• Open Access

Household size: an overlooked confounder when considering inequalities in food choice


Addressing health inequalities is an international concern.1 With diet implicated in the aetiology of many socioeconomically-graded health conditions,2,3 identifying the determinants of dietary inequalities is pertinent.

Recent investigations have explored whether psychosocial factors relevant to food choices (e.g. nutrition knowledge, cost concerns) are socioeconomically patterned and whether they mediate dietary inequalities. Consequently, multiple factors have been identified as promising targets for health promotion. However, it is noteworthy that few confounding variables are typically acknowledged in this field of research. Age and gender are generally considered, however household characteristics, such as household size and composition are rarely addressed. Acknowledgement of these factors could be achieved by providing descriptive information on household characteristics, and may not involve more complex solutions such as modelling household variables or computing household-equivalised socioeconomic measures.

Standard definitions stipulate that a confounder must be associated with both the independent variable and dependent variable.4 Definitions vary as to whether confounding variables may be causually related to the relationship/s that they confound.5 Therefore, confounders are identified on both conceptual and statistical grounds. Whether researchers consider inequalities in food selection, in psychosocial factors, or in both, household characteristics are likely to influence findings. This letter contributes to this field by assessing household size as a confounder of the following relationships:

  • the association between income and food choices
  • the association between income and psychosocial factors.

The data examined are from the Brisbane Food Study, a cross-sectional study of household food purchasers in 2000.6 To focus on household size differences (rather than household composition), only couples (with or without children) were considered (n=529). Four category variables were used to describe household size (2, 3, 4 and 5+) and income (<$25,999, $26,000–51,999, 52,000–77,999, >$78,000). Respondents indicated their usual choice of 16 staple grocery foods (e.g. bread, milk). Grocery choices were classified as ‘regular’ or ‘recommended’ depending on the relative salt, sugar, fat and fibre content of the choices made (e.g. reduced fat milk was classified as ‘recommended’ while full-cream milk was classified as a ‘regular’ choice). A grocery purchasing index was constructed (scaled 0–100) with higher scores reflecting purchasing patterns more consistent with dietary guidelines. Nine psychosocial indexes (scaled 0–100) are considered in this letter:

  1. Taste preferences for recommended grocery choices.
  2. Taste preferences for regular grocery choices.
  3. Nutrition concern (e.g. consideration of the sugar/salt/fat/fibre content of foods).
  4. Satisfaction with diet (belief that the family diet is nutritious and adequate).
  5. Belief that healthy food is expensive.
  6. General food cost concerns.
  7. Knowledge of the relationships between diet and disease.
  8. Knowledge of the nutrient content of foods (e.g. that dairy products contain calcium).
  9. Knowledge of dietary guidelines.

Household income differed significantly according to household size (χ2 < 0.001). Table 1 presents the associations between household size, grocery purchasing and psychosocial factors. As household size increased grocery purchasing behaviour was observed to be less consistent with dietary guideline recommendations, as reflected by lower grocery purchasing index scores (p<0.001). In addition, as household size increased respondents were; less likely to express that nutrition concerns influenced their food choices, and more likely to prefer the taste of regular grocery choices, to indicate that general food cost concerns influenced their food choices, and to believe that ‘healthy’ food was expensive.

Table 1. Estimated marginal mean grocery purchasing and psychosocial index scores across households of varying size (n=529).
Household Size (persons)Grocery Index Scores (0–100)Psychosocial Index Scores (0–100)
  Taste preferences RecommendedTaste Preferences RegularNutrition concernSatisfaction with dietBelief healthy food expensiveGeneral food cost concernsKnowledge diet disease relationshipsKnowledge nutrient content of foodsKnowledge dietary guidelines
  1. R= Referent household size category.

  2. *Statistical significance of the relationship between the relevant household size category and the referent group (2 person household). * p< 0.05, ** p < 0.001.

2 (R)56.566.545.666.968.145.838.485.671.081.4
Overall p-value<0.0010.9440.1060.4060.2190.114<0.0010.0070.5780.949

Thus, these results support the consideration of household factors in analyses considering socioeconomic differences in food choices. The majority of food is known to be consumed at home,7 making consideration of the home context important irrespective of whether individual food and nutrient intake or household food choices are considered. Hopefully, this short appraisal will prompt consideration of household characteristics in future research.