The COVID-19 pandemic and first lockdown orders led to rapid changes in everyday routine activities, which had direct effects on opportunities for crime (Nivette et al., 2021). Immediately after the first stay-at-home orders came into force in many countries in March 2020, crime researchers noted that various forms of property and violent crime had suffered a notable decrease due to the reduced opportunities for offenders to converge with targets in physical settings (Abrams, 2021; Ashby, 2020). Simultaneously, others highlighted that some forms of cyber-enabled and cyber-dependent crime had increased due to the growth in internet use for work and leisure (Buil-Gil & Zeng, 2021; Kemp et al., 2021; Lallie et al., 2021). After the first months of the pandemic, some of the social distancing restrictions were relaxed and rates of offline crime began to bounce back to pre-COVID trends (Balmori de la Miyar et al., 2021; Langton et al., 2021), but there is a gap in research about the mid-term effect of multiple lockdowns on online crime, and few previous studies have compared online and offline crime using the same dataset. To fill these gaps, the present paper analysed crime data recorded in Northern Ireland between April 2015 and May 2021 to analyse the short-, medium- and potential long-term effects of each lockdown on various forms of offline and online crime.
We identified that not all crime types were affected in the same way by the lockdown restrictions. Firstly, we observe that most forms of fraud and cybercrime rose rapidly during the early months of COVID-19 and continued growing up until May 2021. With the exception of drug trafficking, none of the traditional, offline crimes analysed above experienced clear increases during the pandemic, and thus fraud and cybercrime represent an exception to the overly simplistic view that crime decreased during COVID-19. The other crime type which also likely saw increases during the pandemic was domestic violence (Piquero et al., 2021), though some researchers note that it quickly returned to pre-COVID levels when lockdown restrictions were eased (Nix & Richards, 2021). Our data did not allow us to explore trends in domestic violence. In the case of fraud and cybercrime, there were notable differences across crime types. While recorded consumer fraud online, cyber-dependent crime and other fraud experienced notable growth from the first lockdown up until May 2021, a similar increase was not as clear in the case of consumer fraud offline, which decreased after some of the COVID-19 lockdowns. Investment and advance fee frauds, which can be cyber-enabled or not, appear to have decreased when the first lockdown came into place and possibly increased after that. It is possible, if not probable, that those forms of fraud that are enabled by digital technologies rose substantially during the pandemic (Buil-Gil et al., 2021), while non-cyber-enabled fraud suffered little variation or decreased (Kemp et al., 2021). The increase in everyday routine activities that people carried out online, often from under-protected home computers, including remote working, online shopping and online gaming, contributed to an almost-immediate increase in suitable targets on the internet during the months following March 2020. It is also probable that some ‘motivated offenders’ adapted their strategies to take advantage of new online opportunities, but many cyber-enabled and cyber-dependent offences require a set of technical skills which could not have been acquired in such a short time period. It is likely that the increase in cybercrime is due to the combined effect of a generalised increase in online targets of crime and a few offenders who successfully adapted to online opportunities during the pandemic, but further mixed-methods research should be conducted with offenders to analyse if there was a displacement of crime offending from offline to online environments. Moreover, we do not see any indication of cyber-enabled and cyber-dependent crime returning to pre-COVID trends. While our data only allow analysis of changes in crime up until May 2021, it will be important to study cybercrime trends during late 2021 and early 2022 to explore the possibility of a long-term increase in digital offences, which is of clear relevance for policy, practice and academic debate. As some have noted, the increase in online gaming, teleworking, meetings, online shopping and online dating may extend beyond COVID-19 (Nurse et al., 2021; Ofcom, 2021), thus creating new crime opportunities and accelerating the long-term upward trend in online crime.
Secondly, recorded violence, drug crimes and theft from persons experienced immediate drops when each of the three lockdowns came into force, and crime rates then returned to pre-COVID levels after lockdown orders were relaxed. These offences take place primarily in physical places that experienced decreases in mobility during each lockdown, and thus the opportunities for offenders to converge with suitable targets decreased with lockdown restrictions, and crime then returned to normal trends when social distancing measures were eased. Interestingly, while most forms of violent crime appeared to return to the same levels seen before the pandemic, drug trafficking not only bounced back to pre-COVID levels, but rates rose and remained well above those seen before the pandemic. Similar results were found in a study that analysed drug seizures over time in the United States (Palamar et al., 2021), and Langton et al. (2021) observed a peak in drug crime in England and Wales in May 2020, though a similar increase in drug crime was not observed in other countries (Balmori de la Miyar et al., 2021). It is still unclear whether recorded drug trafficking increased as a result of real growth in the supply of drugs or because law enforcement prioritised this type of crime during that period. Drug trafficking could have become more visible with less people walking the streets, or, alternatively, the rise in police recorded drug crime could be the result of greater police resources being dedicated to detecting these crimes. In contrast to violent crime and drug crime, recorded theft from persons decreased with each lockdown and then increased slightly, but rates in May 2021 were still below pre-COVID levels. Could this be because people who become involved in crime returned to the streets quickly after lockdown restrictions were lifted, but those not involved in crime did not leave the home as often as before COVID-19? That would explain why violent crime, in which two persons may mutually become involved in the incident, quickly returned to pre-COVID levels, while theft from persons, in which a crime target is needed, did not return to crime levels seen before the pandemic. That would also explain why residential burglary remained well below pre-COVID levels even in May 2021, due to the continued overall increase in the time spent by residents (capable guardians) at home. Further research is needed to identify changes in offender and victim activities.
