The homicide drop in England and Wales 2004–2014

After decades of rising homicide rates in the late 20th century, much of the Western world witnessed a decline in homicide from the early-mid-1990s. In England and Wales, homicide rates defied this trend and continued to rise for a further decade, peaking in 2004 before declining year on year until 2014. The late onset of the decline in England and Wales presents a quandary for dominant explanations of the broader decline, and has yet to be theorised. This article presents a disaggregated analysis of the homicide drop in England and Wales, identifying subtypes of homicide that appear to have driven the decline. The findings indicate changes in lifestyle, routine activities and social/criminal justice policy as the main drivers of the homicide drop, and contribute to international theory on homicide trends.


Introduction
Homicide rates in England and Wales witnessed consecutive increases between 2014 and 2018, 1 reportedly driven by an upturn in fatal knife-attacks involving teenagers (Ellis, 2019;Younge, 2018). This incline interrupted the longer-term decline in homicide observed from the mid-2000s, that forms the basis of this article. The steady decrease in homicide in England and Wales between 2004 and 2014 followed a similar pattern to other parts of the Western world, including the United States, Australia and much of Europe, where the sudden downturn in homicide from the early-mid-1990s has been well-documented and theorised. The decline in England and Wales began around a decade later and has yet to be the subject of rigorous examination. Using data from the national Homicide Index (HI) database, this article presents the first detailed analysis of the homicide drop in England and Wales and contributes to theory on homicide trends on both a national and international level.
The over-arching finding is that in England and Wales, the decline in homicide was observable across many types of homicide, but particularly evident in particular subtypes: homicides involving younger suspects and victims, occurring in public spaces, and involving intoxicated suspects and victims. Geographically, the analysis revealed that the greatest decline occurred in Greater London and Greater Manchester. Although proportionately, numbers of homicides decreased for both men and women, when calculated as rates within the population of England and Wales, the analysis revealed that rates of female suspects and victims remained stable throughout the period of decline, and that the overall decrease was predominantly driven by declines in male homicide suspects and victims. Similarly, although the percentage of homicides by sharp instrument, compared to other methods of killing, have been steadily increasing since the mid-2000s, the rate of homicides involving sharp instruments declined between 2007 and 2014. In other words, despite decreasing of numbers of kniferelated homicides (calculated as rates within the ever-growing population for England and Wales), stabbings have accounted for a growing proportion of illegal deaths for some time.
The article begins by considering dominant approaches to understanding homicide trends that have been used to explain the most recent decline in other countries. These broadly incorporate cultural theories such as Elias' (1994Elias' ( [1939) Civilising Process and Eisner's (2001Eisner's ( , 2008 'conduct of life' theory, and a variety of alternative explanations routed in Felson's (1987) Routine Activity theory, recognising the potential impact of various factors affecting lifestyle and opportunities for fatal violence to occur. Explanations contextualising the homicide drop with concurrent trends in non-lethal violence are also highlighted as essential to understanding changes in homicide rates, as divergent trends in fatal and non-fatal violence demand an alternative narrative to similar patterns in homicide and other forms of violence (Lehti, 2014).
Cross-cultural research is paramount in developing theoretical frameworks around homicide that are country-specific and capable of informing homicide reduction strategies (Lehti, 2014;Soothill and Francis, 2012), and doing so requires disaggregating national-level homicide data. The article achieves this through presenting an analysis of homicide data for England and Wales (covering the period [2003][2004][2005][2006][2007][2008][2009][2010][2011][2012][2013][2014][2015], using data from the national HI database. By observing whether the decline occurred across all types of homicides or in particular subtypes, a more nuanced understanding is developed. The key findings are discussed in context with concurrent trends in non-lethal violence, the recent increase in homicide and dominant theoretical explanations. Rather than pointing towards a general decline in fatal violence associated with traditional Eliasian (cultural pacification) theory, the findings point to alternative explanations, namely, changes in the lifestyle and routine activities of particular groups, policies aimed at reducing particular subtypes of homicide and the broader expansion of security associated with the general crime drop witnessed across the same period of time.

Perspectives on homicide decline
Examining trends in homicide rates is key to explaining homicide. Through observing peaks and troughs in homicide rates, one can correlate changes in particular subtypes of homicide with extraneous factors and develop nuanced understandings of what drivers or conditions facilitate or prevent fatal violence. Following four decades of increasing homicide numbers, the mid-1990s saw a sharp decline in homicide rates in many regions across the Western world, contextualised by a broader 'crime-drop' which for many criminologists, had not been foreseen (Reiner, 2016;Walby et al., 2016).
