The Impact of the Coronavirus ( Sars-Cov-2 ) Lockdown on Crime in New York and London , March-June 2020 : A Comparative Study

124 Website: www.ijbmr.forexjournal.co.in The Impact of the Coronavirus (Sars-Cov-2) Lockdown ░ ABSTRACT: The objective of this paper is to assess the relationship between The Spring 2020 COVID-19 Lockdown and the levels of crime in New York City (NYC) and London. Our proposition, derived from the Routine Activity Theory (RAT), the ‘breaches’ theory and input from the 2020 research on lockdown and crime, hypothesised that lockdown measures would lead to reductions in crime. The crime categories selected for this study were: homicide, rape, robbery, violence against a person, burglary, theft and vehicle theft. T-test, F-test and the Ordinary Least Squares (OLS) regression calculations were used to test the hypotheses. The four-month lockdown period in 2020 produced a 15% and 31% crime reduction in NYC and London, respectively. In the case of London, the overall results indicate that changes in routine human activities were indeed largely correlated with the reduction in crime. However, crime patterns in NYC in spring 2020 turned out to be inconsistent. A comparison of crime patterns under lockdown proved dissimilarity between NYC and London. The two-city comparison indicates that crime change related to lockdown may vary across crime types, places, and timespans or may have a detrimental effect on crime levels. The study may be considered suitable for replication and elaboration, particularly in view of the extended longevity of lockdown measures.

Website: www.ijbmr.forexjournal.co.in The Impact of the Coronavirus (Sars-Cov-2) Lockdown The analysis examines changes in patterns of the following crime categories, in NYC: murder, rape, robbery, assault, burglary, grand larceny, grand larceny auto, and in London: homicide, rape, robbery, violence against a person, burglary, theft, vehicle theft (theft of or from a vehicle), resulting from the COVID-19 lockdown measures and associated changes to routine activities. The crime terminology differences (US vs. UK) have been applied consistently throughout the paper.
The fundamental presumption for the study stems from the observance that change in life routines alters expected crime levels. This assumption was drawn from the theoretical contribution of environmental criminologists [5] and is rooted in the Routine Activity Theory (RAT) [5] as well as the Crime Pattern Theory (CPT) [7,8]. According to environmental criminology, crime is driven by opportunities. A crime opportunity forms when a motivated offender comes into contact with a suitable target in the absence of capable guardians. The early spring 2020 lockdown measures significantly disrupted the supply of such opportunities to the extent that it is expected that crime levels will have changed correspondingly, for some crime categories probably more than for others.
A further foundation for this study was drawn from the rapidly growing body of work on the crime effects of the pandemic [9][10][11], as discussed in more detail in the Literature Review section of this paper.
This study also exploited the observation that the introduction of lockdown could be viewed as a profound change in everyday patterns of behaviour (not only crime-related) called a 'breach' [12], which results from a technological or social development. Some other scholars, for example Stickle and Felson [13] described the lockdown as a type of 'natural social experiment'. This profound change, widely reported in the current scholarly and popular body of literature, across genres, resulted in numerous repercussions in people's lives, in areas such as education, work and leisure and, last but not least, produced changes in crime patterns. For instance, it was found that two particular measures of lockdown, that is, social distancing and stay-home orders, limited the pool of otherwise common opportunities for theft as well as burglary [14] although not as consistently as had been expected.
Following the assumptions drawn from the RAT and CPT foundations, the 'breaches' proposition as well as the conclusions reported from the contemporaneous observation of the lockdown, this study focused on the changes to crime volume and statistical characteristics in the two, already listed, geographical locations, New York City and London. The locations were a convenience sample, while their respective population size -NYC with 8.4 million and London with 8.89 million -made the comparison feasible. The cities constitute hubs for the US and the UK, respectively and have arrived at comparable levels of civilisational, economic and social development. Under normal circumstances, both NYC and London had experienced a barrage of constant, persistent and rich criminal activity. In order to enrich the scope of the study a comparative approach was chosen, both in the chronological (2019 vs. 2020) and crime categories.
The convenience-dictated choice of locations, even if rationalised by the multiple similarities, also in crime data, involved an important methodological standardisation procedure, namely that referring to the samples. The samples, as utilised in the paper, that is the number of crime incidents committed in the NYC and London area respectively, were counted, processed and presented in a different manner by the respective police authorities, also in the period under question.
That is to say that the respective police records were not equal from the methodological point of view. The London's Metropolitan Police (MET) data presentation was substantially more detailed and comprehensive, while the NYC data wasin comparison -condensed and less detailed. In view of the data differences between the samples, the comparative descriptive analysis in the paper relied predominantly on the relative changes in crime trends, while the means and variance equality tests were applied in the relevant inferential statistics analysis.
The data was drawn from open sources, namely: the New York Police Department (NYPD) records 2 , and the Metropolitan Police resources 3 . The analysis covered March through June 2019 and March through June 2020.
The first eight hypotheses, grounded in the RAT literature, sought to establish the relationship between lockdown and crime, in a comparative manner: both year-on-year and city-tocity. The further seven hypotheses, based on the same crime categories, looked to identify a possible relationship between the number of the Covid-19 cases in both cities, reported daily and aggregated accordingly, treated as a proxy for the independent variable (i.e., lockdown), and crime pattern changes.
The analysis of the trends in the selected crime types in both cities led to the conclusion that there are numerous patterns reflecting the crime evolution in the first four-month lockdown period in 2020: a decrease, increase, initial decrease and then a rise, or fluctuations in the number of crime incidents. These initial observations suggest that the lockdown truly led to the preliminary (albeit short-term) drop in the number of incidents in some selected crime categories. This conclusion was therefore recognised to partly support the RAT, confirming the rule that crime reduction indeed results from both impeded crime opportunities as well as an increased degree of guardianship, particularly in London, where six out of the seven initial Hypotheses were validated. However, the growth/decline trends as well as the dispersion and mean characteristics of the crime distributions proved trend dissimilarity in both locations. Additionally, the supposition which stated that the dispersion of crime would be larger in The Impact of the Coronavirus (Sars-Cov-2) Lockdown 2019 than in 2020 (Hypothesis 8), was only validated for the total aggregated number of crime incidents in NYC. In conclusion, the overall results of the analysis point to mixed conclusions, similarly to the findings of several contemporaneous research studies, namely that the initial drop in the selected crime categories did not prove to be sustainable. The findings from the analysis follow the observation of Stickle and Felton [13] that crime rates "have indeed changed, but unequally across different categories, types, places, and timeframes".
Therefore, the results of this study should be treated with caution. In consequence, the recommendations from the study findings focus primarily on directions for further studies. Policy makers are recommended to consider and estimate the costs of maintaining the lockdown, especially in view of the temporary character of the cost reducing effect on crime as well as certain detrimental repercussions as regards crime trends.
Some editorial conventions have been applied in this paper, in order to avoid unnecessary repetitions. The data and findings are first presented for NYC and then London throughout the paper. Some preliminary statistical test results were moved to the Appendix. Finally, the sources of data in the case of tables and figures were listed in the relevant footnotes and subsequently referred to in the text.

