Spatial analysis of violent crimes in Metro Manila, Philippines

ABSTRACT

Spatial patterns of murder and physical injury in Metro Manila, Philippines were visualised through conditional choropleth maps. Relationship of both crime rates with some demographic variables were investigated while accounting for possible spatial autocorrelation using spatial lag models. Results show that both crime rates tend to cluster in the northern cities of Metro Manila. Furthermore, significant spatial lag coefficients were found only for physical injury rates, with values ranging from 0.49 to 0.62, signifying a positive city-level spatial dependence of physical injury rates in Metro Manila. Moreover, some demographic covariates, such as population density, percentage of young males, education, marriage, and immigration were found to be associated with both crime rates. These results could serve as useful indicators of crime incidence; thus it is recommended that crime monitoring systems include them to aid in resource allocation and program planning for better crime prevention and security management.

KEYWORDS:

• Murder
• physical injury
• spatial analysis
• crime correlates
• demography
• social disorganisation theory

Acknowledgements

The researchers would like to extend their sincerest gratitude to the National Capital Region Police Office’s Regional Investigation and Detective Management Division for providing the essential data utilised in this particular study.

Notes

1. $\rho w_i{y}_i$ is a spatially lagged dependent variable and ${ϵ}_i$ is an independent error term. The coefficient $\rho$ is defined as the spatial coefficient with values that indicate expected increase or decrease in y values of neighbouring spatial units (Ward & Gleditsch, 2007).

2. Rate smoothing is done in order to deal with variance instability resulting from extreme values in population distribution. In particular, Empirical Bayes (EB) employ Bayesian principles to aid in the adjustments of raw rate estimates through shrinkage, which is moving the value of a raw rate towards the overall mean. This effect is more prone to raw rates derived based on a small population at risk, while rates derived based on a large population at risk have tendencies to experience no change (Anselin, Lozano, & Koschinsky, 2006).

Additional information

Notes on contributors

Vio Jianu Mojica

Mr. Vio Jianu Mojica is an Assistant Instructor at the Mathematics Department at De La Salle University. Currently, he is also a Master of Science in Statistics student at the same institution. His research interests are in the area of statistics, particularly in spatial analysis.

Mr. Adelbert Choi is currently a Master of Analytics student at the Royal Melbourne Institute of Technology. He is a Bachelor of Science in Statistics graduate from De La Salle University.

Mr. Robert Neil Leong is currently a PhD student in Public Health and Community Medicine in the University of New South Wales working on economic evaluation of vaccination. He has interests in building and developing spatio-temporal surveillance systems with a focus on infectious diseases.

Mr. Frumencio Co is an Assistant Professor in the Mathematics Department of the College of Science of De La Salle University. His research interests are in the areas of statistics, epidemiology, and biostatistics. He is also involved in interdisciplinary research activities.