The unexpected decrease in homicide was most notable across several US cities, which had previously been witnessing unprecedented rises in homicide rates (Blumstein et al., 2000), but also occurred across much of the West, with England and Wales representing somewhat an outlier, as the homicide rate here continued to soar until 2004 (Shaw et al., 2005). Attempts to explain this homicide drop draw upon a range of theoretical perspectives, some of which derive from earlier explanations of the long-term historical decline in homicide observed between the middle ages and mid-20th century, across much of the Western world (Eisner, 2001;Elias, 1994Elias, [1939; Spierenberg, 2012), and others focusing on contemporary changes in lifestyle and policy which may have compromised the 'opportunity' for fatal violence to occur (Aebi and Linde, 2014;Cohen and Felson, 1979;Lehti, 2014).

Cultural approaches
Cultural theories have been dominant in explaining the long-term decline in homicide, following Elias' (1994following Elias' ( [1939) theory of the Civilising Process, which postulated that throughout history, levels of violence in society have gradually reduced as the State's monopoly of power increased. Elias claimed that as a result of the rise of centralised state governance, individuals became increasingly economically dependent upon each other, were required to co-operate with each other and avoid public displays of violence. The increase in external social control led, Elias argued, to a concomitant growth in internal self-control and reluctance to engage in public conflict, which started in the upper echelons of society and gradually spread to the lower classes. In other words, the shift from feudal society to one characterised by a monopolised state power and consequent changes in the structure of the economy and society meant that over time, people became increasingly restrained and sensitised to violence, and levels of violence (fatal and non-fatal) diminished.
Elias ' (1994 [1939]) theory was subsequently supported by empirical evidence of the decline, primarily provided by Eisner (2001) and also Gurr (1981). Eisner's work also revealed that the majority of the decline was accounted for by a reduction in 'honourbased' public killings between males. This tallied with Verkko's (1951, cited in Eisner, 2008 earlier work on homicide rates during the first half of the 20th century, which found that fluctuations were invariably driven by changes in the volume of male-male killings. Verkko observed that when homicide rates are high, the proportion of female victims tends to be low, and vice versa. Spierenberg (2012) also highlights the gendered nature of homicide trends, with high homicide rates characterised by high levels of male-male homicide: 'whenever and wherever homicide rates are high, this high level nearly always results from a prevalence of fighting among men' (p. 32). Eisner's (2001Eisner's ( , 2008 work has led contemporary understandings of the long-term decline in homicide, and emphasised the importance of interpreting homicide trends in context with social and cultural history. Despite broadly supporting Elias' model, Eisner also found contradictory evidence, including a continuing decline in the homicide rate in England following the Glorious Revolution of 1688, which challenged the monarchic absolutism and ought to (according to Eliasian theory) have led to a subsequent rise in homicide rates: 'these examples suggest that the dynamics of internal pacification do not depend exclusively upon the absolutist social figuration emphasised by Elias' (Eisner, 2001: 631). Eisner (2001Eisner ( , 2008 argued that the increase in state power was insufficient in explaining the long-term decrease in homicide, and other cultural developments must also be taken into account, including the increase in religion, expansion of education, and a disciplining process associated with capitalism (Eisner, 2001). He contended that, combined with the increase of state power, these social changes led to increased self-control and laid the groundwork for 'a more orderly conduct of life' (p. 631).
In considering the more recent decline in homicide, Eisner (2008) argues against country-level explanations, highlighting that they are insufficient in explaining the consistency of the homicide decline across countries with heterogeneous policies and economies. He again emphasises the importance of culture as 'the only phenomenon that travels fast enough to affect such vast areas roughly simultaneously' (Eisner, 2008: 311). Eisner (2008) suggests that, given that changes in homicide rates are largely driven by variations in the volume of male-male killings in public spaces, explanations need to focus on lifestyle, or the 'conduct of life' of these men, and examine how homicide trends link to 'change in norms and expectations about how young men interact in public space' (p. 312).
According to cultural theories of homicide decline then, the decline in homicide across much of the Western world from the early-mid-1990s derived from cultural changes affecting countries with similar norms and values, culminating in a reduction in the number of male-male killings in public spaces. The continuing increase in homicide numbers in England and Wales for another decade presents a quandary for this perspective, given that it has historically followed similar trends to other parts of the Western world, and shares many similar cultural characteristics to countries where the decline began in the 1990s. In light of this, two key questions addressed in the analysis below are first, the extent to which the latterly decline in homicide numbers in England and Wales has been driven by a reduction in male-male killings, and second, has the decline been driven by a reduction of homicides in public spaces?

