on crime displacement, which indicates that longer-term adaptation to other crime types will be the exception rather than the norm (Guerette & Bowers, 2009). This suggests that the prospect of long-term lifestyle changes, such as home working and online shopping, may reduce daytime city populations, which could subsequently keep residential crimes and theft below pre-COVID levels (Langton, Dixon & Farrell, 2021).

It is important that the long-term effects of COVID-19 are a subject of future research, as this may influence strategic planning and police resource allocation. For example, there may be a need for increases in resource in specialist domestic abuse units (Nix & Richards, 2021), and more research evaluating crime prevention initiatives for cybercrime (Brewer et al., 2019).

4.2.3 Policy implication

This review has explored multiple predictors of police demand and has conveyed a large body of research that informs knowledge of a wide range of policing demands. To help understand the demands of the police over different geographies, it is necessary to model predictors of crime and non-crime incident demands on police. Predictors of crime and non-crime incidents are needed as independent variables to explain or predict changes in demand. Statistical models could then be used to predict likely demand at various levels of geography. Therefore, the current review can be used to guide the selection of independent (predictor) variables that might affect or be associated with changes in police demand.

5. Conclusion

This rapid evidence review aimed to provide a summary of predictors that are associated with police crime and non-crime incident demands. Exploring the predictors of police demand is important to understand changes in demand patterns and provide an evidence base for informing future policing decisions.

Using an in-depth and critical methodology, the review searched for papers which investigated predictor-demand relationships that were; UK based, included a large sample, conducted analysis of contemporary data, were conducted between 2011 to 2021 and used a regression analysis. Papers that passed the inclusion criteria were critically appraised based on the quality of analysis for each predictor-demand association.

There were 49 predictors, and 27 separate police demands. Findings were generally mixed for many predictors regarding their associations with crime and non-crime incident demands. Those predictors that demonstrated demands more consistently included increases in population and traffic being associated with increases in crime and RTAs respectively. Events with a political element or with a terrorist influence were associated with increases in hate crimes. Predictors such as individual factors (alcohol, mental health, drug use and past victimisation), family structure, gender, deviant youth behaviour and weak social bonds were associated with increases in violent behaviours (for example, weapon carrying/use, antisocial behaviour, gang membership).

Increases in predictors such as deprivation, education, income inequality, vacant properties and public amenities were associated with increases in crime types such as burglary, robbery, and theft, whereas increases in sanction rates and prison population in an area were associated with decreases in many crime types. Finally, lockdowns from COVID-19 saw shifts in crime trends for burglary, robbery, domestic violence, and cybercrime.

It was evident from this review that while crime demands were generally well researched, future research should also focus on non-crime incident demands, given that these demands dominate police resource. Future research also needs to focus on the long-term effect of COVID-19 to understand future changes in police resource allocation. This review will be important for supporting and informing future resourcing decisions that relate to predictors of crime and non-crime demands.

References

Alsharkas, Z. & Campaniello, N. (2019) Analyse the relationship between unemployment and crime. (Publication No. 231751958) [Undergraduate dissertation, University of Essex] Retrieved from: Analyse the relationship between unemployment and crime . EC 831-6-FY Project in Economics Semantic Scholar

Andrews, R. (2011). Exploring the impact of arson-reduction strategies: Panel data evidence from England. The British Journal of Criminology, 51(5), 839-855.

Aston, E. V. (2015). Are risk factors for drug use and offending similar during the teenage years?. International Journal of Drug Policy, 26(4), 396-403.

Bandyopadhyay, S., Bhattacharya, S., & Sensarma, R. (2011). An analysis of the factors determining crime in England and Wales: A quantile regression approach. Available at SSRN 1866812.

Bell, B., & Machin, S. (2011). The impact of migration on crime and victimisation. London: UK.

Bell, B., Fasani, F., & Machin, S. (2013). Crime and immigration: Evidence from large immigrant waves. Review of Economics and statistics, 95(4), 1278-1290.

Bondy, M., Roth, S., & Sager, L. (2020). Crime is in the air: The contemporaneous relationship between air pollution and crime. Journal of the Association of Environmental and Resource Economists, 7(3), 555-585.

Bonny, E., Almond, L., & Woolnough, P. (2016). Adult missing persons: Can an investigative framework be generated using behavioural themes?. Journal of investigative psychology and offender profiling, 13(3), 296-312.

Bowers, K. (2014). Risky facilities: Crime radiators or crime absorbers? A comparison of internal and external levels of theft. Journal of Quantitative Criminology, 30(3), 389-414.