Opportunity Theories of Crime
Opportunity theories of crime, also associated with environmental criminology [15] investigate the physical and social characteristics of crime, circumstances which bring the offender and the target together, cognitive processes leading to the selection of the crime type and crime location, the influence of laws and procedures onto crime sites as well as the spatial distribution of crime in rural, urban and suburban settings [5]. One of the major examples of an opportunity theory of crime is The Routine Activity Theory.
Routine Activity Theory emerged amongst the numerous efforts undertaken in crime studies which served to explain the rise in crime in the United States after the Second World War [16]. For example, following World War II, the US saw a rise in criminal activity, which stood in contrast to the traditional theories that regarded poverty as the key driver of crime. While many scholars had tried to explain the phenomenon, it was Cohen & Felson [6] who demonstrated that crime patterns can be seen to increase in highly populated areas, in which the likelihood of victims and offenders meeting increases due to the law of large numbers. They determined that criminal activity relies on a few key factors converging in time and space: the offender, the target, and the lack of protection of the target. Routine Activity Approach (RAA) circumscribes that such a convergence of factors is not random and erratic but constitutes an epiphenomenon that occurs alongside the patterns of daily life.
Cohen and Felson [6] noted that repeated, mundane and regularly and/or cyclically maintained spatial and temporal patterns give rise to opportunities through which potential offenders and targets can come into contact or interaction, and which may transform into a criminal incident. Since 1979 the RAT has been refined by numerous authors who contributed to the theory [17] and created a model with six key elements: targets/victims, guardians, places, managers, offenders and handlers. The six elements proposed by the theory were presented as an inner/outer triangle model by Eck [18]. The inner triangle includes the three conditions which may co-exist for the crime to occur: the offender, the victim and the place, as presented by Cohen and Felson [6] and other proponents of the RAT. The outer triangle represents the three protective elements, which could prevent or thwart the crime: a handler (a crime prevention agent), a guardian (e.g., a parent) and a (place) manager (e.g., a store guard on duty). The triangle model is presented in Figure 1. Various scholars pointed out that changes to patterns in every day routines are not the only catalyst which may spur criminal activity. For example, Tilley and Sidebottom [16], elaborating on Cohen and Felson [6], listed several examples of routine activities, brought about by sweeping social changes, which facilitated the rise of crime opportunities, such as the incorporation of women into the labour force post-war in the 20th century, which lead to reduced guardianship of homes and thus created increased opportunities for burglary. Tilley and Sidebottom [16] also pointed out the role of technological progress, such as the invention and proliferation of portable electronic devices, for example cameras, iPods, iPads and mobile phones. Another trend was lifestyle changes, such as the boom in travel and holiday industries, which gave rise to notorious theft incidents and notoriously infamous tourist locations prone to pick-pocketing. On the other hand, the presence or absence of guardians may detain the potential offender from a harmful or offensive action [16] for example, in situations involving children.
The RAT was recently subject to testing via a real-life experiment, defined by one set of authors as "a naturally occurring, quasi-randomised control experiment" [13], while the lockdown measures were introduced worldwide to stop the spread of COVID-19. In this paper, it is anticipated that the  [19], in the aftermath of a war as well as during a cyclical or spontaneous severe economic crisis [20,21]. Some other disruptions caused by or during major sporting events contributed to the malfunction of traffic and public transport [22][23][24]. Ashby [9] identified two major differences between a pandemic and sudden natural disaster, pointing out that a pandemic has a slow onset, and it leaves the physical environment fundamentally undisturbed.
This latter body of literature, although it largely acknowledges the RAT fundamentals, differs from the RAT in the sense that it does not focus exclusively on daily routine alterations caused merely by the change in routine activities but locates the crime pattern changes within the context of multivariable disruptions and/or breakdowns in economic, social or even ontological foundations in human life. Prior to the 'pandemic scholarly literature', in 2006, there emerged a proposition stated by Killias [12], which defines several types of new opportunities ('breaches') for crime, resulting from the changes in everyday patterns of behaviour, which themselves are caused by technological or social development. Other types of breaches, listed by Killias [12], irrespective of their lifespan, result from sweeping, consequence-laden political decisions, such as the end of the cold war, followed by profound systemic changes in post-communist countries or the 1991 re-unification of Germany. Furthermore, Killias [12] argues that certain 'breaches' so exceed the status quo that they may give rise to novel criminal activities not yet envisaged by the lawmakers [12]. Stickle and Felson [13], on the other hand, claim that the recent, lockdown induced changes in crime ("the most salient aspect of the steep fall in crime") are the consequence of the politically empowered legal stay-at-home orders. However, the growing body of the most recent scholarly literature on the effects of a series of lockdowns may yet have to come out with more consistent conclusions. Presently, those conclusions cannot be perceived as unequivocal.