Changes in lifestyles, routines and opportunities for fatal violence
Alternative explanations for the decline in homicide may be broadly considered under the umbrella of routine activity theories, given that rather than focusing on culture, they refer to macro-and micro-level factors that in some way serve to facilitate or reduce the likelihood of fatal violence occurring. Many of these explanations have focused on country-specific factors, such as changes in population, economic trends, the legalisation of abortion and the banning of lead paint/fuel (Lehti, 2014). Indeed, Lehti (2014) emphasises the importance of analysing country-specific trends in homicide and disaggregating homicide to enable a nuanced understanding of 'what kind of persons are dying, and in what contexts they are dying and who are killing them ' (p. 189). He reports that while in some countries (Netherlands, Spain and Switzerland) the decrease was observed across all major subtypes of homicide, the decline in other countries (Sweden, Italy and Germany) was driven by a decrease in expressive homicides. Lehti's (2014) analysis of the homicide drop in Finland revealed that the decline was mainly observed in malemale alcohol-related killings and that there were counter-trends in certain subtypes of homicide, including those involving young female perpetrators. These counter-trends may be masked in aggregate-level analyses of homicide rates and provide support for approaches focusing on nation-specific, disaggregated homicide data.
Similarly, Aebi and Linde (2014) criticise theories based on assumptions of homogeneity across Western countries (e.g. in economic terms) and associate the downwards trend (like the upwards trend in the late 20th century) with changes in youth lifestyle and opportunities, brought about by changes in social control, increased security and the proliferation of the Internet, which has led to young people spending more time indoors. They also highlight that despite numbers of male victims driving overall homicide trends, World Health Organisation (WHO) data show that from the mid-1960s, the number of female victims also increased (and subsequently decreased) proportionately to the changes observed in male victims. Moving away from Eisner's (2008) account of the recent changes in homicide, they conclude that 'any explanations of the trends observed must identify factors that have an influence on the victimization of both genders' (Aebi and Linde, 2014: 568).
One such explanation is the development of technology. The digital age has brought along with it an increasing tendency to socialise online, which has changed the social culture of many countries. The emergence and proliferation of the Internet since the mid-1990s has led to many young people spending large amounts of time in private spaces, rather than socialising in public spaces (Aebi and Linde, 2010). In addition, the expansion of the dark web has led to the introduction of cryptomarkets, which have to a certain extent reduced the volume of face-to-face risky transactions (e.g. surrounding firearm and drug markets) and replaced them with less violent and conflict-ridden online transactions (Morselli et al., 2017). It may be then, that certain subtypes of homicide (i.e. public homicides involving young people and those resulting from firearm/drug-dealing conflicts) have declined as the use of the Internet has grown.
Another potential explanation for the decline in particular subtypes of homicide is the effects of criminal justice policies. Indeed, the drop in the homicide rate in North America, which was accompanied by a broader decline in violent and property crimes, coincided with punitive criminal justice policies. Many commentators accredited the zero-tolerance style policing introduced by Mayor Guilliani of New York in the mid-1990s, although others explained the decline as a result of the stabilising and declining crack cocaine market which had exploded in the late 1980s and correlated with the late 1980s/early 1990s spike in homicide rates across the United States (Baumer et al., 1998;Blumstein et al., 2000;Bowling, 1999).
In England and Wales, a similar zero-tolerance style of policing was introduced in the late 1990s by the 'New Labour' government (Newburn and Jones, 2007), approximately 5 years prior to the decline in homicide beginning in this country. There were also a number of criminal justice interventions deliberately implemented to reduce the number of homicides, following Brookman and Maguire's (2005) Home Office-funded research into the possibilities for reducing homicide. Measures included staggering the closing times of pubs and clubs (to avoid mass exodus), greater police presence in known hot spots and greater use of plastic glasses. Contemporaneously, CCTV has become increasingly widespread, adding to the melting pot of situational crime prevention aimed at reducing public crime and disorder (Zedner, 2007).
As the risk of terror attacks has become omnipresent, security through surveillance and other forms of situational crime prevention have increased, alongside developments in emergency service provision in high-risk areas (Coaffee and Wood, 2006;Zedner, 2007). Hence, it is plausible that the homicide drop in England and Wales has been driven to an extent by changes in 'pre-crime' criminal justice policies (Zedner, 2007) that have served to reduce public disorder and (fatal) violence. Criminologists trying to explain the more general 'crime drop' in England and Wales from the 1990s have posited the 'security hypothesis' (Reiner, 2016) as one of the most plausible explanations for falls in non-violent crime (such as property crime); however, the mass expansion of technology and security has not been considered as a potential explanation for falls in homicide. This is explored in the analysis below through examining the location and circumstances of homicides.