Brennan, I. (2019). Victims of serious violence in England and Wales, 2011–2017.

Brewer, M., & Tasseva, I. V. (2020). Did the UK policy response to COVID-19 protect household incomes?. The Journal of Economic Inequality, 19(3), 433-458.

Brewer, R., de Vel-Palumbo, M., Hutchings, A., Holt, T., Goldsmith, A., & Maimon, D. (2019). Cybercrime prevention: Theory and applications. Springer Nature.

Brosnan, S. (2019). The impact of sports participation on crime in England between 2012 and 2015. Sport in Society, 23(6), 1080-1090.

Budowle, B., Ge, J., Chakraborty, R., & Gill-King, H. (2011). Use of prior odds for missing persons identifications. Investigative genetics, 2(1), 1-6.

Buil-Gil, D., Miró-Llinares, F., Moneva, A., Kemp, S., & Díaz-Castaño, N. (2021a). Cybercrime and shifts in opportunities during COVID-19: a preliminary analysis in the UK. European Societies, 23(sup1), S47-S59.

Buil-Gil, D., Medina, J., & Shlomo, N. (2021b). Measuring the dark figure of crime in geographic areas: Small area estimation from the crime survey for England and Wales. The British journal of criminology, 61(2), 364-388.

Buil-Gil, D., & Zeng, Y. (2021c). Meeting you was a fake: investigating the increase in romance fraud during COVID-19. Journal of Financial Crime.

Butt, E. (2020) Know economic abuse. 2020 Report. Refuge. For women and children. Against domestic violence. Retrieved from: Know-Economic-Abuse-Report-2020.pdf (refuge.org.uk)

Carl, N. (2017). Ethnicity and electoral fraud in Britain. Electoral Studies, 50, 128-136.

Carr, J., Clifton-Sprigg, J., James, J., & Vujic, S. (2020). Love thy neighbour? Brexit and hate crime (No. 13902). IZA Discussion Papers.

Cohen, L. E., & Felson, M. (1979). Social change and crime rate trends: A routine activity approach. American sociological review, 588-608.

Coid, J. W., Ullrich, S., Keers, R., Bebbington, P., DeStavola, B. L., Kallis, C., … & Donnelly, P. (2013). Gang membership, violence, and psychiatric morbidity. American journal of psychiatry, 170(9), 985-993.

College of Policing (2015). College of Policing analysis: Estimating demand on the police service. College of Policing. Retrieved from: [College of Policing analysis: Estimating demand on the police service (https://assets.college.police.uk/s3fs-public/2021-03/demand-on-policing-report.pdf).

Daly, M., & Robinson, E. (2021). High-risk drinking in midlife before versus during the COVID-19 crisis: Longitudinal evidence from the United Kingdom. American Journal of Preventive Medicine, 60(2), 294-297.

Davies, T., & Johnson, S. D. (2014). Examining the relationship between road structure and burglary risk via quantitative network analysis. Journal of Quantitative Criminology, 31(3), 481-507.

Dean, N., Dong, G., Piekut, A., & Pryce, G. (2019). Frontiers in residential segregation: understanding neighbourhood boundaries and their impacts. Tijdschrift voor economische en sociale geografie, 110(3), 271-288.

Devine, D. (2018). The UK referendum on membership of the European Union as a trigger event for hate crimes. Available at SSRN 3118190.

Dixon, A. et al. (2022). ‘Nearly Two Years In: Pandemic Crime Trends in England and Wales to end of 2021’ Statistical Bulletin on Crime and COVID-19, issue 16, University of Leeds.

Evans, S., Alkan, E., Bhangoo, J. K., Tenenbaum, H., & Ng-Knight, T. (2021). Effects of the COVID-19 lockdown on mental health, wellbeing, sleep, and alcohol use in a UK student sample. Psychiatry research, 298, 113819.

Farrington, D. P. (1983). Offending from 10 to 25 years of age. In Prospective studies of crime and delinquency (pp. 17-37). Springer, Dordrecht.

Fitzsimons, E., Jackman, J., Kyprianides, A., & Villadsen, A. (2018). Determinants of risky behaviour in adolescence: Evidence from the UK. Centre for Longitudinal Studies (CLS), London.

Fone, D., Dunstan, F., Williams, G., Lloyd, K., & Palmer, S. (2007). Places, people and mental health: a multilevel analysis of economic inactivity. Social science & medicine, 64(3), 633-645.

