Modelling of Narcotics Crime Incidents in Malaysia: A Generalized Linear Model by Using the Poisson and Negative Binomial Regression Approaches

Abstract

This study analyses the spatial patterns of narcotics crime incidents in Malaysia using Poisson regression and negative binomial regression approaches. The data set comprises the number of narcotics-related arrests categorized by state, gender, age group, ethnicity, and type of offence. Arrests recorded under the Dangerous Drugs (Special Preventive Measures) Act 1985 and Poisons Act 1952 in Perlis by female offenders, aged over 60 and minority ethnic considered as the reference group for this study. The analysis found that the negative binomial regression model is more suitable than the Poisson regression model, as indicated by a lower AIC value. The analysis revealed that each state has a positive and significant incidence rates of narcotics crime compared to Perlis. In terms of demographics, males and individuals aged 19 to 39 years were the most dominant groups among those arrested. The Malay ethnic group showed a high and significant positive association with incident rates compared to other ethnic groups. Offence categories related to drug distribution and possession demonstrated negative relationships, whereas positive urine test results were positively associated with arrest counts. Drug distribution and possession offences usually involve more planned and targeted operations. So, they are linked to specific locations and groups of people. On the other hand, positive urine test results often come from random screenings of large numbers of individuals, which leads to mass arrests. Therefore, the two types of offences show different patterns. The study shows that where crimes happen and who is involved are key factors in explaining why narcotics crime rates vary across Malaysia.