Finally, advances in medical technology may have contributed to the reduction of lethal violence. In recent years, there have been significant improvements in emergency service response times, treatment, life support and trauma response (Harris et al., 2002), rendering serious assaults less likely to result in loss of life (remaining 'assaults' rather than 'homicides'). If homicide rates decline while non-fatal assaults incline, one potential explanation is that fewer assaults are leading to a homicide. This supports Lehti's (2014) reminder of the importance in analysing changes in homicide trends alongside trends in non-lethal violence, as this provides a more accurate insight into changes in violent behaviour. The recent decline in homicide observed across North America and Eastern Europe was contextualised by a more general crime drop; however, in other areas in Europe it was not. Lehti (2014) argues that a drop in homicide accompanied by a drop in non-fatal violence proffers support for cultural explanations, while a decline in homicide with no accompanying decline in general violence indicates alternative explanations. This is explored in the analysis below through a consideration of concurrent non-lethal violent trends in England and Wales during the period of homicide decline. Equally, it is important to note that increased sensitisation to violence could lead to fewer homicides alongside a parallel increase in non-fatal violence -as people become more sensitive to violence, the volume of petty violent incidents reported to the police may increase and inflate the number of violent offences via effective net widening.
The above review of literature on international homicide trends paints a conflicting picture. On one hand, simultaneous trends in aggregate homicide levels across much of the Western world indicate culture as the predominant explanation, given its ability to transgress national borders, and country-specific economies and policies. On the other hand, analyses focusing on disaggregated, national-level homicide trends, and examining homicide in context with broader violence and crime patterns suggest that there are subtle differences in the nature of country-specific trends. This may indicate a variety of driving forces behind fluctuating homicide rates, including changes in routine activities, social and criminal justice policies, and advances in medical technology (Aebi and Linde, 2014;Lehti, 2014), all of which reduce the opportunity and likelihood of fatal violence.
There is unquestionable merit in analysing national-level homicide trends, and disaggregating homicide in order to ascertain the types of individuals and homicides driving change. In light of the diversion between the recent homicide trend in England and Wales and other areas across the Western world, it is particularly important to examine the nuances of the trend here and contribute to international theory on the most recent decline.

Homicide trends in England and Wales
The focus here is on understanding the recent decline in homicide in England and Wales in context with dominant theories on homicide decline, rather than explaining homicide per se. 2 Homicide numbers in England and Wales have historically been comparatively low and the homicide rate remains one of the lowest in the world  -the most recent figures indicate a rate of 12 per million population 3 (Office for National Statistics (ONS), 2019b). Nonetheless, overall trends in England and Wales have tracked closely alongside trends in other developed countries Eisner, 2001Eisner, , 2008Gurr, 1981;Soothill and Francis, 2012), steadily decreasing from the middle ages until the 1960s, before rising for approximately four decades before peaking in 2004 and subsequently declining until 2014. The analysis below covers the period from 1 April 2003 (immediately prior to the peak in rates) to 31 March 2015, just after the rate had begun to rise again, at the same time as an upwards trend has been witnessed across many European countries (ONS, 2019b) and US states (Rosenfeld, 2016).

Methods
Homicide data in England and Wales are recorded in the Home Office-owned HI, which collates data from the 43 police forces and documents details of the suspect and victim demographics as well as victim-perpetrator relationship, incident characteristics and court outcomes. HI data are considered to be some of the most detailed and robust data on homicide in the world and are designated official statistics on homicide in England and Wales (ONS, 2019b). However, the database is not infallible and is subject to missing data, especially for multiple suspects and victims. The data are also subject to a degree of inconsistency across forces, and some variables in the Index have been documented as particularly prone to inaccuracy, for example, intoxication by alcohol and/or drugs appears to be significantly under-recorded (Miles, 2012). Despite these limitations, the Index still provides a good indication of long-term patterns and changes over time. There have been a number of changes to recording practices over the years, most recently following a review in 2000 (Moxon, 2001), which led to a number of changes being implemented and improvements in recording practices (reflected in the data from 2007 onwards). For this reason, the findings for some variables below (including intoxication) are only based upon data from 2007 onwards, due to a large amount of missing data prior to these changes.
The data on which the analysis below was conducted were obtained from the Home Office in an anonymous excel spread sheet and converted into an SPSS database for the purposes of analysis. Some of the analyses were also conducted using R statistical software. A data cleaning process removed all erroneous and duplicate data held within the raw database. This included removing all homicides that had occurred prior to 1 April 2003 (homicides appear on the database on the date when they are recorded by the police rather than the date they occur), removing all cases 'no longer recorded as homicide' (e.g. incidents initially recorded as homicide but subsequently reclassified as suicide or accidental death) and finally, removing three sets of data pertaining to unique events involving multiple victims that would potentially skew the analyses. This included removing all of the 52 London bombing victims, who were killed on 7 July 2005; 20 Morecambe Bay drowning victims, who were killed on 5 May 2004; and 12 Cumbria shooting victims, who were killed on 2 June 2010.