Initial Evidence of the Covid-19 Pandemic Repercussions in the Volume and Intensity of Criminal Cases: Contemporary Studies
The theoretical foundations in this study were complemented and completed by the contribution of the plethora of relevant contemporaneous papers published during the initial stages of the pandemic. As already pointed out, the relevant authors also grounded their research in the RAT and examined the effect of lockdown on the levels and patterns of crime in various locations, primarily urban US environments, sometimes in a comparative fashion. One week after lockdown, all recorded crime had declined 41% with variation by type: shoplifting (-62%), theft (-52%), domestic abuse (-45%), theft from vehicle (-43%), assault (-36%), burglary: dwelling (-25%) and burglary non-dwelling (-25%), compared to their expected rates. Shoplifting is elastic to reduced grocery sector mobility (MEC > 2), burglary dwelling is elastic to increases in residential area mobility (-1), while assault and theft from a motor vehicle are inelastic but still responsive to reduced movement under lockdown (0.48 and 0.69 respectively) Mohler, et al. [14] Daily counts of calls for police service in Los Angeles, January 2, 2020 to April 18, 2020 and Indianapolis, Indiana from 2 January 2020 through 21 April 2020 Differences in means from a baseline period, a regression using daily Google residential mobility Burglary and robbery were significantly lower in Los Angeles, but only marginally lower in Indianapolis; assault/battery calls were statistically unchanged in both locations. The overall effect was notably less than might be expected given the scale of the disruption to social and economic life Pietrawska, et al. [29] Los Angeles in the first three weeks of lockdown, A Safe City Crime data  The conclusions drawn from the analysis of the above findings are as follows: the first weeks of the lockdown produced a decrease in crime incidents in several of the locations under respective investigations, e.g., residential burglaries in Los Angeles, Memphis, San Francisco, Phoenix and Montgomery county [9], although not in Louisville or Boston. Similarly, Ashby's [9] empirical evidence demonstrated that incidents of serious assault in public declined in Austin, Los Angeles, and Louisville, but not in other US cities. Those findings were confirmed by the results conducted on the data from Chicago [9] and in the UK in the first week of the lockdown [28].
However, the results of the study published in the Campedelli, Aziani, Favarin [26] paper, conducted on the Los Angeles data, did not corroborate the supposition that burglary (as a crime category presumably affected by the increased guardianship due to lockdown) underwent reduction. On the contrary, Pietrawska [30] showed that burglaries in Los Angeles increased by 64% although she simultaneously demonstrated that city-wide burglary rates went down by 10%. Another group of scholars, Mohler, et al. [14], who also drew on the Los Angeles data (the number of calls for police service), noted that the overall effect of social distancing and stay-home orders was notably less pronounced than might be expected, including the anticipated rise in domestic violence. Indeed, more calls for police service in relation to domestic disputes were observed, however, they were not accompanied by an increase in domestic violence. Similarly, Piquero, et al. [11] noted that domestic violence did not prove to be on the rise in a consistent manner. Most scholars listed here, who researched the lockdown and crime relationship and who derived their rationale from the RAT school of thought, showed that the initial empirical results of their studies supported the theory, even if those results may not have turned out to be sustainable and were not consistent within various categories, crime types, places, and timespans, as distinctly concluded by Stickle and Felson [13].
In view of the theoretical foundations and the majority of the findings reported in the most contemporary studies, the proposition in this paper was put forward that the lockdown (nicknamed later The Spring 2020 Lockdown) which, according to all contemporary scholars referenced here, unequivocally affected routine patterns of human activity and had a reducing effect on crime in most locations researched, should also be tested and compared in New York City and London. The crime categories under investigation, as elucidated in the introductory section, are in New York City: murder, rape, robbery, assault, burglary, grand larceny, grand larceny auto, and in London: homicide, rape, robbery, violence against a person, burglary, theft, vehicle theft (theft of or from a vehicle).
The hypotheses were formulated in the customary pairs of the null and alternative propositions. In the case of the first eight categories, the independent variable was the time, in this case in four discrete measures, corresponding to the number of months: March, April, May and June in 2019 and 2020; the dependent variable, plotted against the time measures, was the number of crime incidents in the respective months, in four values, corresponding to the months, in the seven selected categories. In the case of Hypothesis 8, the proposition was to find out whether or not the aggregate crime trajectory in both cities in the relevant period was similar.
There followed a set of hypotheses, derived from the rationale which attempted to quantify the lockdown indirectly, via a proxy variable, in view of the impossibility to quantify the measures themselves (i.e., a collection of orders, recommendations and limitations) in their aggregate in an unequivocal, unambiguously quantified manner. As a collective variable, 'lockdown' may be perceived as ambiguous, thus rendering any attempt of a straightforward modelling of the correlations between the containment policies and crime rates also dubious. Therefore, the assumption was put forward to replace 'lockdown' with a proxy variable. It was the number of COVID-19 cases in the respected cities in the corresponding period that became the proxy variable. As those assumptions were not directly inferred from the core body of the underlying literature, the analysis applied to test Hypotheses 9-15 may be perceived as a work-in-progress experiment in the recent outcrop of COVID-19 and lockdown related research. The major intention was to seek an additional The Impact of the Coronavirus (Sars-Cov-2) Lockdown source of rationalisation and support for the first seven, that is, the core hypotheses.
The primary objective of the lockdown measures was the containment and, if possible, eradication of the disease, whilein the absence of a widespread and effective cure, at least initially, as well as a vaccine -it was deemed to be the most significant disease prevention policy. One of the by-products of the lockdown was the hypothesised reduction in crime incidents. Such reasoning is actually present in one of the conclusions made by Stickle and Felson [13].
In the case of Hypotheses 9-15, the intention is to try and establish a possible correlation between the independent (proxy) variable, that is the number of COVID-19 cases in both cities in the relevant period, and the dependent variable, that is the number of crime incidents in the same period.
Although there were criticisms regarding both the counting methodology in the case of the people falling victim to the disease, and the daily versus monthly counts of the variables, our regression tests were conducted in spite of the limitations. Therefore, seven more pairs of the customary null and alternative pairs of hypotheses were formulated.

