How Socioeconomic Status Shapes Food Preferences and Perceptions

SES and Attribute Importance

Healthiness, fillingness, and taste reflect core intrinsic food attributes (Enneking, Neumann, and Henneberg 2007; Rusmevichientong, Jaynes, and Kazemi 2020; Yamim, Mai, and Werle 2020). Thus, understanding how SES shapes food preferences requires a systematic empirical investigation of how poverty and wealth influence the relative weights consumers assign to these critical food attributes.

Humans and other animal species tend to increase energy intake above their level of energy expenditure in the presence of environmental cues that imply that access to food is insecure and reduce their energy intake when they receive cues that access to food is secure (Nettle, Andrews, and Bateson 2017). Given that low-SES consumers are more likely to suffer from food insecurity and malnutrition (Fitchen 1987; Gundersen, Kreider, and Pepper 2011) and high-SES consumers live in a world of food safety and abundance, it is reasonable to expect that low-SES consumers will give more weight to fillingness when choosing from different food options. Further, as several prominent decision-making models reveal (Bettman, Luce, and Payne 1998; Payne, Bettman, and Johnson 1993; Simonson 1989; Simonson and Tversky 1992; Tversky, Sattath, and Slovic 1988), determining attribute weights entails trade-offs. Giving more importance to one attribute usually means giving a lower weight to another. If low-SES consumers are expected to be particularly sensitive to fillingness, it will likely make them less sensitive to other key relevant food attributes (e.g., healthiness and taste).

Previous research suggests that taste is a central attribute in food decision-making (Furst et al. 1996; Glanz et al. 1998; Kourouniotis et al. 2016). People display a generalized tendency to favor taste over other food attributes, and this tendency seems rather persistent across the SES continuum (Aggarwal et al. 2016; Glanz et al. 1998). Along the same lines, studies directly comparing the importance of taste and healthiness in food choices suggest that consumers might be particularly attached to taste (Verbeke 2006). Research has also shown that labels emphasizing taste are more effective in increasing the choice of healthy foods than labels focused on nutritional benefits (Turnwald and Crum 2019).

Thus, two general inferences can be drawn. First, given that poverty (wealth) is associated with food insecurity (safety), one can expect low-SES consumers to put more weight on fillingness than high-SES consumers. Second, assuming that taste is an attribute few are willing to give up on, the increased importance given to fillingness among low-SES consumers should come at the expense of healthiness rather than taste. Formally:

Incipient work has hinted at the idea that food priorities differ across the SES spectrum (Bratanova et al. 2016; Hwang and Choe 2016; Kershaw et al. 2019). Evidence suggests that fillingness bears particular importance among low-SES consumers who choose foods by their perceived filling capacity (Laran and Salerno 2013), display an increased ability to discriminate foods based on their energy density (Cheon et al. 2018; Sim et al. 2018), and believe that larger portions are more appropriate (Best and Papies 2019). However, no research has explicitly and systematically examined how fillingness interacts with other relevant food attributes such as taste and healthiness. Further, the literature often implies or explicitly suggests a trade-off between healthiness and taste but not fillingness (Kershaw et al. 2019; Thomas, Ma, and Gauri 2024). In this research, we fill this gap. It is important to note that while taste and fillingness tend to be positively correlated, examples of unhealthy products that are arguably perceived as tasty but not filling abound (e.g., M&M’s, popcorn, and candy in general), allowing for an analysis of the trade-offs among these attributes. Finally, research is typically limited in terms of the types of foods and the attribute elicitation methods used. In this research, we offer evidence for the primacy of fillingness in low-SES consumers’ minds across qualitative and quantitative studies and directly examine attribute trade-offs via explicit rankings and inferences drawn from conjoint experiments.

SES and Attribute Associations

High- and low-SES consumers may differ not only in the importance they give to intrinsic food attributes but also in the perceived strength of the associations between attributes. Whereas heuristics linking different food attributes (e.g., “healthier = less tasty” and “healthier = less filling”) have proved to be widely used by consumers (Raghunathan, Naylor, and Hoyer 2006; Suher, Raghunathan, and Hoyer 2016) and consequential to food choices (Briers et al. 2020; Paakki et al. 2022; Turnwald and Crum 2019; Turnwald et al. 2019), the role of SES in shaping the strength of these associations has not yet been systematically investigated. However, it is plausible to expect that these widely held beliefs may be particularly strong among low-SES consumers and possibly in ways that hinder healthier eating.