Following this, there were a total of 11,264 observations or rows of homicides in the database. However, this included a large amount of duplicate data as the database contained individual rows for each homicide victim and suspect. For homicide events involving five victims, this meant that the suspect data would be repeated five times, and similarly, for homicides involving only one victim but multiple suspects, there would be a separate row for each suspect, duplicating the victim data for each suspect row. To overcome this, three versions of the database were created: a 'suspect database', containing a single observation for each homicide suspect (excluding multiple victim data); a 'victim database', containing a single observation for each victim (excluding multiple suspect data); and an 'event database', containing a single observation for each homicide event (excluding multiple victim and suspect data). The three respective databases consisted of 7462 unique homicide suspects 4 ; 8038 unique victims; and 7770 homicide events. In the analyses below, suspect-based analyses were conducted using the 'suspect database', victim-based analyses were conducted using the 'victim database' and eventanalyses were conducted using the 'event database'.

Findings
As Figure 1 illustrates, the homicide rate in England and Wales began to decline beginning in 2004, 5 and continued to do so at a steady rate until 2014, when it began to rise again. During this period, the homicide rate decreased from a high of 14.38 per 1 million in the population to a low of 8.45. Through disaggregating the homicide rate by geographic region (based upon police force area), it is clear that the downwards trend was driven by declines in particular regions, most notably Greater London, the North West and closely followed by the West Midlands. Other areas, such as East Anglia and the South West, have seen relatively stable (comparatively low) homicide rates throughout the period of decline.
In terms of the homicide location, the Index records whether the incident took place in a public or private space, and what type of space. Due to changes in recording procedures, the quality of data on homicide location improved significantly from approximately 2007. As Figure 2 illustrates, prior to this, the majority of homicides were categorised as occurring in an 'unknown location'. This chart also shows a shift in the proportion of total homicide occurring in each type of space; from 2007 to 2015, an increasing percentage of all homicides occurred in residential locations. The line chart focusing on homicide rates in each general location from 2007 onwards ( Figure 3) illustrates variations in the decline of homicides occurring in public and residential spaces. The rate of homicides in public areas declined consistently until 2011 then remained stagnate at about three homicides per million for the next 3 years, rising again when the overall homicide rate increased in 2015.
A more detailed analysis of the locations of homicides revealed the rate of victims killed in a 'house or dwelling' peaked in 2010 (0.62 homicides per million), declined until 2014, and then began increasing again concomitantly with the overall homicide rate. Rates of those killed in a 'street, footpath or alleyway' declined until 2012, reaching a low of 0.14 homicides per million before starting to rise. In addition, the rates of homicides in 'open spaces' or around 'licensed premises' approximately halved from 2007 to 2015 (from 0.12 to 0.048 per million and from 0.098 to 0.46 per million, respectively).
These findings indicate a subtle change in the nature of homicide over the period of decline, with a reduction in homicide incidents in public and the overspill areas of pubs and clubs. In 2007, a homicide resulting from a 'fight, brawl, etc.' was the most prevalent recorded circumstance (four homicides per million). This dropped to a low of 1.7 per  Corresponding to the decline in homicides occurring in public spaces, and those involving fights and brawls, there was also a significant decline in the role of drugs and alcohol as factors in homicides. Figure 4 shows the change in rates of suspects and victims who were recorded as under the influence of alcohol, drugs or both at the time of the homicide. Rates of alcohol intoxication have fallen dramatically for both victims and suspects; however, for victims, there have been periods of increases in alcohol as a factor (from 2009 to 2011 and again after 2013) whereas this has been a consistent drop among homicide suspects. Rates of illicit drug use or a combination of drugs and alcohol use among victims and suspects at the time of the homicide have remained low and fairly consistent.
The analysis of method of homicides over time painted a more complex picture. The rates of homicides involving shootings (firearms, crossbow) declined substantially from a high rate of 1.  Analysis of the victim-suspect relationship also failed to provide insight into what underpinned the decline in homicide. The rates of homicides within several categories reduced during the overall decline period of 2004 to 2014. For example, acquaintance homicides declined from 4.47 per million to 2.28; stranger homicides declined from 2.47 to 1.30; homicides involving current or ex-partners declined from 2.28 to 1.65. However, for all categories, there were fluctuations from year to year; there were no discernible drops in any one category, suggesting there has not been a marked change in the relationships between homicide victims and suspects.
