Research Report

Florida TaxWatch used multiple regression analysis to determine which of the factors listed below are significantly associated with crime.

Multiple regression is a statistical technique used to assess the magnitude and direction of relationships between independent variables and dependent variable. The unit of analysis is county. In this TaxWatch Report, eight factors were selected to test the relationship between socioeconomic status, population mobility, community heterogeneity and crime. These variables (called independent variables in multiple regression analysis) are:

1) Unemployment Rate (UNRATE):the proportion of the number of unemployed people to total labor force;

2) Poverty (POORPER): the proportion of poor persons below poverty level to total population;

3) High School Graduation Rate (HSGRAD): the ratio of the number of graduates to the number of first-time ninth graders who entered school four years earlier;

4) Female-Headed Households (FEMHOUSR): the proportion of female-headed households to total households;

5) Mobility (MIGRATIO): the ratio of migration people(both in- and out-) to total population;

6) Young Age Group(YOUNG): the proportion of young people between 15 and 24 to total population;

7) Heterogeneity (LANG98): the number of languages spoken within a county's school district;

8) Urbanization: counties which have 75,000 people or more are classified as "urban(1)" and counties that have less than 75,000 people are classified as "rural(0) as a dummy variable.

These variables were selected because they were the ones most commonly used in criminology literature to represent socioeconomic status, community population mobility and community heterogeneity, as well as crime rates. They were derived mainly from 1998 Florida Statistical Abstract.

The dependent variable for the multiple regression analysis was crime rate per 100,000 population for each county.

The results of the multiple regression analysis are presented in Figure 4. Overall, they shed considerable light on the effect of socioeconomic status, community population mobility and community heterogeneity on the variance in crime rate in Florida's counties. The R2 value (56.9%) indicates that more than half of the crime rate in Florida's counties is accounted for by the independent variables used. According to the F-test result, the overall explanatory power of those independent variables on crime rate is acceptable since the F- statistic (9.564) is much greater than the critical F value (3.01) with df:8,58 at the predetermined .05 level of significance. In other words, the results are statistically significant.

The B coefficients show the magnitude and direction of the relationship between each of the independent variables and crime rate. For example, the B coefficient for HSGRAD indicates that every one percent increase in county high school graduate rate, on average, decreases the crime rate by almost 37 persons per 100,000 population. Beta coefficients indicate the relative significance of each independent variables in relation to crime rate. In other words, the independent variable that has the strongest association with crime rates is LAN98 (the number of languages spoken in a county).

Overall, the regression analysis shows intuitively and theoretically plausible results. All the directions of statistically significant relationships reveal intuitively logical results. Unemployment and heterogeneity are positively related to crime rate. As unemployment or heterogeneity within Florida counties increase, so does crime rate.

We can know whether or not each independent variable is statistically significantly associated with crime rate by comparing each t-ratios with the critical t-value(1.671: one-tailed test, .05 level of significance). When the t-ratio is greater than 1.671, the relationship between the independent variable and the crime rate is statistically significant.

Most noteworthy, the regression result reveals that among the eight variables considered in this report, heterogeneity represented by the number of languages is the most statistically significantly associated with the crime rate. In other words, it shows that those counties having higher heterogeneity tend to have higher crime rates. Another statistically significant independent variable is unemployment rate. Unemployment rates in Florida counties have strong and positive relationships with crime rates. The other six variables are not statistically significant in association with crime rate.