Table 2.

Mean differences between the estimated adjusted post-policy calories purchased from total foods and beverages and estimated adjusted counterfactual post-policy food purchases by educational level and household assets^{1, 2,3}

	Overall				High-In				Not High-In
	Absolute Difference	95% CI	p-value	Relative Difference	Absolute Difference	95% CI	p-value	Relative Difference	Absolute Difference	95% CI	p-value	Relative Difference
Education
<High School	−7·6	−29·3, 14·0	0·49	−1·7%	−46·8	−57·5, −36·1	<0·0001	−23·1%	39·2	24·5, 53·9	<0·0001	15·3%
High School	−13·9	−29·6, 1·8	0·083	−3·1%	−43·2	−51·4, −34·9	<0·0001	−21·8%	29·2	18·9, 39·6	<0·0001	11·7%
College or Greater	−28·1	−47·8, −8.5	0·0050	−5·6%	−59·3	−70·1, −48·4	<0·0001	−26·5% a	31·1	18·6, 43·6	<0·0001	11·3%
		p for interaction	0·095			p for interaction	0·34			p for interaction	0·013
Household Assets
Low	−22·3	−42·1, −2·5	0·027	−4·7%	−52·4	−62·2, −42·6	<0·0001	−25·2%	30·1	16·7, 43·5	<0·0001	11·3%
Middle	−15·6	−34·0, 2·8	0·097	−3·5%	−49·5	−59·1, −40·0	<0·0001	−24·6%	34·0	21·5, 46·4	<0·0001	14·0%
High	−11·8	−30·3, 6·7	0·21	−2·4%	−46·3	−56·6, −36·0	<0·0001	−21·6%	34·5	22·7, 46·2	<0·0001	12·6%
		p for interaction	0·93			p for interaction	0·23			p for interaction	0·27

¹Estimates derived from fixed effects models comparing post-policy nutrient content of purchases to counterfactual post-policy nutrient content of purchases based on pre-policy trends. Covariates included age of household’s main shopper, household head education level, household composition, household assets, and monthly regional unemployment rate, along with indicator variables for calendar months, a linear time trend, an indicator variable for the policy period, and the interaction of time trend, policy period, and household education or assets. Household education was defined as the educational level of the household head. Household assets was created using a factor analysis including a household’s number of bedrooms, number of bathrooms, and number of cars; this index was categorized into tertiles and specified as a set of indicator variables in the model. Pairwise comparisons were conducted between each level of education and household assets.

²Purchase data provided by Kantar WorldPanel Chile.

³p-value for interaction from Wald tests of the equality of the interaction for policy period, linear time trend, and education or assets.

^aThe only statistically significant pairwise comparison was between college and greater and high school education for high-in calories purchased, p=0.02.