A key question of this analysis was the extent to which the overall drop in homicide rates was driven by a reduction in male-male killing. When calculated as percentages of overall homicides, the relative proportions of male and female victim and suspects remained stable, with females consistently accounting for around 30% of victims and 9% of suspects. However, when factoring in changes in the overall population, it is clear that male-male homicides were overwhelmingly driving the overall decline in England and Wales: Figure 5 displays the rates of male and female suspects and victims per million in the population, illustrating that the rate of male suspects actually dropped nearly 63% from 2004 to 2015, and the rate of male victims decreased by 50% over this period. However, the rates of female suspects and victims remained relatively constant throughout the period of decline in homicides.
The analysis of suspect and victim ethnicity is dictated by fairly crude categories that do not necessarily reflect the complexity of ethnicity and, for suspects, the data are based upon self-classification. Notwithstanding this, the analysis revealed that overall, the ethnicity proportions of both suspects and victims has remained stable over the period of decline; Whites making up approximately two-thirds of suspects and three-quarters of victims.
Finally, the analysis of changes in suspect and victim characteristics over time illustrated that although the gap in median age of victims and suspects has remained stable, with victims usually 7 or 8 years older than suspects, the median age for both homicide suspects and victims has increased ( Figure 6). 6 This indicates that the demographics of those involved in homicides have shifted slightly; these events are not a phenomenon limited to the 'young'.
The analyses presented above demonstrate the importance of looking beyond aggregate homicide rates, in order to reveal subtle patterns in the disaggregated data that allow the overall rate to be unpicked. In so doing, it appears that although the decline between 2004 and 2014 was observed across a number of homicide categories, it was particularly apparent in homicides involving young intoxicated men, in public spaces. It also revealed that the steepest decline occurred in Greater London, followed by other large cities. In order to further examine this aspect of the decline, further analyses were conducted focusing only on Greater London. The most noteworthy findings pertain to methods of homicide, shedding some light on what drove the national picture during this time.
As Figure 7 illustrates, during the period of decline, methods of homicide in Greater London decreased across nearly every category, with the sharpest decline observed in the rate of sharp instrument homicides. The rate of sharp instrument homicide remains consistently higher for Greater London compared to national levels (see Figure 8) and   homicides per million to a low of 2.5 in 2014. In London however, rates of sharp instrument homicide peaked in 2007 (driving a national-level peak) and subsequently experienced a major drop of 40% by 2009. Since then, there have been marginal changes, however, the rising numbers of fatal knife-assaults in recent years, which have been widely reported and the subject of much concern, are to a large extent on par with the concomitant population growth in London. Calculated as a rate, sharp instrument homicides in London (and nationally) have actually dropped considerably since the mid-2000s. In addition, the rate of shootings in London decreased from approximately 3.1 per million to 1 per million, and similarly, the rates of homicides involving a blunt instrument dropped from 2.8 to 0.8 per million. The rates of homicides involving some form of beating (hitting/kicking with no weapon, causing to fall) peaked in 2008 and have not substantially altered since 2003.
While it is true that the rates of sharp instrument homicides in London decreased between 2007 and 2014, it is interesting to note that the percentage of homicides involving a sharp instrument increased from accounting for 34% of all homicides in London in 2003, to 52% in 2015. In other words, within the broader context of declining homicide (including those involving knives), fatal knife-violence, in London especially, has become the predominant method of homicide. This finding is crucial in context with the current increase in homicides involving knives that has been widely reported since the turn of 2015. Crucially, although proportionally, knife-related homicides have been increasing, when calculated as rates within the growing population, people are actually safer from knife-crime now than they were in 2007.

Discussion: Understanding the homicide drop in England and Wales
The homicide rate in England and Wales diminished steadily from its peak in 2004 to the year ending April 2014, when the rate began to increase again. Importantly, this decline transgressed a period of economic boom and the subsequent economic crisis (from 2007), which has been followed by long-term recession, austerity and well-documented cuts to welfare and public services, including the police force (Millie, 2014). This context raises questions about economic and political theories of homicide trends (e.g. Currie, 1997;Hall and McLean, 2009;Messner and Rosenfeld, 1999), although the current increase in homicide in England and Wales has recently been theoretically attributed to the social impact of austerity and neoliberalism (Ellis, 2019).
The homicide drop in England and Wales also began a decade later than in other areas of the Western world, presenting a challenge to dominant theories. According to popular cultural theories, changes in homicide simultaneously occurring across developed countries are best explained by changes in culture (Eisner, 2001(Eisner, , 2008Elias, 1994Elias, [1939), particularly affecting the regulation of young men in public spaces. It is unclear how the divergent trend witnessed in England and Wales for approximately 10 years is accounted for within such an approach, given that it conceptually (and historically) belongs to the same group of countries that share similar cultural trends. This reinforces the need for cross-cultural and disaggregated homicide analyses, in order to develop country-specific understandings of homicide trends.