░ 3. HYPOTHESES
Hypotheses grounded in the RAT and the inner and outer triangle of crime literature came first. In NYC and in London, respectively:

░ 4. METHODOLOGY
The data for the study was accessed from publicly available reports published by the NYPD 4 , and the Metropolitan Police 5 , for NYC and London respectively. The timespan for the analysis encompasses March through June 2020 as well as the corresponding period in 2019. The data was published monthly and used for the analysis accordingly. The data for the number of the COVID-19 cases in NYC and London was accessed from online. 6 7 All data sets were accessed in the period March through June 2020 and processed during the months June -September 2020.

Descriptive Statistics
The raw data was organised and processed using STATA 12.0. Descriptive statistics were used to present the data in Tables 2  and 3, respectively. These tables depict the nominal change in the crime incidents, as well as the relative change in the corresponding months of 2019 and 2020. The complete relative change between the number of crime incidents, in each category, for both cities, for the corresponding months in 2019 and 2020, is presented in Table 4.

Inferential Statistics
The analysis of the variables and relationships was conducted at the bi-variate level, also with the help of the statistical software STATA 12.0. Due to the fact that the sample sizes were unequal, preliminary tests were performed. Tests of equality of variances were used as a pre-test for a two-sample comparison of means (paired), and were followed by the tests for equal and unequal variance. Subsequently, when necessary, a two-sample comparison of means test (paired) with Welch's correction for unequal variances was applied. The final results are presented individually for each Hypothesis in Tables 5 and 6. Table 7 summarises the results of the tests for hypotheses 1-7.
An identical approach was applied in the case of testing the properties of the aggregated crime samples. A pre-test and Ttest or F-test were used to test the validity of the supposition that changes in crime will be consistent in New York City and London, as well as to explore further the properties of the aggregated samples as presented in Tables 8-13.   4 https://www1.nyc.gov/site/nypd/stats/crimestatistics/citywide-crime-stats.page 5 https://www.met.police.uk/sd/stats-and-data/met/crime-datadashboard/ 6 https://www.covid19tracker.health.ny.gov 7 https://www.coronavirus.data.gov.uk The mean of the aggregated crime incidents in the month t (t = 1, 2, 3, 4 for March, April, May, June, respectively) was compared for 2019 and 2020 for New York City and London, where: H0 is the mean of the crime incidents across all seven categories in London in the month ( = 1, 2, 3, 4 for March, April, May, June, respectively) in 2019.
In continuation, the mean and variance statistics tests (t-stat and F-stat) for the crime incidents in NYC and London, respectively, were conducted for 2019 and 2020, for equal variances.
Represented numerically: As a follow-up, for unequal variances: Subsequently, a variance statistics test was conducted on the samples from 2019 and 2020 across the seven crime categories, using the variance comparison test as a pre-test for a two-sample comparison of means test (paired), where: For Hypotheses 9-15, the OLS regression method was applied to test the correlation properties: Where: 1 -is the number of COVID-19 cases in New York City in the month ( = 1, 2, 3, 4 for March, April, May, June respectively) of 2020.

