TABLE 6.

CONTROLS FOR HEALTH BELIEFS PREDICT SANITATION BUT DO NOT CHANGE MAIN RESULT

	Dependent Variable: Household Open Defecation
	(1)	(2)	(3)	(4)	(5)	(6)
Village untouchability A−i	.0743***	.0739***	.0732**
	(.0267)	(.0285)	(.0285)
Household untouchability A	−.00276	−.000502	.000161
	(.00961)	(.0107)	(.0107)
Village untouchability B−i				.0586**	.0569**	.0565**
				(.0264)	(.0282)	(.0282)
Household untouchability B				.00513	.0119	.0126
				(.00942)	(.0106)	(.0106)
Health beliefs		F5 = 5.61	F5 = 5.57		F−5 = 5.61	F5 = 5.60
		p < .0001	p < .0001		p < .0001	p < .0001
Doctor social contact			−.0250*			−.0259*
			(.0140)			(.0141)
Other health social contact			.00869			.00861
			(.0143)			(.0143)
All controls from table 2	✓	✓	✓	✓	✓	✓
n (rural households)	27,320	22,016	22,016	27,320	22,017	22,017

Note. The dependent variable is an indicator for household open defecation. “Village untouchability⁻ⁱ” is the fraction of households other than the respondent’s that report practicing untouchability in the respondent’s village. “Open defecation” is an indicator at the household level. “Health beliefs” is the set of dependent variables from table 4. “Doctor social contact” and “other health social contact” are indicators for the household reporting knowing somebody in their family, caste, or community, as in table 5. Each regression includes the most complete set of controls from the main results in table 2: consumption as a cubic polynomial, caste by religion indicators, state fixed effects, and the set of extended controls. The regressions in cols. 1 and 4 are identical to those in panels A and B of table 2, col. 6. Standard errors, clustered by village, are in parentheses.

^*p < .10.

^**p < .05.

^***p < .01.