To successfully navigate the obesogenic food landscape requires constant vigilance, a task that is cognitively depleting and therefore difficult to perform consistently [2, 50]. By rearranging the food environment in a manner that facilitates healthier choices, nudging might help to counter the environmental push to select unhealthy options. This study found mixed evidence for the efficacy of three approaches to nudging healthy dietary choices at a population level. Overall, single or multiple nudges, or multiple nudges concurrent with price reductions did not influence the sale of healthy items at a community pool. Direct observations of a subset of patrons’ purchases, however, showed an approximately 30% increase in sales of healthy items when a signage + taste testing intervention was implemented, an increase that was maintained when prices of healthy items were reduced by 30%, and even when all interventions were removed.
It is unclear why results differed in the full and subsamples. The subsample captured a large proportion of total purchases during the study (40.7%), albeit still a minority. As previously described, inclusion in the subsample was determined exclusively by the time of day purchases were made, and coincided with the busiest times of day. Patrons who purchased items during these times may therefore have differed from those who purchased items at other times in a manner that made them more responsive to changes in the food environment. All observations included the lunch time period, and thus it is possible that patrons may have been selecting a meal or a supplement to a meal, as opposed to a treat or a snack. The proportion of healthy items purchased by individuals in the full (40.8%) and subsample (41.3%) was not different, however. It is also possible that item misclassification on the part of observers might account for our mixed findings. This appears unlikely, however, because there was high congruence between observers and they could visually see all items being ordered and the printed list of items on receipts. It is also possible that patrons ordered healthy items because observers were present, although this too appears unlikely as observers were present during both control and intervention periods and their presence and purpose were not readily apparent. Differences in the statistical procedures used to analyse the two data sets might also be responsible, although our methods are consistent with others .
We implemented two nudges that have received little study, finding mixed evidence for their efficacy. We first tested the efficacy of descriptive menu labels, a strategy commonly used by the restaurant industry to improve consumers’ taste expectations . Wansink et al.  previously showed that sales of targeted (healthy and unhealthy) items in a University faculty cafeteria increased by 27% when they were given descriptive menu labels. We created descriptive menu labels for healthy items only, and although we increased the size of the signs on which they were placed, there was no impact of this change on sales of healthy items. When a taste testing intervention was added the results were mixed, as there was no apparent impact in the full sample, however sales in the subsample increased by 30% relative to baseline. The latter findings are in line with studies showing that repeated exposure to healthier foods can counter children’s naturally neophobic tendencies , and that free product samples distributed to neophobic young adults can increase selection of unfamiliar healthful food products by 14.6% . Our findings in the subsample suggest an even stronger impact of taste testing, as participants were in a natural setting where they were required to pay for a full portion of the healthy items they had tasted. Qualitative studies suggest that children are particularly reluctant to expend money on new food choices or on fruit and vegetables which might taste bad, as opposed to packaged junk foods which always taste the same . Our findings suggest taste testing might help to reduce this perceived risk, thereby nudging purchase of novel, healthier items in some community settings.
Systematic reviews have found subsidies/price reductions on healthier foods to be an effective means of increasing their purchase and consumption in a variety of settings [44, 52, 53]. This has not always proven to be the case in single [54, 55], or combined interventions [44, 52], however, suggesting that price may not always drive food purchases and that some populations are more price sensitive than others. In particular, low income populations, for whom food represents a larger proportion of total expenditures, are predictably more price sensitive [56–60].
In the present study, sales of healthy items remained constant in the full and subsamples when a pricing intervention was added, suggesting that price reductions did not incent purchase of healthy items in this context. Although it is possible that the impact of the signage + taste testing intervention in the subsample might have waned had a price reduction not been added, it appears more probable that price reductions were ineffective because our study sample likely represented a population with higher socioeconomic status (SES). There are several reasons to suspect that pool patrons represented a higher SES group. The study took place in one of the wealthiest jurisdictions in North America [61, 62] and the pool was proximal to several wealthy neighborhoods. In addition, the pool was not readily accessible on foot or via public transit, and had a relatively high entrance fee, factors that have been shown to deter youth in low SES groups from participating in physical activity . Observers also noted that many families appeared well off, paid cash for their purchases, and that few children were overweight/obese. Higher SES populations might not perceive a 30% financial savings to be worth the non-monetary costs (poorer taste, reduced satisfaction) of consuming healthy items. Alternatively, price reductions might have been more effective had other healthy items been targeted, as the efficacy of price reductions differs by item [54, 64]. Many of the healthy items targeted in this study contained fruit, and the demand for fruit is relatively inelastic [56, 57, 65, 66]. Finally, given that the final price of discounted items was not posted (ie. signage placed on healthy items indicated that they were 30% off), it is possible that children or others with limited numerical skills did not understand the potential savings to be had.
Although the addition of a pricing intervention did not appear to influence purchase of healthy items overall, individuals in the subsample who were overweight/obese exhibited a greater sensitivity to the addition of a pricing intervention relative to individuals who were normal weight. It is possible that the greater sensitivity of overweight/obese persons to the signage + taste testing + price reduction intervention may reflect heightened price sensitivity due to their relatively lower SES, as individuals of lower SES tend to have higher body weights compared to those of higher SES [67, 68]. These results contrast with those of Epstein et al. , who observed that obese mothers were less price sensitive than their nonobese counterparts. Males in the subsample also exhibited greater sensitivity to the addition of a pricing intervention relative to females. It is not clear why this was the case. Moderators of price sensitivity have rarely been examined, and gender-specific impacts of pricing interventions were not reported in a recent systematic review . Subgroup-specific findings should be interpreted with caution, however, in light of discrepancies in findings between the full and subsamples, and the moderate kappa coefficients for inter-rater reliability on demographic variables. Our findings in this respect highlight the need to consider effect modification in future studies.