GOV.UK (2011). National standard for incident recording counting rules Home Office. Retrieved from: National standard for incident recording counting rules - GOV.UK (www.gov.uk)

GOV.UK (2013b). Police recorded crime and outcome open data tables. Home Office. Retrieved from: Police recorded crime and outcomes open data tables - GOV.UK (www.gov.uk)

GOV.UK (2017). Rapid evidence assessments. Department of International Development. Retrieved from: Rapid evidence assessments - GOV.UK (www.gov.uk)

GOV.UK (2018). Socioeconomic status. Retrieved from: Socioeconomic status - GOV.UK Ethnicity facts and figures (ethnicity-facts-figures.service.gov.uk)

GOV.UK (2020). Women and the Criminal Justice System, 2019. Ministry of Justice. Retrieved from: Women and the Criminal Justice System, 2019 - GOV.UK (www.gov.uk)

GOV.UK (2021). Counting rules for recorded crime. Home Office. Retrieved from: Counting rules for recorded crime - GOV.UK (www.gov.uk)

Grant, M., Booth, A. (2009). A typology of reviews: an analysis of 14 review types and associated methodologies. Health Information and Libraries Journal, 26, 91-108.

Gray, C., & Hansen, K. (2021). Did COVID-19 Lead to an Increase in Hate Crimes Toward Chinese People in London?. Journal of Contemporary Criminal Justice, 37(4), 569-588.

Guerette, R. T., & Bowers, K. J. (2009). Assessing the extent of crime displacement and diffusion of benefits: A review of situational crime prevention evaluations. Criminology, 47(4), 1331-1368.

Gulma, U. L., Evans, A., Heppenstall, A., & Malleson, N. (2018). Diversity and burglary: Do community differences matter?. Transactions in GIS, 23(2), 181-202.

Hadjipavlou, Redgrave and Desroches (2018). Rethinking Police Demand. Crest Advisory. Retrieved from: Crest-Rethinking-Police-Demand-2.indd (wixstatic.com)

Halford, E., Dixon, A., Farrell, G., Malleson, N., & Tilley, N. (2020). Crime and coronavirus: Social distancing, lockdown, and the mobility elasticity of crime. Crime science, 9(1), 1-12.

Han, L., Bandyopadhyay, S., & Bhattacharya, S. (2013). Determinants of violent and property crimes in England and Wales: A panel data analysis. Applied Economics, 45(34), 4820-4830.

Haynes, R., Jones, A., Kennedy, V., Harvey, I., & Jewell, T. (2007). District variations in road curvature in England and Wales and their association with road-traffic crashes. Environment and planning A, 39(5), 1222-1237.

Hels, T. and Orozova-Bekkevold, I., 2007, The effect of roundabout design features on cyclist accident rate, Accident Analysis & Prevention, 39 (2), 300-307.

Hutchings, E., Browne, K. D., Chou, S., & Wade, K. (2019). Repeat missing child reports in Wales. Child abuse & neglect, 88, 107-117.

Ingram, J., Maciejewski, G., & Hand, C. J. (2020). Changes in diet, sleep, and physical activity are associated with differences in negative mood during COVID-19 lockdown. Frontiers in psychology, 2328.

Ivandic, R., Kirchmaier, T., & Linton, B. (2020). Changing patterns of domestic abuse during COVID-19 lockdown. CEP Discussion Papers (1729). Centre for Economic Performance, London School of Economics and Political Science, London, UK.

Ivandic, R., Kirchmaier, T., & Machin, S. J. (2019). Jihadi attacks, media and local hate crime.

Jackson, S. E., Garnett, C., Shahab, L., Oldham, M., & Brown, J. (2021). Association of the COVID‐19 lockdown with smoking, drinking and attempts to quit in England: an analysis of 2019–20 data. Addiction, 116(5), 1233-1244.

Jaitman, L., & Machin, S. (2013). Crime and immigration: new evidence from England and Wales. IZA Journal of Migration, 2(1), 1-23.

Jennings, C. R. (2013). Social and economic characteristics as determinants of residential fire risk in urban neighborhoods: A review of the literature. Fire Safety Journal, 62, 13–19.

Jones, A. P., Haynes, R., Kennedy, V., Harvey, I. M., Jewell, T., & Lea, D. (2008). Geographical variations in mortality and morbidity from road traffic accidents in England and Wales. Health & place, 14(3), 519-535.

Kawalerowicz, J., & Biggs, M. (2015). Anarchy in the UK: Economic deprivation, social disorganization, and political grievances in the London Riot of 2011. Social Forces, 94(2), 673-698.

