Discussion

The purpose of our project was to see which type of residence was more vulnerable to burglaries. We wanted to investigate which were most susceptible to burglaries among rented or owned homes, and single unit homes or multiple unit homes (ie. Condos). When we completed our analysis, we found the highest degree of correlation was between the number of multiple-unit homes and burglaries, followed by the number of renters and burglaries. Even still, there were significant outliers that consistently had high numbers of burglaries regardless of the variable studied. Therefore, we would like to conduct some more analyses concerning other possible factors influencing the number of burglaries, such as neighbourhood types or the proximity of preventative measures such as police stations.

Obtaining data from different sources made it difficult to be consistent in analyzing the same particular area. In the end, we had to do our analysis based on the data we were able to obtain, instead of the areas we would have liked to focus on such as breaking it down into subregions. Nevertheless, many of our original objectives had to be simplified to make the project achievable under the time allowed.

Both the census data and the crime data we used were incomplete or missing some attributes, possibly due to the methods used in collecting. The census data are based on samples from the population, and areas with small population may have had data suppressed for confidentiality reasons. We found it odd, however, that data regarding housing units was suppressed or otherwise not included, yet seemingly more sensitive data regarding income was included for all of the census tracts. In any event, the lack of data for these census tracts meant that there were no values to correlate the burglaries with.

With regards to the crime data, the number of burglaries given by the San Diego Police Department would be dependent on the number of cases that were reported, and there may have been other cases that were not reported and thus not included. Also, the Police Department reported 77 cases of residential burglaries where the location was not specified. These cases were not included in our analyses, but if we had been able to identify where they took place, we may have had different trends emerging in our data.

This project was an initial attempt at looking at what factors may influence burglars in breaking into a home. The applications of such a study could include collaborating with the police department in decreasing the number of burglaries, possibly by identifying areas where there needs to be more police presence and neighbourhood watch involvement. In the future, we would like to see if there is a similar trend in other cities of similar size for the sake of comparison.

Issues Encountered

The left image (aerial photo)- most of census tract #9902 is actually water, not land, and yet there are indication of one of the highest income values as shown in the center image.

The right image- census tracts # 3800, 5500, 9902 (the hollowed tracts) - had zero value for number of owners, renters, or housing units. However, the population data indicated that there were people living in them. Census tract #9902's crime statistics indicated that there were two counts of larceny, but no other crimes committed. Census tracts #3800 and 5500 had small numbers of burglaries.

The center and right images also indicated that there was a census tract within a census tract. Considering this along with the fact that census tract #9902 is mostly water, this raised the question of how the census tract boundaries were drawn and why certain regions with possibly incomplete census data were included.

These issues had an influence when performing the regression analyses because the burglaries (if any) were not being compared to any values for the other variables, thus affecting the overall trends in the results.

A review of all the regression maps show four census tracts that consistently have high numbers of residential burglaries, independent of which variable they were being correlated to. The census tracts in question were # 7600, 7700, 7901, and 7904. Apart from census tract # 7700 in the housing unit analyses, these tracts always were 2.5 standard deviations above the best fit line (right image shows correlation of population and residential burglaries). The left image displays the total number of residential burglaries per census tract, showing that the values were extremely high to begin with. This suggests that the burglaries in this region were not influenced by the factors we investigated. Further research should be conducted to better understand the underlying trends in this area.
For an alternate analysis click here