Statistics Primer

Complex phenomena like crime emerge due to a variety of factors and circumstances, rather than from a single cause. Multiple linear regression is a helpful tool for researching such complexities. A regression model is a statistical tool that helps researchers understand how different factors relate to an outcome. Think of it as a way of asking, “If we hold everything else constant, what happens to crime when this one factor changes?” 

This analysis compares Utah census block groups using a composite crime index, utilizing data from 2024. Each factor analyzed in the regression model gets a coefficient, which expresses the size and direction of its relationship with crime: a positive number means crime tends to increase as that factor increases, while a negative number means crime tends to decrease as that factor increases. 

The results from this analysis are correlational, rather than causal; therefore, causal claims about the relationship between a given variable and crime rates cannot be made based on this study. In addition, the strongest effect sizes in this study are fairly small, and there is enormous variability in the data, even under conditions where the strongest effect sizes would be expected.

    • The data indicate a negative relationship between crime and population size, with an effect size of .33 percent.  For example, a block group with 2,200 residents is expected to have a 3.3 percent lower crime rate compared to a similar block group with 2,000 residents.

    • The data also indicate a very weak, negative correlation between population density and crime. A 10% increase in population density correlates to a 0.4% decrease in crime rate. 

    How to interpret these findings: 

    This analysis compares crime rates across census block groups of different sizes.  

    This analysis does not test how an increase in population density within a block group will affect crime rates over time. 

  • Description text goes here

  • Description text goes here