Due to limited resources, low-SES consumers are less likely to experiment with cooking or try new food types (Beagan, Power, and Chapman 2015), and they tend to value more familiarity than their high-SES peers when making food decisions (Baumann, Szabo, and Johnston 2019; Daniel 2016; Steptoe, Pollard, and Wardle 1995). Further, they are, in general, exposed to a narrower range of food options than their high-SES counterparts (Bardin, Washburn, and Gearan 2020; Story et al. 2008). This discrepancy in food knowledge, exposure, and inclination to try new possibilities can have important consequences for the associations made by low- and high-SES individuals. Research has shown that a weaker propensity to experience new aspects of life (including new food items) can exacerbate prior heuristics (Haws, Reczek, and Sample 2017; Jung and Kellaris 2004). The less one learns from new experiences, whether for lack of motivation or opportunity, the less likely they are to challenge preconceived associations (Caputo 2014; Flynn 2005). Thus, many well-established beliefs and generalizations with respect to the correlation between food attributes are plausibly stronger and/or more widely accepted among low- versus high-SES consumers. Based on these arguments, we hypothesize that:

One may wonder why we predict no SES differences in the importance of taste in H_1b but a stronger reliance of low-SES consumers on the “healthy = not tasty” intuition in H₂. It is worth clarifying that while H₁ focuses on attribute weights, H₂ deals with attribute associations. This is a critical distinction because two groups can attribute similar importance to taste (i.e., via weights [H_1b]) and still differ in their perception of how tasty healthy and unhealthy foods are (i.e., via associations or heuristics [H₂]), which help explain disparities in choice among members of these groups.

Consistent with our prediction, Pivecka, Kunz, and Florack (2023) find that low- (vs. high-) SES consumers display a stronger “healthy = less tasty” association. Further, Thomas, Ma, and Gauri (2024) show that stronger food cravings reduce the unhealthiness perceptions of hedonic foods. We contribute to this incipient body of work on two fronts. First, we offer a methodological contribution by testing the robustness of the phenomenon with other stimuli, using samples of deprived and affluent residents of a developing country. Second, and most critically, we extend these results by generalizing heuristics to another critical food association (i.e., “healthier = less filling”).

We preregistered all studies. The preregistration protocol, data, code, and materials are available on OSF (https://osf.io/8st4u/). Table 1 displays an overview of the studies.

OPEN IN VIEWER

Table 1.

Overview of the Studies.

Study	Sample	Design	Dependent Variable	Main Findings
1	358 residents from wealthy and deprived regions of Rio de Janeiro, Brazil	2 (neighborhood SES: low vs. high) × 3 (food pair replicates) between-subjects	Consequential food choice (selection of healthy or unhealthy food from pair)	Low-SES consumers chose healthy foods to a much lower extent than high-SES consumers.
2a	247 residents from wealthy and deprived regions of Rio de Janeiro, Brazil	Single factor, between subjects (neighborhood SES: low vs. high)	- Hypothetical food choice between two dishes (healthy and unhealthy) - Thought listing, open-ended justifications for choices	Low-SES consumers chose the healthy dish to a much lower extent than high-SES. consumers. Low-SES consumers mention fillingness more often and healthiness less often as a justification for their choice relative to high-SES consumers.
2b	101 residents from a deprived region of Rio de Janeiro, Brazil	Survey	- Fillingness importance when making food choices - Reasons for fillingness importance	- Most low-SES consumers considered fillingness an important attribute. - Financial considerations and the feeling of satiety were the most important reasons.
3a	200 residents from wealthy and deprived regions of Rio de Janeiro, Brazil	Single factor, between subjects (neighborhood SES: low vs. high)	Food ranking of three attributes: healthiness, fillingness, and taste	Low-SES consumers prioritized fillingness to a greater extent and considered healthiness as a less important attribute than high-SES consumers. No SES differences in the importance of taste.
3b	383 participants from a Brazilian online panel	Single factor, between subjects (SES/environmental harshness manipulation: low vs. high)	Food ranking of three attributes: healthiness, fillingness, and taste	Consumers in the low- (vs. high-) SES condition prioritized fillingness to a greater extent and considered healthiness as a less important attribute. No SES differences in the importance of taste.
4	250 residents from wealthy and deprived regions of Rio de Janeiro, Brazil	2 (neighborhood SES: low vs. high, between) × conjoint analysis with three attributes (healthiness, fillingness, taste) and three levels (low, moderate, and high)	Selection of a product from a hypothetical pair varying on taste, healthiness, and fillingness	Low-SES consumers prioritized fillingness to a greater extent and healthiness to a lower extent than high-SES consumers. No SES differences on the importance of taste.
5	300 residents from wealthy and deprived regions of Rio de Janeiro, Brazil	2 (neighborhood SES: low vs. high, between) × 2 (food items: healthy vs. unhealthy, within) mixed design	Perceptions of healthiness, fillingness, and taste	Low-SES consumers perceived healthier foods as less filling and less tasty than unhealthy ones when compared with high-SES consumers.
6	1,063 residents from a deprived region of Rio de Janeiro, Brazil	Single factor, between-subjects (control vs. high fillingness)	Food choice between healthier and unhealthier options	High-fillingness (vs. control) condition can increase low-SES consumers’ choice of the healthier option, but the effect varies across types of intervention.