Khangura, S., Konnyu, K., Cushman, R., Grimshaw, J., & Moher, D. (2012). Evidence summaries: the evolution of a rapid review approach. Systematic reviews, 1(1), 1-9.

Kirchmaier, T., & Villa-Llera, C. (2020). COVID-19 and changing crime trends in England and Wales. Centre for Economic Performance, (013).

Langton, S., Dixon, A., & Farrell, G. (2021). Six months in: pandemic crime trends in England and Wales. Crime science, 10(1), 1-16.

Levi M & Smith R. (2021). Fraud and its relationship to pandemics and economic crises: From Spanish flu to COVID-19. Research Report no. 19. Canberra: Australian Institute of Criminology.

Livingston, M., Kearns, A., & Bannister, J. (2014). Neighbourhood structures and crime: The influence of tenure mix and other structural factors upon local crime rates. Housing Studies, 29(1), 1-25.

MacArthur, G. J., Smith, M. C., Melotti, R., Heron, J., Macleod, J., Hickman, M., … & Lewis, G. (2012). Patterns of alcohol use and multiple risk behaviour by gender during early and late adolescence: the ALSPAC cohort. Journal of public health, 34(suppl_1), i20-i30.

Malleson, N., & Andresen, M. A. (2016). Exploring the impact of ambient population measures on London crime hotspots. Journal of Criminal Justice, 46, 52-63.

Matthews, K., Shepherd, J., & Sivarajasingham, V. (2006). Violence-related injury and the price of beer in England and Wales. Applied Economics, 38(6), 661-670.

Mawby RI. (2016). Policing policies in cities dependent on tourism: Developing an evidence-based approach. International Journal of Police Science & Management. 2016;18(3):195-205.

Mburu, L. W., & Helbich, M. (2016). Crime risk estimation with a commuter-harmonized ambient population. Annals of the American Association of Geographers, 106(4), 804-818.

Neanidis, K. C., & Rana, M. P. (2021). Crime in the era of COVID-19: Evidence from England. Available at SSRN 3784821.

Newton, A., Partridge, H., & Gill, A. (2014). In and around: Identifying predictors of theft within and near to major mass underground transit systems. In Safety and Security in Transit Environments (pp. 99-115). Palgrave Macmillan, London.

Nivette, A. E., Zahnow, R., Aguilar, R., Ahven, A., Amram, S., Ariel, B., … & Eisner, M. P. (2021). A global analysis of the impact of COVID-19 stay-at-home restrictions on crime. Nature Human Behaviour, 5(7), 868-877.

Nix, J., & Richards, T. N. (2021). The immediate and long-term effects of COVID-19 stay-at-home orders on domestic violence calls for service across six US jurisdictions. Police practice and research, 22(4), 1443-1451.

O’Brien, K. S., Forrest, W., Greenlees, I., Rhind, D., Jowett, S., Pinsky, I., … & Iqbal, M. (2018). Alcohol consumption, masculinity, and alcohol-related violence and anti-social behaviour in sportspeople. Journal of science and medicine in sport, 21(4), 335-341.

ONS.GOV.UK (2020). Crime in England and Wales: year ending June 2020. Office of National Statistics. Retrieved from: Crime in England and Wales - Office for National Statistics (ons.gov.uk)

ONS.GOV.UK (2021). Crime in England and Wales: year ending June 2021. Office of National Statistics. Retrieved from: Crime in England and Wales - Office for National Statistics (ons.gov.uk)

Page, N. (2015). Investigating trends and determinants of violence-related injury in England and Wales (Doctoral dissertation, Cardiff University).

Piatkowska, S. J., & Stults, B. J. (2021). Brexit, Terrorist Attacks, and Hate Crime: A Longitudinal Analysis. Social Problems.

Prady, S. L., Pickett, K. E., Croudace, T., Fairley, L., Bloor, K., Gilbody, S., … & Wright, J. (2013). Psychological distress during pregnancy in a multi-ethnic community: findings from the born in Bradford cohort study. PloS one, 8(4), e60693.

Quddus, M. A. (2008). Modelling area-wide count outcomes with spatial correlation and heterogeneity: An analysis of London crash data. Accident Analysis & Prevention, 40(4), 1486-1497.

Quick, M., Li, G., & Brunton-Smith, I. (2018). Crime-general and crime-specific spatial patterns: A multivariate spatial analysis of four crime types at the small-area scale. Journal of Criminal Justice, 58, 22-32.

Rao, R., Mueller, C., & Broadbent, M. (2021). Risky alcohol consumption in older people before and during the COVID-19 pandemic in the United Kingdom. Journal of Substance Use, 1-6.