And lastly,
Where: is the average number of crime incidents across all seven categories (homicide, rape, robbery, violence, burglary, theft, vehicle theft) in New York City in the month t (t = 1, 2, 3, 4 for March, April, May, June respectively) of 2020.
is the average number of crime incidents across all seven categories (homicide, rape, robbery, violence, burglary, theft, vehicle theft) in London in the month t (t = 1, 2, 3, 4 for March, April, May, June respectively) of 2020.

░ 5. FINDINGS
The findings are presented for both cities and the periods under scrutiny in the established order, that is, New York City followed by London and crime classification according to the rationale listed in the introduction section. The presentation opens with a monthly and total crime incidence on an absolute and rate basis, year-to-date change in the relevant months of 2019 compared to the same period in 2020.   Table 3: London: a comparison of crime incidents in absolute numbers and crime rate change, March through June 2019-2020, respectively (own rendering, processed from the source 9 ). Table 2 shows that, contrary to the assumptions, the number of incidents in New York City in three crime categories: murder, burglary and vehicle theft, did not drop after the introduction of lockdown. In comparison to March 2019, not only did the records show a higher volume of offences in March 2020 but they confirm that in these three categories, crime continued to rise throughout the lockdown period, month-on-month. While the number of observations for murder is small and therefore sensitive to any changes, which makes the relative comparison dubious, the total absolute number growth in these three categories: murder (+33 incidents), burglary (+1717) and grand larceny auto (+2788), respectively, came to +4538, and alone constituted a rise of 87% in the four months in question (4538/5158). On the other hand, the total absolute number decrease in rape (-223), robbery (-704), assault (-1003), grand larceny (-5222), respectively, came to -1930 and constituted a drop off of 14% (-1930/13788). In the case of robbery incidents, the downward trend did not come into effect until April. As an aggregate, the number of crime incidents in the four months under consideration went down by 4504 and almost 15% (-4504/30500) respectively, year-on-year.

Data Presentation
As can be inferred from  Table 4: London and New York: crime incidents 2019-2020, relative change year-to-year (source: as in Table 2 & Table 3).
The year-to-year four-month comparison indicates that, with the exception of murder and burglary and car theft in New York City, in all the other crime categories as well as in burglary in London, the lockdown was associated with sudden drop offs in crime, with the total mean reduction of 15% and 31% in NYC and London, respectively, throughout the fourmonth period. The relative change in each month was respectively: 4.7 times, 1.5 times, 1.7 times and 4.3 times larger in London as compared to New York City. The lockdown was introduced a week later in London than in New York City, but it seemed to exert an almost immediate negative (i.e., preventive) effect on the crime rate in the former. It can be explained by the fact that the almost overnight result of the stay-home recommendation increased the presence of capable guardians, that is, the residents, thus preventing burglaries. Indeed, burglary in London remained at an approximately steady level of -35% across all four months, which stood in contrast to NYC. On the other hand, the increase in the rates of murder, burglary and grand larceny auto across all four months in NYC, respectively, allows already to draw an initial inference that Hypotheses 1, 5 and 7 will not be validated for that city.

Hypotheses 1-7: Analysis and Findings
The following 14 figures illustrate the differences in the change trends (functions) for the selected crime categories in the period March through June 2020 as compared to the corresponding period in 2019. The dependent variable in each case is the number of the relevant crime incidents in absolute numbers. The blue curve describes the trend in 2019, whereas the orange curve depicts the corresponding trend in 2020. The graphs are positioned side-by-side, city-to-city, to illustrate the trend development in each crime category comparatively. The y-axis has been adjusted to scale for each plot, given the variation in the count of recorded incidents, so that the presentation was consistent in size.    Table 2; the source for Figures 3, 5, 7, 9, 11, 13 and 15 is as in Table 3.            As can be inferred from Table 5, Hypotheses 2 and 6 (rape and grand larceny) have been validated. Hypothesis 3 (robbery) shows a p-value of 0.052, while Hypothesis 5 (burglary) indicates a p-value of 0.051. These hypotheses should be retested on a more detailed sample (daily crime records) to draw more precise conclusions. In the case of Hypothesis 7 (grand larceny auto) the p-value is smaller than 0.05, however the tstat coefficient is negative, therefore this hypothesis cannot be validated. In fact, there was a rise in this crime category, as indicated by the results in Tables 2 and 4. After the equality of means tests analysis it can be concluded that The Spring 2020 Lockdown in NYC only had a clear crime reducing effect in the rape and grand larceny categories. Nevertheless, the data does not provide enough support to formulate the opposite conclusion, i.e., that it is possible to deduce that lockdown had a crime inducing effect on the crime count in NYC.
An analogous procedure was applied in the case of London.  Table 7: Hypotheses 1-7, validation.
In continuation, the same procedure was performed for all crime type incidents grouped together.
The aggregated figure was the total sum of all relevant crime incidents in each relevant month. Thus, in Tables 8 and 9, N is the aggregated number of the crime categories (7). Again, the initial analysis started with an equality of variances test, used as a preliminary test for the comparison of means test (paired) across crime categories, which was conducted in two versions -for equal and unequal group variances (in two samples for 2019 data and 2020 data).  Table 9: London: a pre-test for a two-sample comparison of means test (paired), N is the aggregated number of crime categories.