Method

Results

Food choices

To assess how high- and low-SES individuals differed in their food preferences, we conducted logistic regressions including SES as an independent variable (1 = low SES, 0 = high SES) and food choice (1 = healthy food, 0 = unhealthy food) as the dependent variable, controlling for age, gender (1 = female, 0 = other), marital status (1 = married, 0 = other), BMI, and current level of hunger. We ran this model considering all three pairs as our focal analysis and, as a robustness check, we exploratorily examined each pair separately (Figure 1).

OPEN IN VIEWER

Figure 1.

Food Choices Across SES (Study 1).

Low-SES consumers chose the healthy food items to a much lower extent than high-SES consumers (β = −1.78, SE = .24; 95% CI = [−2.26, −1.30], p < .001, OR = .17; Table W5). Exploratory analyses revealed similar results irrespective of product pair (Web Appendix C) and showed that when comparing the proportion of healthy versus unhealthy choices within each group, low-SES consumers preferred the unhealthy options (61.80%) over the healthy ones (38.20%, p < .001), whereas high-SES consumers displayed the opposite preference (22.78% vs. 77.22%, p < .001).

Discussion

Study 1 shows that, over and above differences in supply, differences in preferences may also influence the dietary choices of high- and low-SES consumers. Low-SES participants chose unhealthy foods to a much greater extent than their high-SES peers. This occurred for different pairs of foods, including branded and unbranded items. Although the current study provides evidence for the importance of demand-side considerations in nutritional inequality, it does not shed any light on the psychological reasons underlying these differences.

Method

Results

Thought-listing protocol

To assess how high- and low-SES individuals’ justifications for food choices differed, we conducted logistic regressions with SES as an independent variable and each of the coded justifications as dependent variables, using the same covariates. Overall, low- (vs. high-) SES consumers cited healthiness to a much lower extent (19.20% vs. 80.33%; β = −2.50, SE = .39; 95% CI = [−3.27, −1.73], p < .001, OR = .08; Figure 2). Critically, fillingness justifications were overwhelmingly present among low-SES consumers (79.20%), but none of the high-SES consumers mentioned it. Given the absence of variability in the dependent variable among the affluent, we were unable to run a logistic regression, but a proportions test revealed that low-SES consumers considered fillingness to a much greater extent (p < .001). Exploratory results showed that low-SES participants cited taste more often, but spontaneous considerations of this attribute were relatively high for both groups (85.60% vs. 59.02%, p < .001), which resonates with the hypothesis that consumers, in general, are unwilling to give up on taste. Although differences appeared, the other aspects were considered to a much lesser extent (Figure 2).

OPEN IN VIEWER

Figure 2.

The Influence of SES on Aspects Considered in a Thought-Listing Task (Study 2a).

Discussion

Study 2a replicates the finding that low-SES consumers choose unhealthy foods to a much greater extent than their high-SES counterparts using a meal rather than a single product choice. Further, it shows the importance of healthiness, fillingness, and taste in food decision-making (the three most mentioned attributes). Critically, it reveals that thoughts of fillingness and health, more so than taste, represent the main differentiators between the two SES groups. Thus, this study offers initial evidence that the primacy of fillingness among low- (vs. high-) SES consumers may help explain differences in food preferences.

Method

Results

Not surprisingly, and in line with Study 2a, fillingness was considered important by 89 out of 101 low-SES consumers (88.12%; p < .001; against 60%).

To assess the importance of each potential reason why fillingness is a relevant attribute, we conducted one-sample t-tests contrasting the mean score of financial considerations and feeling of satiety against both the scale midpoint (4) and the other three reasons (familiarity, health concerns, and social signaling; Figure 3). As expected, feeling of satiety (M_satiety = 6.82, SD_satiety = .95) was significantly higher than the scale midpoint (i.e., 4, p < .001), and familiarity (M_familiarity = 4.22, SD_familiarity = 2.86), health (M_health = 3.32, SD_health = 2.79), and social signaling reasonings (M_signaling = 1.97, SD_signaling = 1.77). Likewise, the same pattern was observed for financial considerations (M_financial = 6.91, SD_financial = .65; all p's < .001). An exploratory analysis contrasting the feeling of satiety with financial considerations shows that both were equally highly valued (p = .39).