Requia, W. J., Mohamed, M., Higgins, C. D., Arain, A., & Ferguson, M. (2018). How clean are electric vehicles? Evidence-based review of the effects of electric mobility on air pollutants, greenhouse gas emissions and human health. Atmospheric Environment, 185, 64-77.

Ristea, A., Kurland, J., Resch, B., Leitner, M., & Langford, C. (2018). Estimating the spatial distribution of crime events around a football stadium from georeferenced tweets. ISPRS International Journal of Geo-Information, 7(2), 43.

Sager, L. (2016). Estimating the effect of air pollution on road safety using atmospheric temperature (No. 251). Grantham Research Institute on Climate Change and the Environment.

Sarstedt, M., & Mooi, E. (2019). Regression analysis. In A Concise Guide to Market Research (pp. 209-256). Springer, Berlin, Heidelberg.

Shaw, C. R., & McKay, H. D. (1942). Juvenile delinquency and urban areas. University of Chicago Press.

Shen, J., Wang, T., Zheng, C., & Yu, M. (2020). Determinants of bicyclist injury severity resulting from crashes at roundabouts, crossroads, and T-junctions. Journal of advanced transportation, 2020.

Shepherd, J. (1990). Violent crime in Bristol: an accident and emergency department perspective. The British Journal of Criminology, 30(3), 289-305.

Smith, V., & Wynne-McHardy, E. (2019a/b). An Analysis of Indicators of Serious Violence, Findings from the Millennium Cohort Study and the Environmental Risk (E-Risk) Longitudinal Twin Study (Vol. 1). Research Report 110, Home Office. The World Bank, (2019) Nigeria: Overview, https://www. worldbank. org/en/country/nigeria/overview.

Steinberg, L., & Monahan, K. C. (2008). Age differences in resistance to peer influence. Developmental psychology, 43(6), 1531.

Su, C. W., Dai, K., Ullah, S., & Andlib, Z. (2021). COVID-19 pandemic and unemployment dynamics in European economies. Economic Research-Ekonomska Istraživanja, 1-13.

Sun, Y., Huang, Y., Yuan, K., Chan, T. O., & Wang, Y. (2021). Spatial patterns of COVID-19 incidence in relation to crime rate across London. ISPRS International Journal of Geo-Information, 10(2), 53.

Tarling, R., & Dennis, R. (2016). Socio‐Economic Determinants of Crime Rates: Modelling Local Area Police‐Recorded Crime. The Howard Journal of Crime and Justice, 55(1-2), 207-225.

Tilley, N., Pease, K., Hough, M., & Brown, R. (1999). Burglary prevention: early lessons from the crime reduction programme, PRCU Research Paper 1. Home Office Research, Development and Statistics Directorate.

Wang, C., Quddus, M., & Ison, S. (2009). The effects of area-wide road speed and curvature on traffic casualties in England. Journal of transport geography, 17(5), 385-395.

Whitworth, A. (2012). Inequality and Crime across England: A Multilevel Modelling Approach. Social Policy and Society, 11, pp 27-40.

Whitworth, A. (2013). Local Inequality and Crime: Exploring how Variation in the Scale of Inequality Measures Affects Relationships between Inequality and Crime. Urban Studies. 2013;50(4):725-741.

Wickham, S., Taylor, P., Shevlin, M., & Bentall, R. P. (2014). The impact of social deprivation on paranoia, hallucinations, mania and depression: the role of discrimination social support, stress and trust. PloS one, 9(8), e105140.

Williams, M., Burnap, P., Javed, A., Liu, H., Ozalp, S. (2019). Hate in the Machine: Anti-Black and Anti-Muslim Social Media Posts as Predictors of Offline Racially and Religiously Aggravated Crime. The British Journal of Criminology, 60(1), 93-117.

Wu, D., & Wu, Z. (2012) Crime, inequality and unemployment in England and Wales, Applied Economics, 44:29, 3765-3775.

Zeng, Q., & Huang, H. (2014). Bayesian spatial joint modeling of traffic crashes on an urban road network. Accident Analysis & Prevention, 67, 105-112.