And thirdly, some of the crime types with the most obvious seasonal patterns, including public order offences and possession of weapons, criminal damage, and bicycle theft, all of which tend to occur at much higher rates during summer, show a very similar seasonal variation during the pandemic. Crime records decreased with the first lockdown (March 2020) and increased during summer, after the second lockdown (October 2020) crime started to decrease during autumn and reached minimum levels with the stay-at-home order of the third lockdown in winter (January 2021), and after winter crime records began to increase again. Given the close correspondence between the traditional seasonal patterns in crime and the lockdown periods, it becomes difficult to fully comprehend the extent to which changes in crime are due to a continuation of pre-COVID crime seasonality or lockdown restrictions. In the case of criminal damage and bicycle theft, our model results—both the ITS and ARIMA coefficients—provide some support to the hypothesis that social distancing orders significantly affected crime trends, and thus we can expect that changes in crime are due to the combined effect of lockdown restrictions and seasonal variation. This is particularly evident in the case of bicycle theft, which displays a large, unusual decrease during the months following the second lockdown, when further restrictions related to the closure of cafes, hospitality and non-essential shops were introduced. The trend in shoplifting during the pandemic retains some similarities with that of bicycle theft, but both are markedly different from all other offline crimes. Shoplifting records suffered a very large drop immediately after both stay-at-home orders and progressively returned to pre-COVID levels during the following months, but with lower levels recorded by the end of the second lockdown (November/December) than when the second lockdown came into place in October. This is also likely to be the result of the stricter restrictions in place by the end of November, when cafes, hospitality and non-essential shops were closed. This is also shown in our ARIMA models, which indicate large decreases in both bicycle theft and shoplifting by the end of the second lockdown and with the third COVID-19 lockdown.
Our study also identified that not all COVID-19 lockdowns in Northern Ireland had the same effect on crime. The first lockdown, which was defined by a stay-at-home order and restrictions on all non-essential social and business activity, had an overall negative effect on most types of street crimes, due to a reduction in opportunities for the physical convergence between offenders and suitable targets. Similarly, the stay-at-home order of the third lockdown, in January 2021, had evident effects on mobility and crime opportunities. In contrast, the effect on crime trends of the second lockdown, which involved the closure of schools, universities and the hospitality sector but not a stay-at-home order, was less evident and non-significant in many cases. The trends in crime during the months following each lockdown also varied. In the case of the first and third lockdowns, most offline crime types suffered an immediate decrease and then progressively returned to the pre-COVID trend, while we observe the opposite trend during the months following the second lockdown. Records of some crimes, including bicycle theft and shoplifting, were lower by the end of the second lockdown, in November/December 2020, than in October. This is at least partly explained by the hardening of COVID-19 restrictions in late November, when the Northern Ireland Government imposed the closure of cafes, hospitality, non-essential shops and gyms. A similar pattern is observed in the case of fraud, where those fraud types that can take place offline suffered a decrease at the end of the second lockdown due to the additional social distancing measures, while online fraud experienced an increase in December 2020 due to the closure of shops and businesses and increased online shopping over the festive period.
While the findings presented in this article are first-of-its-kind and contribute to the criminological literature about the short-, mid- and long-term effects of rapid social changes on crime (offline and online), these are not free of limitations. The main threat to the validity of our findings is related to the use of police-recorded crime statistics as a primary source of data. Police-recorded crime data are known to be severely affected by measurement error arising from underreporting and underrecording, and it is yet unknown the extent to which the COVID-19 pandemic not only affected crime but also the measurement properties of crime statistics (Wallace et al., 2021). This may be particularly problematic in the case of cybercrime, given the low reporting rates that define these offences (van de Weijer et al., 2019). Future research is needed to explore if crime reporting and recording practices changed during COVID-19, thereby illuminating the extent to which research using police-recorded crime data to study changes in crime may be affected by measurement error. Moreover, due to data availability we analyse changes in crime across months, which may mask internal heterogeneity across days and weeks. Future research should analyse smaller temporal units of analysis where possible.