OPEN IN VIEWER

Figure 3.

Reasons Fillingness Is Important Among Low-SES Consumers (Study 2b).

Discussion

Study 2b shows that low-SES consumers largely value filling foods not only for financial reasons but also because they highly value feelings of satiety/a full stomach. These factors take precedence over other plausible possibilities, such as familiarity, health considerations, and social signaling. Our next studies demonstrate how fillingness considerations shape food choices by shedding light on the roles of attribute importance and associations.

Participants and Procedure

Two hundred participants took part in this study (M_age= 41.90 years, SD = 18.75; 45% female, 55% male). Half of them were residents of the South Zone of Rio de Janeiro, and the other half lived in Favela da Maré. Research assistants unaware of the study’s hypothesis approached passersby in the street and asked whether they were willing to take part in a short survey. The research assistants provided those who agreed with a general context (food purchase) and brief explanations about three different food attributes, presented in random order: healthiness (“how healthy is a food item, that is, how good it is for the health of the person eating it”), fillingness (“how filling is a food item, that is, how much it satiates hunger for a good chunk of time”), and taste (“how tasty is a food item, that is, how appetizing it is to eat it”). Then, participants ranked the items in order of importance:

When purchasing food items, what importance do you generally attribute to each of these three aspects? Please indicate (1) for the most important, (2) for the second most important, and (3) for the least important.

Results

We conducted separate ordered logistic regressions to assess how SES shapes the weight given to the three food attributes (i.e., healthiness, fillingness, and taste). The models included SES as the independent variable (1 = low SES [deprived neighborhood], 0 = high SES [wealthy neighborhood]), BMI, current level of hunger, gender (1 = male, 0 = female/other), marital status (1 = married, 0 = not married), and age as control variables, and the ranking measure for the attribute under consideration as the dependent variable. For ease of interpretation, we reversed the ranking measure so that higher scores indicate greater relative importance.

In line with H_1a, low-SES consumers prioritized fillingness to a greater extent than high-SES consumers (β = 1.00, SE = .48; 95% CI = [.05, 1.95], p = .038, OR = 2.72) but considered healthiness as a less important attribute (β = −1.22, SE = .51; 95% CI = [−2.22, −.21], p = .017, OR = .29). Consistent with H_1b, there were no SES differences on the relative importance of taste. Both high- and low-SES consumers reported weighing heavily on this attribute (β = .07, SE = .43; 95% CI = [−.77, .91], p = .872, OR = 1.07; see Figure 4, Panel A). For analyses with alternative model specifications and interactions, see Web Appendix E.

OPEN IN VIEWER

Figure 4.

The Influence of Measured and Manipulated SES on Attribute Importance/Weights (Studies 3a and 3b).

Participants and Procedure

Two hundred seventy-three participants took part in this online study. We recruited participants from Netquest, a large online panel. We targeted participants with household incomes below R$4,000 (about four minimum wages) to ensure our manipulation would be arguably credible for all respondents. Following the preregistration protocol, we excluded 19 participants who did not consent to participate and 71 participants because they reported a household income above the target level. Our final sample thus includes 283 participants (M_age = 39.38 years, SD = 12.73; 65% female, 35% male). The survey procedure was the same as in Study 3a, and participants also ranked the attributes in order of importance as the main dependent variable.

Results

We conducted the same analysis as in Study 3a. In line with H_1a, low-SES consumers prioritized fillingness to a greater extent than high-SES consumers (β = .65, SE = .22; 95% CI = [.20, 1.09], p = .004, OR = 1.91) and considered healthiness as a less important attribute (β = −.49, SE = .22; 95% CI = [−.93, −.05], p = .029, OR = .61). Consistent with H_1b, there were no SES differences on the relative importance of taste. Both high- and low-SES consumers reported weighing heavily on this attribute (β = −.13, SE = .22; 95% CI = [−.57, .30], p = .541, OR = .87; see Figure 4, Panel B).

Discussion

Studies 3a and 3b show that low- and high-SES consumers report weighing food attributes differently when choosing what to eat, regardless of whether SES is measured or manipulated. Relative to high-SES consumers (or those prompted to feel high SES), low-SES consumers (or those prompted to feel low SES) report giving more importance to fillingness, less importance to healthiness, and equally high importance to taste (consistent with H₁). Our results thus indicate that, by prioritizing satiety, low-SES consumers tend to give up on nutritional value rather than palate.