Month
The results indicate that the mean of the aggregated crime incidents in the respective months and in total in 2019 was not larger than the mean in 2020, both for New York City and London.
There followed a test of comparison of means test in 2019 and 2020 across crime categories using a comparison of means test (paired) for unequal variances (  Table 13: London: a pre-test for a two-sample comparison of means test (paired), N is the aggregated number of crime categories.
The results indicate that the mean of the aggregated crime incidents in the respective months and in total in 2019 was not larger than the mean for the crime aggregates in both cities in 2020.
The equality of variances test results also indicate that the variance of the aggregated number of the incidents in the seven crime categories in 2019 was not significantly different in each consecutive month from that of 2020 in both cities. However, the results of the test conducted on the aggregate for the four months in NYC show that the 2020 variance was smaller than in 2019, hence indicating the smaller dispersion of crime in that year.
The null hypothesis in this test assumes that the variance/dispersion of the number of crime incidents in 2020 does not differ significantly in relation to the previous year, whereas the alternative hypothesis states that the variance/dispersion is significantly lower in 2020 than in the previous year.

Hypothesis 8: Analysis and Findings
An analogous procedure was applied in the case of Hypothesis 8. Firstly, a comparison of crime tendencies in 2020 was presented graphically (Figure 16). Figure 16: A comparison of the crime trends in absolute numbers, New York and London, 2020 (source: as in Table 2 & Table  3). To sum up, in six categories, with the exception of homicide in London, the mean of crime in London was higher than that in New York City throughout the four months, in both 2019 and 2020, whether under lockdown or not. The only exception is the category of murder, which in NYC was higher both in 2019 and 2020. Thus, Hypothesis 8 has been falsified. The crime trends in London and in New York City under the lockdown period are dissimilar. In other words, lockdown should not be perceived as an 'equalising' factor in the analysis of crime trends, and certainly cannot be treated as such in isolation from other variables.

Hypotheses 9-16: Analysis and Findings
Regression analysis was conducted on the crime data from New York City and London, accessed from the source listed in Footnote 6 (this was the source for Figure 17, and, together with the previously listed NYPD data, the source for Figures: 19, 21, 23, 25, 27, 29 and 31) and from the source listed in Footnote 7 (this was the source for Figure 34, and, together with the previously listed MET data the source for Figures: 36,  38, 40, 42, 44 and 46) from mid-March through June 2020.
Although, as pointed out earlier, there had been concerns expressed by numerous statisticians and other scholars that the calculation methodology and presentation of the COVID cases may not have followed the expected standards, that is the case count was not rigorous and may have included cases counted twice or more times, the data for this experiment had been assumed as credible since they were published by the respective governmental bodies.
The analysis involved an OLS regression test. The independent variable in these tests was the number of COVID-19 cases and the dependent variable was the number of crime incidents in each category. The tool applied for these tests was the OLS method, with four observations for each test, representing the number of months (Figure 18-31).                 The results from the OLS calculations applied to New York are presented in the Table 18. There followed an identical analysis for the data from London. The results, as applied to London, are presented in Table 19. The individual and comparative results, for both cities, are discussed after the presentation of the graphic and numerical findings.             The OLS results are presented in Table 19.  In view of the p-values, none of the 14 OLS models turned out to be statistically significant. The smallest p-value is that for murder in New York City (0.178).
New York City: The analysis of the models demonstrates that the function coefficient is negative for: rape, robbery and grand larceny. This indicates a negative correlation between the variables; as the number of COVID-19 cases grows, the number of crime incidents in the corresponding crime category drops off. The function coefficient for the other four crime categories is positive, which means that as the number of COVID cases grows, the number of crime incidents in the categories under scrutiny also grows.
London: The analysis of the models demonstrates that the function coefficient is negative for: homicide, violence against a person, burglary, robbery, theft and vehicle theft. The inference therefore is that there is a negative correlation between the number of COVID case and the number of crime incidents in those categories. This means that as the number of COVID cases grows, the number of crime incidents in the categories under study drops off. Only in the case of rape is there a positive correlation, that is as the number of COVID cases grows, the number of rape incidents in London goes up.
If it is accepted that there indeed is a correlation between the number of COVID-19 cases and crime reduction (Hypotheses 9-15), it can be concluded that three of these hypotheses have been validated in New York City: hypothesis 10 (rape), 11 (robbery) and 14 (grand larceny). In the case of London, six of the hypotheses have been validated: 9 (homicide), 11 (robbery), 12 (violence against a person), 13 (burglary), 14 (theft) and 15 (vehicle theft). These results are consistent with the Hypotheses 1-7 validation process. ░ 6. DISCUSSION AND CONCLUSIONS