Appendix

Table A1: High-level groupings, the 41 predictor headings, and the additional predictors (equalling 49 predictors) that were encapsulated in the predictor headings

Higher-level grouping	Predictor heading	Predictors encapsulated in predictor heading
Socioeconomic Factors	Deprivation Unemployment Education Income Inequality Socio-demographic (Family Structure) Socio-demographic (Socio-Economic Status) Affluent Area	Female Lone Parent
Population	Population Density Population Churn Total Population Immigration Prison Population Population at Risk
Traffic Environment	Traffic Volume Road Length Road Layout Air Pollution
Demographics	Gender Young People (15 to 24 years of age) Young People (16 to 29 years of age) Ethnic Composition Young People (10 to 15 years of age) Other Age Groups Deviant Youth Behaviour	Youth Conceptions Peer Group Informal Social Control
Geographic Location	Average House Price Police Force Factors Land Use Housing Mix Seasonal Effects Vacant Properties Public Amenities	Sanctions Detection Rate Student Households Weather/Climate
Events	Political Events Terror Attacks Major Sporting Events	Stadium Events
Individual Factors	Alcohol Mental Health Drug Use Past Victims
COVID-19	COVID-19 Effect on Demands COVID-19 Effect on Predictors Unemployment due to COVID-19

Appendix A2: Higher-level groupings, predictor headings and associated authors

Socioeconomic factors

Deprivation: Brennan, 2019; Carl, 2017; Haynes et al., 2007; Jones et al., 2008; Kawalerowicz & Biggs, 2015; Livingston et al., 2014; Mburu & Helbich, 2016; Page, 2015; Tarling & Dennis, 2016; Quddus, 2008; Quick et al., 2018.

Unemployment: Alsharkas & Campaniello, 2019; Andrews, 2011; Bandyopadhyay et al., 2011; Buil-Gil et al., 2021b; Brosnan, 2019; Davies & Johnson, 2014; Gulma et al., 2018; Han et al., 2013; Matthews et al., 2006; Page, 2015; Quddus, 2008; Tarling & Dennis, 2016; Wang et al., 2009; Whitworth, 2012; Whitworth, 2013; Williams et al., 2019; Wu & Wu, 2012.

Education: Alsharkas & Campaniello, 2019; Buil-Gil et al., 2021b; Gulma et al., 2018; Smith & Wynne-McHardy, 2019b; Whitworth, 2012; Whitworth, 2013; Williams et al., 2019.

Income inequality: Alsharkas & Campaniello, 2019; Bandyopadhyay et al., 2011; Brosnan, 2019; Buil-Gil et al., 2021b; Han et al., 2013; Page, 2015; Smith & Wynne-McHardy, 2019a; Whitworth, 2012; Whitworth, 2013; Wu & Wu, 2012.

Socio-demographic (Family Structure): Aston, 2015; Livingston et al., 2014; Smith & Wynne-McHardy, 2019a; Smith & Wynne-McHardy, 2019b.

Socio-demographic (SES): Aston, 2015; Smith & Wynne-McHardy, 2019a.

Affluent area: Mburu & Helbich, 2016.

Population

Population density: Buil-Gil et al., 2021b; Davies & Johnson, 2014; Kawalerowicz & Biggs, 2015; Tarling & Dennis, 2016; Quick et al., 2018; Whitworth, 2012; Whitworth, 2013; Wu & Wu, 2012.

Population churn: Andrews, 2011; Tarling & Dennis, 2016; Quick et al., 2018; Whitworth, 2012; Whitworth, 2013.

Total population: Andrews, 2011; Davies & Johnson, 2014; Malleson & Andresen, 2016; Whitworth, 2012; Wang et al., 2009.

Immigration: Bell & Machin, 2011; Bell et al., 2013; Buil-Gil et al., 2021b; Dean et al., 2019; Jaitman & Machin, 2013.

Prison population: Han et al., 2013.

Population at risk: Haynes et al., 2007.

Traffic environment

Traffic; air pollution: Bondy et al., 2020; Davies & Johnson, 2014; Haynes et al., 2007; Jones et al., 2008; Quddus, 2008; Sager, 2016; Wang et al., 2009.

Demographics

Gender: Alsharkas & Campaniello, 2019; Aston, 2015; Brennan, 2019; Brosnan, 2019; Buil-Gil et al., 2021b; Fitzsimons et al., 2018; Smith & Wynne-McHardy, 2019a; Smith & Wynne-McHardy, 2019b.

Young people (15 to 24): Bandyopadhyay et al., 2011; Han et al., 2013; Kawalerowicz & Biggs, 2015; Tarling & Dennis, 2016; Williams et al., 2019.

Young people (16 to 29): Livingston et al., 2014; Whitworth, 2012; Whitworth, 2013.

