Duke University Press

Table 4

Logistic models of mortality by neighborhood inequality exposure, using three strategies of adjusting for selection

Gini	Unadjusted	Regression-Adjusted	IPT-Weighted
First	—	—	—
Second	0.83^***	0.72^**	0.71^**
	(0.24)	(0.25)	(0.25)
Third	0.52	0.37	0.46
	(0.26)	(0.27)	(0.26)
Fourth	0.73^**	0.51^*	0.75^**
	(0.25)	(0.26)	(0.26)
Fifth	0.60^*	0.38	0.67^*
	(0.26)	(0.27)	(0.27)
Notes: Effects are log-odds ratios of mortality risk. Positive coefficients indicate increased risk. All models adjust for baseline covariates: birth cohort, race, gender, southern region, and educational attainment. Time-varying covariates are employment, family income, neighborhood average family income, neighborhood population size, and neighborhood proportion Black. The first model does not adjust for time-varying covariates; the second model includes time-varying covariates as regression controls; and the third model uses time-varying covariates to inform IPT weights.
^p < .05; ^p < .01; ^**p < .001 (for two-sided tests of no effect)

Gini	Unadjusted	Regression-Adjusted	IPT-Weighted
First	—	—	—
Second	0.83^***	0.72^**	0.71^**
	(0.24)	(0.25)	(0.25)
Third	0.52	0.37	0.46
	(0.26)	(0.27)	(0.26)
Fourth	0.73^**	0.51^*	0.75^**
	(0.25)	(0.26)	(0.26)
Fifth	0.60^*	0.38	0.67^*
	(0.26)	(0.27)	(0.27)
Notes: Effects are log-odds ratios of mortality risk. Positive coefficients indicate increased risk. All models adjust for baseline covariates: birth cohort, race, gender, southern region, and educational attainment. Time-varying covariates are employment, family income, neighborhood average family income, neighborhood population size, and neighborhood proportion Black. The first model does not adjust for time-varying covariates; the second model includes time-varying covariates as regression controls; and the third model uses time-varying covariates to inform IPT weights.
^p < .05; ^p < .01; ^**p < .001 (for two-sided tests of no effect)