The measurement of attribute importance often asks for a multimethod examination (Van Ittersum et al. 2007). Although direct rating and ranking approaches are simple and easy, they assume respondents are willing to and capable of properly assessing attribute weights and trade-offs. Another common approach is to indirectly infer attribute weights from choices (Ben-Akiva, McFadden, and Train 2019). Thus, in Study 4, we rely on a choice-based conjoint analysis to assess the robustness of the results obtained in Studies 3a and 3b.

Method

Results

We performed three separate logistic regressions for each food attribute to test our hypotheses. The models included food choice as the dependent variable (1 = chosen, 0 = not chosen) and SES (1 = low SES, 0 = high SES), dummies for moderate and high levels of the attribute under analysis (the lower level served as the baseline), and the two-way interactions between SES and attribute dummies as independent variables. We also controlled for choice set and attribute order effects, sociodemographic characteristics, and diet-related information, and we clustered the standard errors at the individual level. Although we report the results of separate regression models for ease of interpretation, Tables W13–16 in Web Appendix G present the results of a single model containing all six two-way interactions.

In line with H_1a, low-SES consumers chose filling foods to a greater extent than high-SES consumers (Figure 5, Panel A). The significant interaction between SES and the medium-fillingness dummy (β = .96, SE = .18; 95% CI = [.60, 1.32], p < .001, OR = 2.62) and the even stronger effect for the high-fillingness dummy (β = 1.21, SE = .21; 95% CI = [.81, 1.62], p < .001, OR = 3.37) corroborate our hypothesis. Simple slopes analyses show that low-SES consumers chose low-fillingness foods less than high-SES consumers (β = −.72, SE = .11; 95% CI = [−.95, −.50], p < .001, OR = .48), but this effect reverses for medium-fillingness foods (β = .24, SE = .10; 95% CI = [.04, .44], p = .019, OR = 1.27) and becomes even more positive for high-fillingness foods (β = .49, SE = .13; 95% CI = [.24, .74], p < .001, OR = 1.63).

OPEN IN VIEWER

Figure 5.

Attribute Importance Inferred from Food Choices Across the SES Continuum (Study 4).

Further, as anticipated in H_1b, low-SES consumers were less likely to choose healthier food items than high-SES consumers, as revealed by the negative and significant interaction between SES and the medium-health dummy (β = −1.02, SE = .21; 95% CI = [−1.44, −.61], p < .001, OR = .36) and the even stronger interaction effect for the high-health dummy (β = −2.02, SE = .22; 95% CI = [−2.46, −1.59], p < .001, OR = .13; Figure 5, Panel B). Again, simple slopes analyses support our predictions by showing that low-SES consumers chose low-health foods more than high-SES consumers (β = 1.01, SE = .13; 95% CI = [.75, 1.26], p < .001, OR = 2.74), but this effect reverses for highly healthy foods (β = −1.02, SE = .12; 95% CI = [−1.26, −.77], p < .001, OR = .36). No significant differences were found for moderately healthy foods (β = −.02, SE = .11; 95% CI = [−.24, .20], p = .873, OR = .98). Finally, again in line with H_1b, taste did not differentially influence choices across social classes (Figure 5, Panel C). Neither the interaction between SES and medium-taste dummy (β = .20, SE = .20; 95% CI = [−.19, .60], p = .312, OR = 1.23) nor the interaction effect for the high-taste dummy reached significance (β = −.11, SE = .21; 95% CI = [−.54, .31], p = .593, OR = .89).

Discussion

In a choice-based context, Study 4 provides further evidence that SES shapes the relative importance of food attributes and does so in a systematic manner. Relative to their high-SES counterparts, low-SES consumers in our sample were more likely to give weight to fillingness (H_1a), less likely to care about healthiness, and as highly concerned about taste (H_1b) when making food choices. Study 5 takes a step further to explore whether high- and low-SES consumers also vary in the strength of the perceived associations between these attributes.

Method

Results

Strength of attribute associations: healthier = less filling

We regressed fillingness perceptions on SES (1 = low SES, 0 = high SES), type of food (1 = healthy, 0 = unhealthy), and the interaction between them using a linear regression model with diet and sociodemographic controls and standard errors clustered at the individual level. The interaction reached significance (β = −1.79, SE = .14; 95% CI = [−2.06, −1.52], p < .001, r = .60). As Panel A of Figure 6 shows, low-SES consumers perceived healthier foods as significantly less filling than unhealthy foods (M_{LowSES Healthy} = 3.53, SD = 1.34; M_{LowSES Unhealthy} = 4.77, SD = .81; β = −1.23, SE = .12; 95% CI = [−1.47, −.99], p = <.001, r = .64). This effect was not only weaker but also reversed among high-SES consumers; that is, they perceived healthy foods as significantly more filling than unhealthy ones (M_{HighSES Healthy} = 5.97, SD = .74; M_{HighSES Unhealthy} = 5.41, SD = .42; β = .55, SE = .06; 95% CI = [.43, .68], p = <.001, r = .58). Although the interaction is consistent with H₂, the reversal (rather than attenuation) among high-SES consumers was unexpected (see Web Appendix H for a summary of individual scores across socioeconomic groups and a justification of the reversal and the role of ice cream in explaining it).

