This study explored the role of five socioeconomic indicators in explaining socioeconomic differences in children’s diets, finding that each indicator was independently associated with at least one dietary outcome. In general, lower SEP was associated with poorer dietary behaviours, specifically higher intake of non-core foods, sweet drinks and more unhealthy behaviours; and lower intake of fruits, vegetables and less healthy behaviours. This is consistent with previous reports of socioeconomic gradients in children’s dietary intake [2–6, 34, 35]. Overall, the amount of variance explained by predictive CCR models was small, highlighting that a range of other factors are also important for children’s dietary intake.
However, this effect size is consistent with the magnitude of variance in children’s dietary intake explained by socio-demographic factors alone in other studies . The dietary intake of children in this study did not meet suggested dietary targets, and the small differences in dietary intake between children of different socioeconomic groups suggest that all children, irrespective of SEP, would benefit from health promotion to improve dietary intake. However the necessity to target improvements in diets of low SEP children should not be dismissed. Socioeconomically-related differences in the diets of adults are generally larger than those identified for children [37, 38], which may be attributed to declines to dietary quality during adolescence and therefore low SEP children are an important target group to improve eating habits before declines in dietary quality during adolescence and adulthood . The effect of SEP appeared to be particularly important for girls’ sweetened drink intake, which was significantly and negatively associated with all measured SEP variables. This is in contrast to another study which found SEP predicted boys’, but not girls’ soft drink consumption . A target area for future nutrition promotion may be encouraging the reduction of sweetened drink intake among low SEP children.
The effect of SEP differed between boys and girls, consistent with previous studies [4, 18]. In particular, gender differences were identified in associations of fruit intake with SEP. In contrast to boys, fruit intake was higher among girls of low educated mothers, low income households and living in more disadvantaged neighbourhoods. This finding is difficult to explain and requires further investigation. Fruit, unlike vegetables, typically requires little preparation or cooking skills, and may be perceived as an easy to consume snack [40, 41]. Low SEP women have reported lower levels of food preparation and cooking skills , and so might be more likely to provide fruit for a handy snack or part of a meal. If so, it is unclear why this was the case only for girls and not boys. Alternatively, these findings may be due to more general gender stereotypes around food, eating and health. From an early age, girls are socialised differently with regard to food and this may result in more health consciousness, more weight concerns, and different health beliefs and attitudes, which are translated into different dietary patterns [43–45]. Girls may perceive their food environments differently to boys, reporting more restrictions around eating than boys, despite parent reports indicating no differences in rule imposition between boys and girls . However, it is unclear how SEP may impact on gender stereotypes around eating and this warrants future consideration.
Multiple socioeconomic factors independently predicted many of the dietary outcomes, suggesting the need to consider multiple socioeconomic indices when explaining gradients in children’s dietary intake. To provide further insight into potential covariance and conceptual overlap among SEP indicators, adjusted bivariate CCR regression models for each individual SEP indicator and dietary outcome were conducted (results available as Additional file 1: Table S1 online). The patterning of results between adjusted bivariate associations and multivariate models was similar, however there was attenuation or amelioration of associations between SEP and diet in some multivariate models compared with bivariate associations, suggesting some overlap between SEP variables. Turrell et al.  similarly found that in multivariate models, independent associations of SEP indicators with food purchasing changed compared with original bivariate associations, and suggested that this attenuation occurred as a result of unmeasured effects due to overlap between SEP indicators. However, correlations between SEP indicators are weak, and Turrell et al.  cautioned that researchers should not assume that each SEP indicator is tapping a similar underlying construct, and therefore indicators should not be used interchangeably. Correspondingly, the findings of this study highlight a risk of misrepresenting socioeconomic gradients if only a single measure of SEP is used. The differences between adjusted bivariate associations and multivariate models suggest that had results solely relied upon bivariate associations, the claims about SEP-diet indicators would have been misstated. Over- or underestimation of socioeconomic gradients in dietary intake may have implications for the development and implementation of dietary intervention strategies or health promotion campaigns. Distinct facets of SEP may influence children’s dietary intake by conceptually different pathways, and therefore inclusion of SEP variables in analyses should be driven by conceptual considerations. Further still, an individual’s SEP is defined by more than one SEP variable, and these variables may each influence dietary intake differently, but each may also mediate associations between other SEP indicators and diet. For example, education may influence occupation and income, which may then impact on diet; in this case occupation and income would be mediating variables. Formal tests of mediation were beyond the scope of this investigation, but future studies should consider the mediating role and interplay between different SEP variables in associations of SEP and children’s diet. We refer readers to Turrell et al.  which provides a detailed discussion of the stability and robustness of different SEP indicators in relation to adult diet. Given the evidence for the importance of each SEP indicator independently in relation to eating, in the following section we consider theoretical conceptual pathways of influence for SEP on children’s diet.
Mother’s education level was the most consistent predictor across all dietary outcomes. Low education was associated with higher intake of non-core foods and sweetened drinks, and more unhealthy behaviours, as well as lower vegetable intake and less healthy behaviours. Maternal education has been identified as the most consistent predictor of children’s diet, and low maternal education has been associated with poorer dietary intake across a range of dietary outcomes [1, 3, 5, 23, 34, 47]. Higher education may enable individuals to better access, assimilate and put into practice health information [48, 49], and may therefore better equip parents to understand and make use of health and nutrition information. Parents of lower education may have poorer nutrition knowledge [49, 50], and consider health less often when making food choices for themselves and their children [48, 51]. Conversely, parents of higher education may place more importance on healthy eating, which may be related with better nutrition knowledge . Nutrition knowledge and health consideration may be related with healthier dietary intake among mothers and children, and may inform the types of foods parents provide for their family [49, 51, 53, 54].
