Demographic Correlates of Humanizing Language in Media Coverage of Crime: Evidence from the Boston Globe, 1976-84

We merge FBI data on homicides in Massachusetts between 1976 and 1984 with the corresponding Boston Globe articles covering those homicides to study how the race, age, and sex of homicide victims and offenders interact to shape the quantity of media coverage (number of articles) and quality of media coverage (use of humanizing language as determined by NLP-assisted content analysis). Our analysis reveals rich patterns of differential media coverage across groups and subgroups, even after controlling for a battery of homicide-specific confounding factors. Specifically, we document that: (i) among male homicide victims, the probability of humanizing coverage is significantly higher for white victims than for black victims, with a difference of 30 percentage points for juveniles (under 18) and a difference of between 10 and 16 percentage points for all other age groups; (ii) there is no statistically significant difference in the quantity or quality of coverage across races among female victims, except those between the ages of 18 and 29, among whom blacks are significantly less likely to receive humanizing coverage than whites; (iii) both female and juvenile victims tend to receive more coverage on average, and this coverage tends to be more humanizing when it occurs; and (iv) victims of homicides with black offenders receive significantly more humanizing coverage on average, but this effect disappears after controlling for other homicide-specific confounding factors. Our analysis highlights the importance of viewing differential media coverage of violent crime through an intersectional lens and how content analysis can be applied to produce comprehensive measures of humanizing language.