OPEN IN VIEWER

Figure 6.

The Influence of SES and Type of Food on Perceived Fillingness and Taste (Study 5).

Discussion

Study 5 provides evidence that SES systematically shapes food attribute heuristics. Low- and high-SES consumers vary in the strength with which they associate different food attributes, with low-SES consumers strongly perceiving healthy food items as particularly not filling and not tasty compared with high-SES consumers (H₂).

Method

All four intervention studies were preregistered and tested how increasing the fillingness of healthy options influenced subsequent food choices. However, not all interventions shared the same design (e.g., in Intervention 1, baseline preferences were assessed in wealthy and deprived regions, whereas Intervention 4 focused on the deprived region only). For the sake of parsimony, the procedures and results reported here emphasize the comparison between intervention and control conditions within the deprived region, Favela da Maré, where the problem of unhealthy eating is prominent. Specifics about the stimuli and designs used in each study, the preregistered analyses, and the supplementary statistical details are described in Web Appendix I.

Results

To assess the effectiveness of the high-fillingness interventions, we conducted a series of logistic regressions including experimental condition as an independent variable (1 = high fillingness, 0 = control) and food choice (1 = choice of healthy food, 0 = choice of unhealthy food) as the dependent variable. We also conducted a one-tailed Bayes factor (BF₁₀) analysis to evaluate the evidence for our intervention results.

The results are described in Figure 7. Relative to its respective control condition, the portion size intervention (Intervention 1) proved to be effective in promoting healthy eating among low-SES consumers (β = 3.36, SE = 1.03; 95% CI = [1.33, 5.38], p < .001, OR = 28.69, BF₁₀= 1,942.75). The description via message claim on the package intervention (Intervention 2) was marginally significant (β = .43, SE = .26; 95% CI = [−.07, .94], p = .091, OR = 1.54, BF₁₀ = .97), pushing low-SES consumers into healthier dietary choices. Finally, neither the description via advertisement intervention (Intervention 3, β = .06, SE = .36; 95% CI = [−.65, .76], p = .874, OR = 1.06, BF₁₀ = .14) nor the food swap intervention (Intervention 4, β = .01, SE = .27; 95% CI = [−.52, .54], p = .972, OR = 1.01, BF₁₀ = .14) reached significance. The Bayes factor analysis highlights that the portion size intervention (Intervention 1) had a substantial effect, as indicated by a high BF₁₀ value. The description via message claims and advertisement showed either an inconclusive result with a BF₁₀ value close to 1 (Intervention 2) or strong evidence for the null hypothesis (Intervention 3). Finally, the food swap intervention (Intervention 4) had low BF₁₀ values, suggesting strong evidence for the null hypothesis.

OPEN IN VIEWER

Figure 7.

Effectiveness of High-Fillingness Interventions Among Low-SES Consumers (Study 6).

Discussion

Study 6 shows that the effectiveness of fillingness-focused interventions varies depending on the specific strategy used. In line with findings from Cadario and Chandon (2020), behaviorally oriented interventions in the form of portion size enhancements are particularly promising (Intervention 1). None of the other interventions seemed particularly promising, as reflected in the Bayes factor analysis.