Findings for the effect of mother’s occupation and employment on children’s dietary intake were mixed. On the one hand, employment in higher status occupations was associated with more fruit intake and less sweetened drink intake, consistent with previous studies [18, 19]. Unexpectedly, boys of mothers in higher status occupations reported less fruit intake and healthy behaviours and more unhealthy behaviours. This was explored further by considering the effect of maternal employment on children’s dietary intake, and we found that employment was associated with dietary intake independently of mother’s occupation. Children of employed mothers were likely to consume more sweetened drinks, fewer fruits and vegetables and engage in less healthy behaviours. Time spent in employment may impact on the time mothers have to engage in activities that can positively influence children’s food intake, such as supervising breakfast consumption, eating family meals and engaging children in food shopping. Time-poor mothers may be more reliant on takeaway meals, contributing to lower consumption of fruit and vegetables . Neumark-Sztainer  reported that adolescents of unemployed mothers consumed family meals more frequently, and this was related to more positive dietary behaviours. Younger children of employed mothers had a higher sweetened drink intake and a lower likelihood of consuming fruits and vegetables as a snack . Sweeting and West  found children of mothers in part-time employment to have a significantly lower risk for unhealthy eating, suggesting these mothers may have an economic advantage over homemakers and a time advantage over mothers who are employed full-time . Occupation determines working conditions such as time spent in employment, flexibility of working arrangements and differential exposure to work place stressors, and these factors may affect children’s dietary intake [24, 25]. Given this, it is likely that a combination of factors related to both occupation and employment influence children’s food intake.
This study found that higher income independently predicted more fruit intake and less sweetened drink intake and unhealthy behaviours. Income appeared less frequently as a predictor of children’s dietary intake than education, occupation and employment. This is consistent with previous literature, except that previous authors have also found income to be positively associated with vegetable consumption [1, 28, 35]. Income reflects the financial resources available for food purchasing, accessing resources and health professionals. Low income families may prioritise non-food expenses such as rent/mortgage, utility bills and school fees over food and health expenditure, as these expenses are less flexible , and must allocate a greater proportion of their overall income to purchasing food. Purchasing groceries according to dietary guidelines costs low income families 35-44% of their disposable income compared to approximately 20% for families of average income [57, 58]. Purchasing patterns of low-income families indicate they may purchase fewer fruits and vegetables, foods high in fibre, low in fat, sugar and salt [59, 60]. Healthy foods, particularly fruit and vegetables, may be perceived to be more expensive  and low-income adults are more likely to report price as a barrier to fruit and vegetable consumption [62, 63].
SEIFA appeared least frequently as a predictor of dietary intake, with an effect of smaller magnitude compared with the other socioeconomic indicators. Living in a more disadvantaged area predicted higher sweetened drink intake and more unhealthy behaviours, and lower fruit intake for boys. The types of foods readily available for purchase in close proximity to the family home may influence children’s food consumption . More access to convenience stores and fast food restaurants may contribute to higher intake of processed snack foods, displacing intake of fruits and vegetables [65, 66]. However the evidence for disproportionate access to food stores by neighbourhood disadvantage is mixed. Studies conducted in the USA suggest that disadvantaged neighbourhoods may have fewer supermarkets and more convenience stores, resulting in higher prices and lower availability of healthy foods [67–71]. Evidence from Australian studies is less consistent, with some studies suggesting better access to supermarkets and greengrocers in more advantaged neighbourhoods [72, 73], but other authors finding no difference in store availability between low and high SEP neighbourhoods . Evidence of neighbourhood variation in takeaway and fast-food outlets is mixed, with studies showing no socioeconomic differences ; a higher density of fast-food outlets in low socioeconomic areas ; and conversely closer proximity to fast-food restaurants in more advantaged areas . It is likely that the relationships between diet and neighbourhood are culturally and contextually specific, and may differ by region, state and country.
Strengths and limitations of study
As this study was cross-sectional causality cannot be inferred, and although cross-sectional studies offer a snapshot of current associations, factors resulting in behaviour change cannot be identified. Individual SEP may predict participation in research, with lower participation rates among individuals of low SEP, and non-responders to dietary surveys may also differ from responders on dietary intake and attitudes [77, 78]. This indicates a risk of a recruitment bias, whereby parents more ‘concerned’ about health and nutrition may have opted to participate which may have influenced the types of responses in questionnaires. However, parents were offered a $30 voucher as compensation for time taken to participate in the study, to encourage some parents to participate who may not have otherwise done so. There were no differences in scores reported for dietary intake between children who did and did not participate in phase two; however families who participated in phase two resided in neighbourhoods of higher SEP than those who did not. This potential respondent bias must be recognised as a limitation of this study, despite achieving a socioeconomically stratified participant sample and providing incentives for participation. Finally, the limitations of children reporting their own dietary intake need to be recognised, in terms of reporting errors and poorer recall of intake. However, the CNQ has been shown to have acceptable validity and reliability in children of this age group  and children of this age are capable of self-reporting dietary intake . The strengths of this study include the relatively even distribution of the sample across socioeconomic strata. Multiple dimensions of SEP were considered simultaneously, allowing for independent and shared effects of demographic variables to be determined. Analyses were conducted separately for boys and girls, enabling the identification of sex-specific socioeconomic predictors of dietary intake. SEP data were not reduced into broad categories for analysis (i.e. low versus high), therefore increasing the sensitivity to detecting gradients across SEP strata. Online administration of the CNQ allowed for in-built measures, such as forced question response, to minimise missing data and reduce errors associated with data entry.