Discussion

Nearest neighbor analysis

Before using the nearest neighbor analysis tool in Crimestat, it was expected that crimes would likely be clustered together, occurring within a similar area, resulting in certain neighborhoods experiencing more crime than others. The nearest neighbor results confirm with what was expected, with crimes being clustered instead of randomly dispersed. Building upon this is the nearest neighbor fuzzy mode which displays how many crimes occur within 1000 metres of a point. This provides a visual aid in understanding which areas of the city have a high amount of violent crimes within a small area. The results of the fuzzy mode analysis may differ with a different neighbor radius. Additionally, due to errors when converting the census tract polygons containing population information into point data, nearest neighbor hierarchical clustering was not able to be conducted which may have provided additional important information in second and third order clusters.

Kernel Density

Smooth point estimates created using the ‘Kernel Density’ tool in Crimestat provides a continuous surface of crime density across the study area, as opposed to a hot spot analysis which provides clusters of crimes. Kernel density was employed over other types of interpolation because it is more appropriate for single point locations (Levine, 2010). The kernel density is more useful as a visual aid in understanding how the density of crimes varies across the study area. A dual kernel density would preferably have been employed, but problems in creating point data for population and crime meant this could not be done. Dual kernel density has an advantage over single kernel density because it normalizes the results by the population. With a higher population we expect that there is more chance for crimes to occur, so the resulting continuous surface is more representative of areas which are experiencing a high number crimes compared to the amount of people who have the potential to commit crimes.

Overlay Hot Spot Analysis

Overlay hot spot analysis was used as an alternative method to risk terrain modelling, a system developed to create an understanding of risk based on features of the physical environment. As this program costs money, overlay hot spot analysis was an acceptable alternative. Bars, poverty, and crimes hot spots were overlaid to understand the effects that these variables have on crime in the landscape. Poverty has been linked as a generator of crimes, but bars were a feature of the physical environment that were also successful in presenting areas that may be at high risk of crimes. Kennedy et al argue through their theory of risky places that the features of a landscape have a spatial influence and can enhance the likelihood of a crime. (2018). Bar hot spots are generally in areas that experience high amounts of crime, which is likely not a coincidence. Other variables such as gas stations, liquor stores, bus stops, and drug stores likely also have effects on the distribution of crimes throughout the city, but would need to be explored through a risk terrain model.

Unexplained Neighborhoods

Although the methodologies previously used have been successful in displaying the spatial distribution of crimes for most of the city, the north-eastern district of the city remains largely unexplained by the methods used. The nearest neighbor fuzzy analysis map and the kernel density map both show that this area experiences high amounts of crime, but with a less significant amount of clustering. Instead this area experiences crimes that are spread throughout multiple neighborhoods, and are also occur through consecutive time series as explained by the emerging space-time hot spot analysis.

Race

Race is one demographic variable that has not yet been explored. This was mainly due to the fact that racism is commonly associated with crimes, but is not always the reason that a crime is occurring in that area. However, it is still important to look at race a reason why crime exists in certain areas, and may indicate why so much crime exists in the north-eastern part of the city.

Black or African American population was normalized by total population in each census tract to show the percentage of black people per census tract. The results reveal a clear divide, a high concentration of black population living in the north of the city, where many census tracts have a 86%-99% black population. An article in The Guardian claims that the city of St. Louis has long been divided, but the northern part of the city became the victim of “white flight”, where homeowners sold and relocated once black neighbors moved in (Aufrichtig et al, 2017). This also affected housing prices for middle-class black people who soon left as well, leaving behind a neighborhood with high amounts of poverty (Aufrichtig et al, 2017). The north of the city is also plagued with blighted properties, abandoned homes that again reinforces the idea of the theory of risky places (Aufrichtig et al, 2017). But abandoned homes may only be part of story when it comes to crimes in northern St. Louis.

Conclusion

The methods employed to understand the spatial distribution of crimes across St. Louis City provided useful insight into areas that experience clustering of crimes. It is clear that areas of the city experience clustering and hot spots of violent crimes, but it is still not understood why these crimes occur, especially in the north of the city. Solutions to these problems additionally may not be easy, especially if multiple physical and social factors are responsible. The overlay hot spot analysis was useful in showing how physical features can influence violent crimes, however risk terrain modelling would be preferred as the primary method for understanding risk generators in the city. Additionally the northern part of the city with dominant black community experiences crime dynamics unrelated to race, and likely deeply tied to the history of the area and culture around violence. Police presence and community work in these areas may help to reduce violence, as well as a reduction in blighted properties which has shown to reduce crimes in previous studies (Valasik et al, 2018).