Statistical Results

The statistical results were quite effective in setting the context in the city wide scale. Several assumptions can be made just by analyzing the crime types. The increase of B&E Commercial and Mischief (which includes property damage) in the lockdown year 2020 can be caused due to the shutdown of establishments that have been emptied out. These establishments would be left empty during this period which could have been broken into and damaged. B&E Residential decrease over the years could be due to more people staying at home and transitioning into work-from-home which makes it harder for criminals to break into homes that are occupied. Auto Theft decrease can be caused by the reduction of number of vehicles available out in the open and in the streets. Because, most people are at home during the lockdown year 2020 and restriction year of 2021, the cars are mostly parked in their garages. This would make the theft of vehicles much more difficult. The large decrease in regular thefts can be explained by the reduction of people outside which reduces the chance of thefts. Thefts tend to often occur in busy places and during lockdown 2020 and restriction 2021, there would be significantly less crowded areas.

Summarize Percent Change

This analysis was done in the neighbourhood scale where each neighbourhood showed increase and decrease trends in crime. We must be very careful in overanalyzing and critically interpreting the results here as the percentage change in raw crime counts can often be sensitive to minor changes and outliers. For example, an increase from crime count of 1 to 2 would result in a 100% positive change and 2 to 1 would be a 100% negative change. This is why the count labels and the exact perchance change was added into the resulting map to reduce these effects. B&E Commercial increased considerably during 2020 (Lockdown) in neighbourhoods that encompasses stores and establishments that are often along major roads and neighbourhood boundaries. Downtown, Kitsilano, Fairview, Mount Pleasant, Arbutus Ridge, Kerrisdale, Renfrew-Collingwood, Kensington-Cedar Cottage all contain major road networks like Oak Street, Main Street, Broadway Street, Robson Street, Burrard Street, Granville Street, and etcetera. In 2021 (Restriction) where these establishments were reopened, the B&E Commercial reduced in most neighbourhoods. Neighbourhoods with many smaller establishments that was further away from the central hub of the city reopened much later or remained closed throughout 2021, which can explain the increase in B&E Commercial in Victoria-Fraserview and Killarney. These neighbourhood results of B&E Commercial mirror the statistical results of increase in 2020. B&E Residential in 2020, unlike the statical results, showed increases in West side of Vancouver. Kitsilano, Arbutus Ridge, West Point Grey, and Shaughnessy all had increase in B&E Residential while the East side of Vancouver, West End, and Downtown mostly had declines. It looks like neighbourhoods with apartments and condos had decreases but neighbourhoods with houses had increases in B&E Residential. 2021 had all neighbourhood except West Point Grey decrease in B&E Residential. Mischief in 2020 lockdown year was similar to the B&E Commercial neighbourhoods and saw an increase in mostly those same areas. Because VPD includes property damage in Mischief, it explains why it is closely connected to B&E Commercial and the statistical results. Auto Theft decreased in both COVID years in all neighbourhoods except in Marpole in 2021 where it had a small increase. The decrease in Auto Theft in all neighbourhoods confirms the theory in the statistical results that explained the decrease in Auto Theft due to less cars in the streets and inside garages. Theft also saw an overall decrease in both years. This can also be explained due to less people in crowded areas targeted by pickpockets and thieves. There is an unusual increase in Theft in Renfrew-Collingwood in 2020 and further increase in 2021. The counts and the percent change is not small which means that it has less chance of being a coincidence or an outlier. This is a notable increase that can be further analyzed in future studies to understand why this increase of Theft is occuring in Renfrew-Collingwood.

Average Nearest Neighbor

Every year generally had an increase in the z-scores (less clusters). These results were expected because of the general decrease in crime trends. The exception in B&E Commercial and Mischief in 2021 were the z-score decreased instead meant there was more clustering. This was also expected with the running theme of the relation between B&E Commercial and Mischief to the lockdown year.

Kernel Density

The spatiality of the clusters and its changes were generally similar to the statistical results and the percent change results. As B&E Commercial increased in 2020 lockdown year, there were increases in kernel density in the major urban regions and roads that harbour establishments. The largest cluster is in Downtown as this is the neighbourhood with the most number of establishments that could be broken into. For B&E Residential, the density all around the city is decreased slightly in 2020 and further in 2021 which is explained by the decrease in overall B&E Residential around the city in 2021. The slight increase in B&E Residential in the Westside of Vancouver in 2020 in the percentage change results can be observed here with the small density increase in the West regions. Unlike other analyses results, Mischief showed no major density patterns and trends. Auto Theft, expectedly deceased in density significantly all around Vancouver in both 2020 and 2021 COVID years. Theft density dropped accordingly to the statistical decrease over the years. Again, Theft patterns and density only show in parallel pattern along the major roads.