Ethnic composition: Brennan, 2019; Buil-Gil et al., 2021; Carl, 2017; Davies & Johnson, 2014; Dean et al., 2019; Fitzsimons et al., 2018; Gulma et al., 2018; Kawalerowicz & Biggs, 2015; Livingston et al., 2014; Matthews et al., 2006; Mburu & Helbich, 2016; Quick et al., 2018; Smith & Wynne-McHardy, 2019a; Tarling & Dennis, 2016; Whitworth, 2012; Whitworth, 2013; Williams et al., 2019.

Other age groups: Brennan, 2019; Gulma et al., 2018; Hutchings et al., 2019; O’Brien et al., 2018; Quddus, 2008.

Young people (10 to 15): Davies & Johnson, 2014; Fitzsimons et al., 2018; Livingston et al., 2014; Smith & Wynne-McHardy, 2019a.

Deviant youth behaviour: Aston, 2015; Brennan, 2019; Smith & Wynne-McHardy, 2019a; Smith & Wynne-McHardy, 2019b; Whitworth, 2012.

Geographic location

Average house price: Buil-Gil et al., 2021b; Matthews et al., 2006; Tarling & Dennis, 2016; Whitworth, 2012; Whitworth, 2013.

Police force factors: Bandyopadhyay et al., 2011; Whitworth, 2012; Wu & Wu, 2012.

Housing mix: Davies & Johnson, 2014; Livingston et al., 2014; Smith & Wynne-McHardy, 2019a.

Land use: Bowers, 2014; Buil-Gil et al., 2021b; Haynes et al., 2007; Kawalerowicz & Biggs, 2015; Matthews et al., 2006; Mburu & Helbich, 2016; Page, 2015.

Vacant properties: Davies & Johnson, 2014; Mburu & Helbich, 2016.

Public amenities: Mburu & Helbich, 2016; Newton et al., 2014.

Seasonal effects: Matthews et al., 2006; Page, 2015; Sager, 2016.

Events

Political events: Carr et al., 2020; Devine, 2018; Piatkowska & Stults, 2021.

Terror attacks: Devine, 2018; Ivandic et al., 2019; Piatkowska & Stults, 2021.

Major sporting events: Matthews et al., 2006; Ristea et al., 2018.

Individual factors

Alcohol: Coid et al., 2013; O’Brien et al., 2018.

Mental health: Coid et al., 2013; Smith & Wynne-McHardy, 2019a; Smith & Wynne-McHardy, 2019b.

Drug use: Brennan, 2019; Coid et al., 2013; Smith & Wynne-McHardy, 2019a; Smith & Wynne-McHardy, 2019b.

Past victim: Brennan, 2019; Smith & Wynne-McHardy, 2019a.

COVID-19

COVID-19 effect on demands: Buil-Gil et al., 2021a; Buil-Gil et al., 2021c; Gray & Hansen, 2021; Halford et al., 2020; Ivandic et al., 2020; Neanidis & Rana, 2021; Nivette et al., 2021; Sun et al., 2021.

COVID-19 effect on predictors: Brewer & Tasseva, 2020; Daly & Robinson, 2021; Evans et al., 2021; Ingram et al., 2020; Jackson et al., 2021; Rao et al., 2021; Su et al., 2021.

Unemployment due to COVID-19: Kirchmaier & Villa-Llera, 2020.

Table A3: Summary of socio-economic predictor-demand associations

Demand	Deprivation	Unemployment	Education	Income inequality	Socio-demo (family structure)	Socio-demo (SES)	Affluent area
Arson		x
Burglary	x	x	x	x
Crime rate
Criminal damage		x		x
Cybercrime
Dark figure		x	x	x
Drugs		x		x	x	x
Fraud	x	x		x
Hate crime		x	x
Miscellaneous		x		x
Possession of weapons	x				x	x
Property crime	x	x	x	x	x
Robbery	x	x	x	x
Sexual offences		x		x
Theft	x	x		x			x
Vehicle crimes		x	x	x
Violence against the person	x	x	x	x	x
Violent crime	x	x	x	x	x
All crime	x				x
ASB
Domestic abuse
Gangs
Missing persons
RTAs	x	x
Alcohol (COVID-19)
Unemployment (COVID-19)
Poverty (COVID-19)

Table A4: Summary of population predictor-demand associations

Demand	Population density	Population churn	Total population	Immigration	Prison population	Population at risk
Arson		x	x
Burglary	x	x	x	x	x
Crime rate
Criminal damage	x	x
Cybercrime
Dark figure
Drugs	x
Fraud				x
Hate crime
Miscellaneous
Possession of weapons				x
Property crime	x	x		x
Robbery	x	x	x		x
Sexual offences	x				x
Theft	x		x	x
Vehicle crimes	x	x	x	x
Violence against the person				x	x
Violent crime	x	x	x	x
All crime				x
ASB
Domestic abuse
Gangs
Missing persons
RTAs			x	x		x
Alcohol (COVID-19)
Unemployment (COVID-19)
Poverty (COVID-19)