Discussion
As exemplified by the results of this research study, the introduction of lockdown can be deemed as to produce a reduction in certain crime categories in both cities, with an overall 15% and 31% drop in New York City and London, respectively. The initial drop off in crime was more pronounced and sustained in London. The relative change in each month there was respectively: 4.7 times, 1.5 times, 1.7 times and 4.3 times larger in London as compared to NYC. However, several findings seem to distort this seemingly unambiguous picture. Murder, robbery, assault, burglary, and grand larceny auto did not experience a drop in NYC. Neither did violence against a person in London. The incidence of murder (homicide) should probably be excluded from the conclusions, as there are few observations (more for NYC) and conclusions in this regard may be distorted. On the other hand, the fact that 33 more persons in 2020 fell victim to murder in NYC under the Spring Lockdown than in the same period a year earlier, should certainly provoke further investigation.
As already stated, the year-to-year four-month relative comparison based on the results presented in Table 4 indicates that the crime trajectory in London during the Spring 2020 Lockdown indeed occurred in accordance with the theoretical foundations put forward by the RAT and subsequent contributors, even given the relatively short period of observation. However, the relative changes to crime in NYC proved to be contrary to the expectations, since the initial fall in crime occurred distinctly only in the category of rape, assault and grand larceny. Robbery incidents were on the rise in March although since April they followed the expectations. However, the hypothesised decline in burglary and grand larceny auto -anticipated to fall due to the (expected) increased guardianship, intensified by the stay-home ordersnot only did not go down but, in fact, increased by almost 89%. As already stated, such a change could be seen as contradictory to the long-established assumptions, even despite the short period of observation. The sources of this detrimental change would therefore need to be analysed in conjunction with the lockdown and other hypothesised variables. The lockdown variable itself would need to be analysed and deconstructed thoroughly if credible answers were to be found. Given the brevity of the period of observation, the analysis may be subject to multi-angled criticism, since it seems to stay in contrast with some theoretical foundations as well certain findings in contemporary literature.
The results drawn from the inferential statistics tests allow for the rejection of the null hypotheses 2 and 6 (rape and grand larceny) for NYC, and the null hypotheses 1, 2, 3, 5, 6 and 7 (homicide, rape, robbery, burglary, theft and vehicle theft) for London. The results of these tests allow to support the conclusion that lockdown -in accordance with the expectations -did act as a possible deterrent as far as some categories of crime are concerned. However, there are few similarities between the cities in this aspect. The results of the statistical inference tests indicate that the dispersion of crime between the respective years and between the cities do not allow to draw a conclusion that the crime processes showed similarity in the two cities. It is therefore unjustified to ascribe the changes in crime trajectories in the respective cities in the four months under investigation to the lockdown itself, although -as already pointed out -the course of crime development in London may tentatively indicate an initial negative correlation (as lockdown is introduced and tightened, the number of the selected crime incidents falls).
The additional correlation tests which attempted to establish a relationship between COVID-19 and crime, where COVID-19 cases were assumed to be a proxy variable, seem to confirm those initial conclusions although there are several limitations regarding the correlation analysis.

Limitations of the Study
One of the most important limitations in this study is the number of observations. Both data sets presented the data in monthly intervals only. This made the data presentation and testing overly simplistic. It was not possible, for example, to correlate the exact dates of the lockdown introduction in London (23 March 2020) and New York City (16 March 2020) with the corresponding week in the crime data. Similarly, it was not possible to correlate the date on which lockdown started to be lifted: 13 June 2020 in London; 8 June 2020 in NYC and 15 May 2020 in the state of New York. The sparse frequency of the data did not allow for more sensitive and reliable testing of the RAT. Daily data would allow for a more precise test, such as the SARIMA test, used by Ashby [9].
Another issue is the overt generalisation used to establish the variable 'lockdown'. In reality, this variable should be deconstructed into several partial variables, such as 'stayhome' orders, 'work from home' recommendations, 'fill in public transport to a reduced capacity' orders, etc. A question remains whether these variables could be quantified. Nevertheless, building a model which could accommodate such a sophisticated break-down is beyond the feasibility of this study.
One more limitation of this study is the fact that it only presents a limited number of statistical results, derived from processing a selected number of publicly accessible sources. It is possible that other research tools as well as data sets could enrich the propositions put forward in this study, for example a comparative study focused on the 'strict lockdown' as opposed to 'soft lockdown/no lockdown' countries and their crime data from 2020. Another proposition could be a case study, focused on Sweden alone, as a country with virtually no or few Spring Lockdown procedures or, finally, a possible comparison of Sweden and another country, for instance, The Czech Republic (these two countries are similar in the population size but reacted very differently to  in regard to crime levels in Spring 2020. It could also be interesting to compare the crime data from Sweden with those of Germany and the UK, as representing two countries with considerably strict lockdown measures. Again, such a research study is beyond the framework of this paper but might be suggested as a possible research avenue.