The findings presented above illustrate that the decline in homicide numbers observed across England and Wales was broad in the sense that there were decreases in rates across many types of suspect-victim relationships and demographic attributes, methods, locations, regions and circumstances; but also specific in that certain subtypes of homicide decreased more substantially than others. The decline was most apparent in metropolitan areas (especially Greater London and Greater Manchester), in homicides occurring in public places, in homicides deriving from 'fights and brawls' and around licenced areas, in homicide events involving intoxicated (by alcohol) suspects and victims, and in homicides involving younger, male, suspects and victims. This indicates that a reduction in fatal assaults arising from young men engaging in violence outside pubs and clubs after consuming alcohol has, to a certain extent, driven the decline. Eisner (2001Eisner ( , 2008 argues that changes in homicide rates are explained by changes in 'conduct of life', pertaining to expectations and moral codes surrounding young men and their behaviour in public spaces. These findings offer support to this theory, given the apparent decline of male-male homicides occurring in public spaces and involving alcohol intoxication, which are also contextualised by a concurrent decline in drinking for those aged 16-24 years between 200516-24 years between and 201716-24 years between (ONS, 2018. However, rather than supporting cultural change per se, or indeed Elias' (1994Elias' ( [1939) concept of cultural pacification, the findings also corroborate Linde's (2010, 2014) argument that homicide changes are driven by more specific changes in lifestyle, norms and routine activities (such as less frequent drinking in public places, increased socialising in private, etc.). This is further supported by the dominance of Greater London and Greater Manchester (and to some extent, the West Midlands) in the decline, which indicates something particular about these areas and their inhabitants. As discussed earlier, since the turn of the century, policing and criminal justice policies have sought to reduce public disorder and crime, including zero-tolerance policing, problem-oriented policing and situational crime prevention, alongside specific policies aimed at homicide reduction and prevention (recommended by Brookman and Maguire, 2005). Combined with this, the increasing magnitude of the terror threat and accompanying focus on security and 'pre-crime' measures (Zedner, 2007) has enhanced both levels of surveillance and the capacity for emergency services to respond to incidents. Concurrently, advances in medical technology and trauma response have improved fatality rates (Harris et al., 2002). The effects of such measures are likely to be more acute in urban areas identified as high-risk (in terms of levels of violence/homicide and terror threats). It may be then, that a combination of changes in lifestyle, physical and social environment, targeted criminal justice policy and emergency responses to potentially lethal violent incidents effected a reduction in lethal violence resulting from public, alcohol-related assaults involving young men. This would also explain the increase in the proportion of homicides taking place in private spaces, which have not been the focus of security policies, would not have necessarily benefitted from enhancements in emergency service responses and where more socialisation now takes place.
What remains unclear is how the more recent (2014)(2015)(2016)(2017)(2018) increase in the homicide rate fits in with such an explanation. Crime prevention and criminal justice policies did not suddenly lapse, indeed, security measures and levels of surveillance have never been higher. Furthermore, the recent upturn in homicide challenges explanations of decline based on advances in medical technology (Aebi and Linde, 2014;Harris et al., 2002;Lehti, 2014). In the absence of disaggregated homicide data for the past 4 years, it is only possible to speculate (and draw upon public reports) at this time that the current increase is specific to a particular subtype of homicide, which appears to be fatal knife-attacks involving young men; tentatively linked to the increase in violence associated with county lines 7 and the long-term effects of austerity on vulnerable young people (Ellis, 2019;Townsend, 2019;Younge, 2018).
The importance of disaggregating data not only by country, but by demographics and contextual circumstances, is aptly demonstrated by the analyses focusing on Greater London. Disaggregating the decline in homicide by geographical area revealed that the overall decrease witnessed across England and Wales was largely driven by a reduction in homicide numbers in London (followed by other metropolitan areas), despite the growing population levels. Crucially, the proportion of homicides in Greater London involving knives (and young males) experienced a steady increase throughout the period of decline; however, the rates by population of knife-related homicide witnessed a dramatic drop between 2007 and 2009, before stabilising. Focusing on aggregate rates of homicide, and on proportions rather than rates then, can be misleading, as evidenced by the media reporting of knife-homicide over the past 5 years. Rather, the important question ought to be why, in the context of declining numbers of homicides, are a growing proportion of illegal killings conducted using knives?
Finally, an important question concerns whether there has been a more general decrease in violence alongside the decline in homicide. The broader landscape of nonlethal violence has been found to vary across other parts of Europe (Lehti, 2014) and is essential in understanding the full context of homicide trends. Official statistics on nonfatal violent crime in England and Wales are reported by the ONS, who report on police recorded offences and the annual Crime Survey for England and Wales (CSEW).