Theoretical Contributions

Although emerging research has uncovered the important and often overlooked role of demand in guiding SES differences in dietary choices (Allcott et al. 2019; Bratanova et al. 2016; Cheon and Hong 2017; Laran and Salerno 2013; Thomas, Ma, and Gauri 2024), no study to date has systematically assessed how exactly poverty helps shape food preferences and perceptions. Our results provide evidence that poverty influences not only the relative importance of specific food attributes but also the perceived strength of the associations between these attributes. By showing that relative to high-SES consumers, low-SES consumers care more about fillingness (but not taste) than healthiness and are particularly prone to believe that healthy items are neither filling nor tasty, our findings help explain why low-SES consumers display unhealthier diets than high-SES consumers. In addition, we contribute to the literature on food heuristics. While previous research shows that consumers rely on food heuristics when deciding what to eat (Raghunathan, Naylor, and Hoyer 2006; Suher, Raghunathan, and Hoyer 2016), we demonstrate that negative attribute associations involving healthy foods (e.g., healthy = not filling) are not homogeneous across consumer segments. Relative to high-SES consumers, low-SES consumers tend to believe that healthy items are neither tasty nor satiating. Lastly, we address recent appeals for (1) further research on the psychology of low-SES consumers (Carey and Markus 2016; Pham 2016; Shavitt, Jiang, and Cho 2016), (2) deeper examination of the food insecurity agenda (Food and Agriculture Organization of the United Nations et al. 2020), and (3) investigation of food decisions in more naturalistic contexts (Haws et al. 2022). Although food insecurity affects about 25% of the world's population (Food and Agriculture Organization of the United Nations et al. 2020), research on the food decisions of high- and low-SES consumers has not yet systematically examined this group. Most of the research in the area investigates how induced states of scarcity or subordinate rank (Bratanova et al. 2016; Cheon and Hong 2017; Laran and Salerno 2013) influence decision-making. We instead investigate how chronic and extreme experiences with scarcity and abundance help shape food-related decisions.

Societal and Marketing Implications

Our findings show that even in environments where opportunities and resources are similar, low-SES individuals still prefer unhealthier food options than high-SES individuals. This is consistent with recent studies showing that the influence of food cost and availability on nutritional inequality may have been overstated (Cummins, Flint, and Matthews 2014; Handbury, Rahkovsky, and Schnell 2016; Pechey and Monsivais 2016) and that much of the variance across the socioeconomic spectrum seems to result from differences in preferences for (un)healthier foods (Allcott et al. 2019). Policy makers, however, have primarily focused on aspects related to the supply of healthy foods (Bitler and Haider 2010; Handbury, Rahkovsky, and Schnell 2016). For instance, in the United States between 2010 and 2019, policies such as the Healthy Food Financing Initiative, which provides resources to improve access to and affordability of healthy foods in underserved communities, have leveraged more than $1 billion (Allcott et al. 2019). Likewise, the Biden–Harris Administration’s national strategy on hunger, nutrition, and health (White House 2022) disproportionally focused on promoting access to and affordability of healthy foods. More broadly, the Food and Agriculture Organization of the United Nations et al. (2023) suggests restricting the marketing of unhealthy foods and beverages and enforcing health taxes. Thus, while most policy interventions are tailored to make healthy foods cheaper and conveniently available to low-SES consumers, which are important and valuable goals, our findings suggest that combining supply and demand forces is critical. Specifically, even though making healthy foods more accessible and affordable or restricting access to unhealthy foods may influence the dietary behaviors of low-SES consumers, increasing the availability of healthy items that are convincingly filling may be an equally promising avenue for promoting healthier diets. Along these lines, one key policy implemented by Brazil’s national government is the creation of a food basket that contains healthy, affordable, and nutritious food staples that are especially filling (e.g., rice and beans). Recently, taxes for these products were reduced to 0% to encourage their consumption among lower-income consumers. While this strategy resonates with the findings from our studies, governments and marketers could also explicitly highlight the fillingness of the healthy items in this basket.

The food industry has long used fillingness in its communication strategies. This approach, however, has been primarily implemented to promote unhealthy items. For instance, Burger King has used strategies such as “Aren’t you hungry?,” “Hungry yet?,” and “Are you hungry or starving? Starving!!” to advertise its hamburgers. Likewise, the frozen food brand Hungry Man often advertises its fried chicken as “Filling, no bones about it,” and mentions that “Hunger, this town ain’t big enough for the two of us.” Snickers chocolates are often marketed via statements such as “You’re not you when you’re hungry! Snickers satisfies,” and questions such as “How whiny does hunger make you?” Based on our findings, however, marketers could also rely on similar communication strategies to advertise healthy foods.

Study 6 tested a set of exclusively demand-side interventions. This study shows that although it is possible to increase the choice of healthier foods among low-SES consumers if these options are perceived as filling, food preferences are difficult to shift, and interventions vary in effectiveness. These results dovetail with recent meta-analytic findings on the success of different nudges in promoting healthier diets. In line with Cadario and Chandon (2020), we found that behaviorally oriented nudges such as portion size enhancements are particularly effective in inducing healthy eating. Cognitively oriented strategies in the form of descriptive labels offered mixed findings, with results pointing to either a small positive or a null effect. Finally, increasingly used by marketers and policy makers (IGD 2023), the food swap intervention did not lead to significant differences between the control and high-fillingness conditions. However, it is worth noting that across conditions, a nontrivial percentage (31%) of our participants swapped their initial choice for a healthier version. We suspect that the social pressure due to the face-to-face setting of our studies likely washed away possible intrinsic differences in attribute weights. Consistent with this rationale is the finding that when effective in e-commerce or self-checkouts, food swaps tend to have relatively small effects and are often limited to a few product categories (IGD 2023; Koutoukidis et al. 2019; Schruff-Lim et al. 2024).