Table A5: Summary of traffic environment predictor-demand associations

Demand	Volume of traffic	Road length	Road layout	Air pollution
Arson
Burglary			x
Crime rate				x
Criminal damage				x
Cybercrime
Dark figure
Drugs
Fraud
Hate crime
Miscellaneous
Possession of weapons
Property crime
Robbery
Sexual offences
Theft
Vehicle crimes
Violence against the person
Violent crime				x
All crime
ASB
Domestic abuse
Gangs
Missing persons
RTAs	x	x	x	x
Alcohol (COVID-19)
Unemployment (COVID-19)
Poverty (COVID-19)

Table A6: Summary of demographic predictor-demand associations

Demand	Gender	Young people (15 to 24 years)	Young people (16 to 29 years)	Ethnic composition	Young people (10 to 15 years)	Other age groups	Deviant youth behaviour
Arson
Burglary		x	x	x	x	x	x
Crime rate				x
Criminal damage		x	x	x			x
Cybercrime	x
Dark figure				x
Drugs	x						x
Fraud				x
Hate crime
Miscellaneous
Possession of weapons	x						x
Property crime	x			x
Robbery		x	x	x			x
Sexual offences
Theft				x
Vehicle crimes		x	x	x			x
Violence against the person	x			x	x
Violent crime		x	x	x			x
All crime				x
ASB	x			x
Domestic abuse
Gangs
Missing persons						x
RTAs						x
Alcohol (COVID-19)
Unemployment (COVID-19)
Poverty (COVID-19)

Table A7: Summary of geographic predictor-demand associations

Demand	Average house price	Police force factors	Land use	Housing mix	Seasonal effects	Vacant properties	Public amenities
Arson
Burglary	x	x		x
Crime rate
Criminal damage	x	x
Cybercrime
Dark figure	x		x
Drugs		x
Fraud		x
Hate crime
Miscellaneous		x
Possession of weapons				x
Property crime			x	x
Robbery	x	x
Sexual offences		x
Theft		x	x			x	x
Vehicle crimes	x	x
Violence against the person	x	x	x	x	x
Violent crime	x	x	x		x
All crime			x	x
ASB
Domestic abuse
Gangs
Missing persons
RTAs			x		x
Alcohol (COVID-19)
Unemployment (COVID-19)
Poverty (COVID-19)

Table A8: Summary of event predictor-demand associations

Demand	Political events	Terror attacks	Major sporting events
Arson
Burglary
Crime rate
Criminal damage			x
Cybercrime
Dark figure
Drugs
Fraud
Hate crime	x	x
Miscellaneous
Possession of weapons
Property crime
Robbery
Sexual offences
Theft			x
Vehicle crimes
Violence against the person			x
Violent crime
All crime			x
ASB
Domestic abuse
Gangs
Missing persons
RTAs
Alcohol (COVID-19)
Unemployment (COVID-19)
Poverty (COVID-19)

Table A9: Summary of individual predictor-demand associations

Demand	Alcohol	Mental health	Drug use	Past victims
Arson
Burglary
Crime rate
Criminal damage	x
Cybercrime
Dark figure
Drugs
Fraud
Hate crime
Miscellaneous
Possession of weapons			x	x
Property crime
Robbery
Sexual offences	x
Theft
Vehicle crimes
Violence against the person
Violent crime	x	x
All crime
ASB	x
Domestic abuse
Gangs	x	x	x
Missing persons
RTAs
Alcohol (COVID-19)
Unemployment (COVID-19)
Poverty (COVID-19)

Table A10: Summary of COVID-19 predictor-demand associations

Demand	COVID-19 (effect on demands)	COVID-19 (effect on predictors)	Unemployment (due to COVID-19)
Arson
Burglary	x		x
Crime rate
Criminal damage	x
Cybercrime	x
Dark figure
Drugs	x		x
Fraud	x
Hate crime	x
Miscellaneous
Possession of weapons	x
Property crime
Robbery	x
Sexual offences
Theft	x		x
Vehicle crimes	x		x
Violence against the person	x
Violent crime			x
All crime
ASB	x		x
Domestic abuse	x
Gangs
Missing persons
RTAs
Alcohol (COVID-19)		x
Unemployment (COVID-19)		x
Poverty (COVID-19)		x