Table 2.

Estimated rate ratios (95% confidence intervals) of sub-city homicide rates associated with a 1SD higher value of social and built environment characteristics in males and females in single exposure and multiple exposure models.

	Single exposure	All social environment	All built environment	Fully adjusted
Males
Social environment
Education	0.87 (0.84–0.90)	0.87 (0.84–0.90)		0.87 (0.84–0.90)
GDP	0.83 (0.78–0.89)	0.86 (0.80–0.92)		0.87 (0.81–0.92)
Gini	1.43 (1.25–1.64)	1.40 (1.22–1.60)		1.28 (1.10–1.48)
Built environment
Isolation	1.11 (1.05–1.18)		1.16 (1.09–1.24)	1.13 (1.07–1.21)
Population density	1.22 (1.07–1.39)		1.12 (0.98–1.27)	1.06 (0.93–1.20)
City size	1.19 (1.04–1.35)		1.29 (1.12–1.48)	1.16 (0.99–1.35)
Population growth	0.91 (0.85–0.98)		0.94 (0.88–1.01)	0.97 (0.90–1.03)
Females
Social environment
Education	0.89 (0.86–0.93)	0.89 (0.86–0.93)		0.90 (0.86–0.93)
GDP	0.90 (0.85–0.95)	0.92 (0.87–0.97)		0.92 (0.87–0.97)
Gini	1.23 (1.11–1.37)	1.22 (1.10–1.36)		1.21 (1.07–1.36)
Built environment
Isolation	1.07 (1.02–1.12)		1.09 (1.04–1.15)	1.07 (1.02–1.12)
Population density	1.01 (0.92–1.12)		0.96 (0.87–1.07)	0.93 (0.84–1.03)
City size	1.06 (0.96–1.16)		1.14 (1.02–1.27)	1.05 (0.93–1.18)
Population growth	0.96 (0.91–1.01)		0.96 (0.91–1.01)	0.98 (0.93–1.03)

The outcome is homicide rates at the sub-city-level, education is at the sub-city-level and all the other exposures are at the city-level. All models assume a Negative Binomial distribution with constant dispersion parameter and include sub-city-level and city-level random intercepts, country-level fixed effects and a dummy for age. A logarithmic link was used, and an offset equal to the log of the corresponding population divided by the undercounting correction factor (itself multiplied by 100,000) was included. All exposures were standardized to have a mean of 0 and a standard deviation of 1.