During the course of our study, we met numerous limitations that restricted our research. First, the census tract shapefile we downloaded from Chicago Data Portal did not have population data in its attribute table. We had to download the population data as a separate Excel file. Problems arose when we attempted to perform a table join, in which resulted ArcMap to crash in three different attempts. This might due to a limitation in the software (ArcGIS) or a limitation in hardware (inadequate processing power or ram). We resorted to use a third party sourced census tract for Chicago, that had no metadata and unreliable x,y coordinates. We were able to deduce the column that represents population, however, without reliable x,y coordinates we were unable to conduct risk adjusted nearest neighbor hierarchical spatial clustering analysis. Therefore we only conducted standard nearest neighbor hierarchical spatial clustering analysis that does not take population into account. With reliable x,y data we would definitely be able to better describe crime patterns in Chicago at the local level. The second limitation we ran into during our research was the lack of x,y coordinates in the four crime layers. Crime events that lack x,y coordinates account for 10-18% of the total respective crime categories. This could prove statistical significant when conducting kernel density, fuzzy mode, and hierarchical nearest neighbor clustering analysis.
The third limitation we ran into was normalizing crime data in respect to sport stadiums. In this project, we wanted to prove Bernasco and Block’s theory which suggest popular public venues like major sport stadiums act as sites that facilitate criminal activities (2010). Unlike CTA train stations, sport stadiums do not open 24/7 or all year round, they only open during their respective seasons. For example, Soldier Field is a football stadium that serves as the home field for the Chicago Bears. It only opens for half of the football season (each team in the NFL play in other stadiums as away games for half of the season) which ranges from August to early January (depending if the Bears make the playoffs). On top of that, it only opens on Sundays, Monday, or Thursdays for home games. Furthermore, there are five professional sport teams that compete in the big four sports of North America. Therefore in order to normalize crime data in respect to sport stadiums that can be compared to the overall crime density in Chicago or crime density around CTA train stations, one would have to obtain each team’s schedule for 2016 and intersect the schedules with raw crime data we downloaded from Chicago Data Portal. This requires a great deal of data mining which is far beyond our expertise.
In conclusion, our analysis did partially prove Bernasco and Block’s theory of crime attraction. From our analysis, for all four crime categories, crime density around train stations was much higher than the city average. Crime density also increased as the buffers got closer to the train stations, indicating that train stations in Chicago indeed act as crime attractors. However, we were unable to prove sporting stadiums as crime attractors since crime density around stadiums was much lower than the city average in all four categories. In addition, crime density decreased as the buffers got closer to stadiums. However, this does not prove that stadiums are not crime attractors. As explained in the limitation section, it was very difficult to normalize crime data in respect to stadiums. But with the right skill set (data mining for example), one may able to get the right data and conduct a more accurate analysis in regards to sport stadiums as crime attractors. To further this research, one may look at the correlation between certain crimes and certain crime attractors. In this research, for example, we found that train stations attracted more crimes like robberies and criminal damages than homicides and narcotics. Other crime attractors like ATMs, community centers, or the Lower Wacker (in Chicago) may exhibit different trends. Furthermore, one may consider to incorporate demographic data, which can potential play a role in crime attractor analysis.