In testing the portion size enhancement intervention, we show that while healthy foods are not seen as particularly tasty or filling per se, introducing vegetables to a dish can increase its overall perceived fillingness and, as a result, push low-SES consumers to healthier diets. This intervention has theoretical and practical implications. Whereas previous studies examining portion size nudges focus on how increasing plate or serving size affects energy intake (Diliberti et al. 2004; DiSantis et al. 2013; Van Ittersum and Wansink 2012), the current study shows that adding healthy items to a dish that was already perceived as healthy can shift cross-class food preferences. In other words, portion size interventions can increase the willingness to consume healthy over unhealthy foods rather than simply their consumption within a given category. While one could argue that our intervention simply encouraged low-SES participants to choose the healthy option based on its value for money, Study 2b shows that the feeling of satiety is as important as financial considerations among low-SES consumers, and as such, high-fillingness interventions are likely to add value over and above price considerations. Another important aspect concerning this type of intervention refers to its actionability. Indeed, unlike sheer framing interventions, increasing portion size involves enhancing costs for marketers. To address potential cost barriers, marketers could search for high-margin products that are filling and healthy (such as the basic food basket in Brazil outlined previously), and, most critically, governments could subsidize (or reduce the tax of) nutritious and satiable foods.

Limitations and Future Research

The results of this work can help develop several other food-related research inquiries. While our investigation is focused on the demand side of dietary choices, it is also possible that other factors driven by supply-side barriers help shape demand for food items. For example, the lack of familiarity and previous experiences with healthy foods among low-SES consumers might be initially driven by their higher prices or lack of availability in low-SES neighborhoods. But as unfamiliarity leads to greater reliance on heuristics (Haws, Reczek, and Sample 2017), we reason and find that it also helps shape food preferences (i.e., the demand). Thus, future research might investigate how this cycle of external barriers, internal preferences, and food choices unfolds in a longitudinal study.

Further, the effectiveness of our interventions proved to be widely dependent on the strategy used. Although the portion enhancement study seemed particularly promising, one could also raise a few potential alternative explanations. For example, since the high-fillingness healthy option also included more ingredients, variety seeking could explain our results. However, because healthy foods are not perceived as palatable by low-SES consumers, it is unlikely that the inclusion of healthy items in a dish influences their decisions via the increased variety of not-as-palatable ingredients rather than the fillingness increment that such inclusion provides. Further, one could also argue that our results are a sheer consequence of the fact that, after the inclusion of more ingredients, the dish offers a more attractive value for money. While reasonable, the results from Study 2b show that fillingness is relevant to low-SES consumers not only due to financial considerations but also—and to the same extent—because these consumers value the feeling of satiety. Notwithstanding these caveats, we encourage future research to devise and test new interventions to promote healthier diets. Following Cadario and Chandon (2020) meta-analytical results, future research could focus on promising strategies and test other behaviorally oriented interventions.

Along these lines, despite favoring taste similarly, low-SES individuals are predicted to hold particularly negative perceptions about the taste of healthy items relative to high-SES individuals. Thus, we should expect low-SES consumers to increase their choices of healthy foods when these items are also perceived as tasty. In other words, successfully increasing the perceived palatability of a healthy food item should increase its attractiveness among low-SES consumers. Although previous studies have shown that labeling healthy food items with taste-focused messages not only increases their consumption (Turnwald et al. 2019; Turnwald and Crum 2019) but also enhances expectations of taste experience (Turnwald et al. 2019), this corollary hypothesis has not yet been tested within the SES context. If supported, taste-focused labeling on healthy food options could also help policy makers design interventions to reduce nutritional inequality.

In summary, although often neglected in relevant food policies, the role of SES differences in food preference and perceptions of nutritional inequality cannot be understated. As the current research shows, intrinsic preferences for fillingness and perceptions about healthy foods among low-SES consumers are consequential to food choice and, potentially, to a plethora of health outcomes (Darmon and Drewnowski 2015). As such, learning more about how SES-based differences in preferences and perceptions shape decision-making may help researchers more deeply understand the phenomenon and policy makers develop more effective interventions that will ultimately reduce pervasive nutritional inequality.