Concluding Remarks
The mixed findings which emerged from this study follow -to some degree -the body of the recent literature on the relationship between lockdown and crime change. The authors, whose contribution to criminology was summarised in Table 1, also presented mixed results. Most of those papers documented the drop off in the crime rate after the introduction of lockdown, although not universally. In the case of London, the theoretical foundation, that is, the proposition that an increased level of guardianship which serves as a protective element and reduces the number of burglaries, was confirmed by the empirical evidence almost to the letter but several of the results observed in New York City stay in contradiction to the RAT.
There was no homogeneity in proving the overall 'beneficial', that is the crime reducing effect of lockdown. The findings presented in this paper cannot be perceived as unconditionally supporting such conclusions albeit they are drawn from a sample of two cities only and cannot be viewed as representative.
The similarity of the sample statistical properties between the years 2019 and 2020 (with the exception of the aggregated total crime count in NYC) and the dissimilarity of the crime processes between the cities demonstrated by the comparison of means tests allows to infer that the criminal patterns remained similar to the pre-lockdown conditions and characteristics. It is true that there was a general drop off in crime incidents. However, it is impossible to speculate that the cost of law enforcement maintenance during those four months of the Spring Lockdown in 2020 must have decreased accordingly. If we take into consideration the fact that the number of calls for police service increased (Table 1) and the police servicing of the lockdown, that is quarantine checks, monitoring social distance restrictions, controlling the outlets that remained open etc. the costs of police operations during the period under study might have stayed similar or higher.
One of the quantifiably justified positive conclusions regarding London is that there were lives saved -as the number of homicide incidents decreased while rape offences occurred less frequently in both cities -but, sadly, violent crime, such as murder and robbery was on the rise in New York City consistently throughout the period under study. It is possible the protests and the ensuing commercial outlets break-ins and looting were partly responsible for the growth in crime in NYC in the latter part of the Spring Lockdown.
Therefore, it seems that the results of this study present a cautionary tale. Rather than adding to the evidence in favour of the grand criminal theories, they seem to support the argument by Stickle and Felson [13], discussed in the literature review, namely that the size and intensity of the lockdown and its subsequent impact on crime -otherwise known as the largest criminological experiment in historyposes a challenge to criminologists in their quest to explain this unprecedented lockdown-provoked upheaval in crime trends. In consequence, the conclusions drawn from the comparative analysis of the crime trends in NYC and London allow to support one of the arguments presented by Stickle and Felson [13], that is that the changes in crime under lockdown are uneven spatially, chronologically and in the matter of category rather than unequivocally confirm the propositions of the RAT.
Perhaps the social, economic and civilisational changes which have permeated the world since the 1980s are presenting a new challenge which would allow enriching the RAT and the follow-ups to the theory. It would certainly seem that the lockdown phenomenon cannot be perceived as a single 'breach' or be compared to a war or economic depression. It has affected every single aspect of life -distorting work, education, transport and communication, leisure and holiday patterns as well as negatively influencing mental health, and curtailing civil liberties and human rights, to mention some. Similarly, it has resulted in an asymmetric crime evolution. Its imprint should be analysed from every angle. No theory can comprehensively explain lockdown induced changes in crime volume and characteristics, as they are bringing along new challenges and costs. The phenomenon of lockdown and its consequences on crime development may require sophisticated tools and a large pool (drawn from different countries) of data which could be processed comparatively. Such a comparative analysis, conducted on a data set encompassing a year, could throw light on the more detailed, precise and verifiable effects of the lockdown on crime.

Recommendations for Further Studies, Policy Makers and Law Enforcement
The primary recommendation for the policy makers is to weigh the pros and cons of a lockdown very carefully. The recommendation for the law enforcement officers is to estimate the costs of a possible re-introduction of lockdown, particularly in view of the evidence from New York City. Policy makers and the law enforcement personnel are also recommended to consider and estimate the costs of maintaining lockdown, especially in view of the temporary character of the lockdown cost reducing effect.
There are several research directions which could be drawn upon this study, some of which have already been listed. It would be interesting to decompose the variable 'lockdown' and relate it with the reduction in the number of burglary incidents, based on the results from London. A cost-benefit assessment of maintaining lockdown vis-a-vis expected savings from the reduction in crime is another possibility. Finally, it is recommended that a more comprehensive comparative study should be carried out. It is advised, however, that large, detailed and reliable sources of crime data, which would present it at more frequent intervals, be accessed in such future endeavours.

░ ACKNOWLEDGEMENT
A valuable inspirational and analytical contribution to this paper has been the 2020 MSc in Crime Science dissertation of Caroline Byczynski, conducted at University College London (UCL).