Regression (GLR & GWR)

In the GLR analysis, decrease in AICc score during 2020 and 2021 hints that the relationship between the variables became stronger with COVID years, however the Deviance Explained remained mostly same. What is certain is that the AICc scores in the GWR was significantly less than the GLR model. The large difference in the AICc values between GLR and GWR shows us the benefits of using the GWR that better fits the local model in explaining local spatial dynamics and relationships. The absence of significance of the unemployment rate variable that resulted in 2019, during 2020 and 2021 meant that unemployment rate became a non-factor in the relationship to crime in COVID years. However, the significance of unemployment rate variable in 2019 is quite questionable as the GWR result maps show a standard deviation range of mere 0.17 to -0.09. Nonetheless, unemployment rate variable is more dependant on crime in Downtown and South part of Vancouver. As for the rest of the GWR maps, higher numbers in standard deviation and darker tones of blue is associated with dependency of the explanatory variables to the five crimes and lower numbers in standard deviation and darker tones of red is associated with more independence. For example, with population age 15~64 variable in 2021, for every 1 unit increase, there is a 2.24 to 20.88 increase in the five crimes. In the local model, this means that there is almost ten times the value in some parts of the city. With population age 65+ variable in 2019, the GLR coefficient value was around 14.03, however, the range in the GWR deviation results showed -7.19 to 26.89. This is an example where the limitation of the global model of GLR is quite apparent. It fails to highlight where the contribution to population age 65+ is much stronger or weaker than other areas. Generally, there is an independence between the variables in all of South part of Vancouver in all three years and a positive relationship and dependance of the variables in the Eastside of Vancouver. In the corner of Dunbar-Southlands and Kerrisdale, a positive relationship and dependance appears during year 2021.

Limitations & Future Analysis

Due to time constraints, the GLR and GWR was conducted on the five types of crimes that were bundled. In the future, regression model using each types of crime in each year might show a better relationship, correlation, and intricacies between the variables. Also, Moran’s I spatial autocorrelation using the residuals was skipped after the GLR as similar cluster analysis of Average Nearest Neighbor was already performed. However, performing Moran’s I and comparing it to the Average Nearest Neighbor might show new phenomena. Other population characteristics that could have been used as explanatory variables would have been desirable. As more time passes, updated data for the census, and crime data for 2022 will be completed. These data can be valuable in detecting further trends. Lastly, the VPD crime data did contain a time and date field. Unfortunately, they were divided into separate columns of minute, hour, day, and month and not under one time and date field. The limitations of knowledge in combining these divided fields into one time and date field meant that interesting temporal analyses were missed.

Conclusion

Overall, despite several reports of crime decreasing during lockdown but increasing again after the lift of lockdown in various parts of the world, Vancouver saw a decline in crime in both counts and density in 2021 (Restriction). 2020 (Lockdown) had the expected decrease in crime but unexpected increase in B&E Commercial and Mischief due to establishments that were closed down with no security. B&E Residential, Auto Theft and Theft all decreased with more people at home, cars parked in garages and less gathering of people. The statistical results gave the effects of COVID in crime on the city scale, while the percent change analysis focused in on the neighbourhood scale. B&E Residential showed overall decline in 2020 on the city scale, however, we noticed the increase in B&E Residential specifically in many family housing neighbourhoods contrasting the apartments and condos abundant neighbourhoods that saw a decline in B&E Residential. This showed the importance of neighbourhood scale analysis. Similarly for the GLR and GWR models of regression, the GWR analysis proved more fitting in explaining the local relationships that occurred around Vancouver. In the future, studies should acknowledge the benefits in analyzing local scales and changes as opposed to solely conducting city level analyses. It appears that COVID effects crime in multiple scales and this should be remembered for further analysis. Also, more analysis in the neighbourhood scale for crime trends effected by COVID in Vancouver will greatly benefit the policing strategies to avoid over-policing in certain neighbourhoods. For example, large increase in Theft over the lockdown year of 2020 and further increase in 2021 in Renfrew-Collingwood can be analyzed to find the cause for future crime reduction and prevention. The patterns of crime in Vancouver has changed due to COVID and because we do not know when or if the COVID pandemic will ever completely end, these future studies will be imperative moving forward.