The latest report on violent crime indicates that alongside the decline in homicide, forms of non-fatal violence have also witnessed a long-term steady decline. The number of violent offences estimated by the CSEW peaked in 1995 at 3.8 million, and there has subsequently been a statistically significant 68% reduction to 1.4 million violent offences estimated in the year ending March 2018 (ONS, 2019a). These figures suggest that violent crime as a whole declined throughout the period of decline in homicide numbers and thus refute the idea that medical advances may explain a reduction in fatal assaults (one would expect to see an increase in non-fatal violence if this was the case).
Importantly however, according to Walby et al. (2016), applying a 'new methodology' to CSEW data whereby the cap on repeat victimisation is removed, overall figures for violence indicate that it has been rising rather than decreasing since 2009. This analysis revealed that violence against women and domestic violence against women in particular has been rising since 2009. Violent crime against men (overall) has been falling; however, domestic violence against men has also been increasing since 2009. These findings not only reiterate the importance of methodology and of disaggregating violence data, but also have implications for the findings in this analysis. First, these figures suggest that the decrease in homicide may not be wholly contextualised by a broader decline in general violence, as would be expected by proponents of cultural theories.
Notwithstanding this, the fact that violence against men has been declining supports both cultural theories and those based around changes in lifestyle and routine activity theories (e.g. Aebi and Linde, 2014;Eisner, 2001Eisner, , 2008. Second, the enhancement of medical technology and emergency service provision may be serving to reduce the number of homicides but simultaneously leading to the inflation of non-fatal violent events. And third, it may be that factors serving to prevent 'public' homicide (such as criminal justice policies) are leading to an increase in more private forms of violence and homicide, which ties in with the observed increases in domestic violence (towards men and women) reported by Walby et al. (2016) and the observed increase in proportions of homicides occurring in private spaces.

Conclusion
Overall, a complex picture emerges. On one hand, the most recent homicide drop in England and Wales appears to have been general in that it is observable across all categories. On the other hand, the patterns emerging through disaggregated analysis of homicide data reveal that some subtypes of homicide decreased more substantially than others, and point to one particular subtype of homicide as a key driver in the declinefatal assaults involving intoxicated young men, in public spaces, deriving from fights and brawls. Crucially, the decline has been much more apparent in metropolitan cities, with Greater London experiencing the most dramatic decline in numbers, and importantly, a reduction in the rate of homicide per population. Within the context of this decline, the proportions of certain types of homicide have increased, including those occurring in private spaces and homicides involving knives.
The findings proffer little support for Elias ' (1994 [1939]) Civilising Theory, under which one might expect to observe a much more general and continuing decline in homicide. Clearly, cultural norms and expectations of behaviour are important, and to this extent, the findings resonate with Eisner's (2001Eisner's ( , 2008 concept of 'conduct of life'. However, the findings also indicate factors beyond changes in cultural norms as key in explaining the decline in homicide. Rather (or in combination with cultural explanations), the analysis revealed patterns broadly supporting Aebi and Linde's (2014) integration of lifestyle theory with routine activities; that is, the findings indicate that a plethora of social and criminal justice policies (including those aimed specifically at public forms of homicide), lifestyle changes (brought about by not only cultural change but also major advances in technology), and advances in emergency service and medical responses may have combined to reduce the opportunities for fatal assaults.
To add to the complexity, the more recent increase in homicide in England and Wales, recorded between 2014 and 2018, challenges this and other dominant explanations of homicide decline. It also highlights the need to disaggregate homicide into subtypes -as Brookman (2005) and Brookman et al. (2017) contends, different forms of homicide demand different explanations -and so do trends in particular subtypes. Although the absolute numbers and rates of knife-related homicide decreased substantially during the period of decline, the proportion of all homicides involving knives has increased. The implications are positive on one hand, indicating that media reports may have over-exaggerated the risk of knife-violence. On the other hand, the predominance of knives as a method of homicide (calculated as a percentage of the total) justifies an ongoing focus on tackling knife-crime. Despite the limited data available on the 2014-2018 increase, it appears that knives may be a driving factor, which would fit with the findings reported here. There is plausible argument developing that the increase may be linked to County Lines violence (Townsend, 2019) and a result of the inevitable social implications of the expansion of neoliberalism and long-term austerity witnessed in England and Wales since the economic crash in 2007 (Ellis, 2019). However, it is essential for further research to focus particularly on the micro and macro dynamics of this subtype of homicide and the individuals involved, in order to fully understand the context of fatal knife-crime and develop effective prevention strategies.