Investigation of the Factors Affecting the Homicide Counts in the USA by Quantile Regression

Abstract

This study has conducted in the United States of America (USA) in the 1960s, 1970s, 1980s, and 1990s to determine whether divorce rates, unemployment rates, and the population had an impact on homicide numbers. Initially, all variables were examined and interpreted geographically on the map by districts. Subsequently, the stationary circumstance of variables has been tested with the Augmented Dickey-Fuller (ADF) Test, which is one of the unit root tests. After it has found that the variables did not need to stabilize, regression analysis has performed by the Least Squares (LS) method. Quantile regression, which is an alternative method to the LS method, has been used since all the resulting models do not have a normal distribution. These models have been created with 3 diverse quantile values for each period. Among these models, the ones with the highest correlation coefficient are the models having the 0.75 quantile value. Therefore, the results have been obtained from models with the 0.75 quantile value. Hence, for the homicide counts in the USA, those have found that the country population had a positive effect in the 1960s, the country population and the divorce rates had positive effects in the 1970s, the country population had a positive effect in the 1980s, and the country population and the unemployment rates had positive effects in the 1990s. Furthermore, the unemployment rates in the 1970s and 1980s had a negative effect on the homicide counts in the USA.

Keywords

• homicide counts
• geoda
• least squares method
• quantile regression
• the usa homicide