Tobacco use and household expenditures on food, education, and healthcare in low- and middle-income countries: a multilevel analysis

Table 2 Unadjusted regression model of logged household expenditures for food, education and healthcare

	Food			Education			Healthcare
Variables	Random-slope	Random-intercept	Fixed-effects	Random-slope	Random-intercept	Fixed-effects	Random-slope	Random-intercept	Fixed-effects
	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)
Current tobacco use (ref. No use)
Daily	−0.064**	−0.051**	−0.052**	−0.208**	−0.183**	−0.184**	−0.137**	−0.121**	−0.122**
Occasional	−0.083**	−0.050**	−0.050**	−0.138**	−0.065**	−0.066**	−0.093*	−0.043	−0.043
Constant	4.666**	4.657**	4.656**	1.517**	1.498**	1.547**	1.584**	1.574**	1.455**
SD (Daily)	.081 (.015)^a	—	—	.137 (.027)^a	—	—	.148 (.028)^a	—	—
SD (Occasional)	.103 (.029)^a	—	—	.120 (.035)^a	—	—	.143 (.044)^a	—	—
SD (Constant)	.517 (.058)^a	.520 (.059)^a	—	.699 (.079)^a	.691 (.078)^a	—	.662 (.075)^a	.652 (.074)^a	—
Hausman test^b	—	[χ^c(2) = 4.05 (p = 0.132)]		—	[χ^c(2) = 7.59 (p = 0.023)]		—	[χ^c(2) = 5.82 (p = 0.055)]
Likelihood ratio test^c	[χ^c(2) = 59.86 (p < 0.001)]		—	[χ^c(2) = 43.86 (p < 0.001)]		—	[χ^c(2) = 47.16 (p < 0.001)]		—
Observations (countries)	53,185 (40)			50,732 (40)			50,602 (40)

Notes: **p < 0.01, *p < 0.05; SD, standard deviation of estimated coefficient; ^astandard error in parentheses; ^bRejecting the null hypothesis would favor fixed-effects to random-intercept; ^cRejecting the null hypothesis would favor random-slope over random-intercept

ISSN: 1471-2458