Media Consumption and Fear of Crime: Evidence of the Need for an Intersectional Approach

ABSTRACT

Race/ethnicity, gender, and exposure to various types of media have been identified as important predictors of fear of crime. However, previous research largely fails to take a truly intersectional approach when testing this relationship. Utilizing a unique data source that oversamples for minority respondents and includes measures for social media, Internet, and traditional media consumption the current study attempts to fill this gap. Our research finds evidence that the link between media consumption and fear of crime varies significantly across intersectional subsamples of race/ethnicity and gender. This identifies a need for future intersectional research on fear of crime.

KEYWORDS:

• Fear of crime
• public opinion
• media
• intersectionality
• quantitative methodology

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

1. The survey instrument asked respondents if they identified as White, Black, or Hispanic/Latinx. We acknowledge the words “Hispanic” and “Latinx” can convey different meanings; however, they are referenced collectively in this research to best reflect the original measurement.

2. Ethnicity supersedes race, so Whites, Blacks, and Hispanic/Latinx respondents are distinct groups.

3. Cronbach’s tests the reliability of a measure when it is derived from multiple measures of a similar concept (Santos, 1999). Ranging from 0 to 1, higher scores indicate that the measures elicit similar responses from respondents.

4. A reviewer suggested we include a measure of political status. While the data does include a measure of political orientation, it was excluded due to a lack of variability across race and gender. For example, only 15 respondents identified as Black, female, and conservative. However, when the variable is included it does not reach significance in the full sample model and does not significantly alter any of the key findings.

5. Z = (b_{1 –} b₂)/√(SEb₁² + SEb₂²)

6. This degree of disaggregation does pose potential problems related to statistical power and Type II error. That is, the subsample size in conjunction with the number of covariates included in our models increases the risk that the models will fail to detect statistically significant relationships. Sample size estimation using standard parameters, however, suggests this danger is limited to extremely small relationships. Specifically, estimation using a statistical power level of .8, probability level of .05, and 15 covariates indicates minimum required sample sizes of 950, 139, and 68 to detect statistically significant effect sizes of .02 (small), .15 (moderate), and .35 (large), respectively. In the case of this study, all subsamples maintain an n > 185.