In the full sample sales of healthy items did not differ in the pre- and post-intervention phases, however in the subsample sales of healthy items remained approximately 30% above baseline values in the post-intervention period. This result may indicate that patrons in the subsample who increased their purchase of healthier items in the signage + taste testing and signage + taste testing + price reduction phases learned to prefer the healthier menu items, and therefore continued to purchase them when all the interventions were removed. Alternatively, if the same patrons did not frequent the pool on a weekly basis the finding that sales of healthy items remained 30% above baseline values in the post-intervention phase might indicate that something other than the interventions, such as differences in the clientele or an unobserved change at the pool, was responsible for the ~30% increase in sales of healthier items observed. A marked shift in the clientele coinciding with the start of the signage + taste testing intervention appears unlikely, however, and observers were almost continuously present at the pool and did not observe any changes other than those implemented as part of the intervention. Thus, the most plausible explanation is that the same patrons returned to the pool on a weekly basis and findings represent a true impact of the interventions on their food purchases.
Limited impacts of the interventions are perhaps unsurprising in light of the social ecologic framework, which suggests that health behaviors are shaped by reciprocal interactions among individual, social, and environmental factors. Nudging is a very subtle technique, perhaps too subtle to counter the powerful influence of other environmental factors, such as food marketing, or individual factors such as food preferences or purchasing intentions. Indeed, the impact of nudging on food selection in many studies has been relatively small [7, 15, 17] and inconsistent. Nudges that have proven effective in one context [15, 70, 71] have had null [51, 72], or even opposite impacts in others , and outcomes sometimes differ widely for individual items [15, 70, 71, 74]. Our findings are similar, as nudges that were not effective in the full sample were effective among a subsample of patrons, and their impact differed significantly according to the weight status and sex of the purchaser, suggesting differential sensitivity to specific food environment characteristics. Nudges might be more effective if incorporated within multicomponent interventions, or if carefully matched to the particular circumstances of a target population and setting.
Other explanations for our findings might include the fact that many healthy menu items were similar to the contents of patrons’ home-packed lunches, making them less attractive compared to many unhealthy menu items which could not be brought from home due to temperature restrictions (eg. ice cream cones, grilled cheese sandwiches). Second, given that few children were overweight, parents may not have perceived a need to closely regulate children’s intake of unhealthy items . Third, the interventions may have had limited reach. Although the signage advertising the new names and price reductions on healthy items was colorful and prominent, it may not have captured the attention of consumers in the few seconds typically allocated to food selection in away-from-home settings [51, 75–77], particularly given the excited atmosphere . Similarly, not all patrons participated in taste testing. Fourth, we only promoted the sale of the most healthy items on the menu. Other studies have combined healthy and moderately healthy items into a “healthier” category. Our results may have differed had we also promoted the sale of moderately healthy items, as the taste profiles of these foods are more compatible with consumer taste preferences. Lytle  has suggested that when food access is limited by factors such as low-income, individuals may be more susceptible to influences within the physical food environment. Thus, it is possible that the higher SES of the study sample might also underlie their relative insensitivity to the interventions. Finally, health is only one of many things that individuals value. Children, in particular, have difficulty perceiving the long-term health consequences of dietary choices, and tend to prioritize taste, particularly when eating outside the home [30, 80]. Visits to the pool were a fun family outing and therefore parents may have been more likely to allow indulgences and to accede to children’s food requests [81, 82].
Compared to purchases made by adults alone and/or by children and adults together, when children were alone they purchased more unhealthy items and items with significantly more calories. Children perceive that purchase and preparation of fruits and vegetables are adult tasks . Thus, it may be wise for parents to at least accompany children during food selection. Notably, however, adult choices were not substantially better than the choices made by children alone, a finding also observed by others [80, 83, 84]. In qualitative studies parents admit that they purchase unhealthy foods for their children because other concerns, such as convenience and cost sometimes take precedence over health [85–87]. Indeed, adults may be equally susceptible to environmentally-cued food selection, suggesting that all groups may benefit from increased availability of healthy options in recreational sports settings. It is important that adults select healthier options not only for their children, but also for themselves, to support health, and because parental role modeling significantly influences the dietary behaviors of children [88, 89].
In contrast to industry’s contention that healthy items do not sell in recreational sports settings [24, 25], healthy items were popular among pool patrons and represented 40.8% of items sold. Their share of gross profits was somewhat lower, at 34.1%, as the cost to purchase raw ingredients was higher for healthy foods relative to the profit they generated. Managers can find ways to further minimize food costs, however, as minimizing food costs was not an explicit study goal. In addition, lower profits on healthy items could be offset by increasing the price of unhealthy items [58, 90]. None of the interventions increased overall sales volumes as has been seen in other studies [51, 91, 92], a beneficial finding from a public health perspective, but one that is contrary to the profit motive of industry; however, neither did they adversely affect gross profits, and all were relatively inexpensive to implement and administer. This study also identified a number of non-monetary challenges related to offering healthy items that were encountered by industry. The importance of working with the food industry to improve food environments has been recently highlighted [21, 93] and it will therefore be important to address these barriers to ensure they do not impede much needed improvements to food environments.
Strengths and limitations
Researchers implemented all interventions in conjunction with concession staff and monitored them closely to ensure high fidelity. Thus, null results cannot be attributed to poor execution of interventions. The study was performed in a real-world setting with all of its constraints and supports, increasing the validity of findings. An important strength of this study was that anonymous sales data were augmented by objective measures of food selection in a subsample of patrons for whom selected demographic characteristics were recorded. These strengths are balanced by several limitations, as observer error in this respect may have introduced bias. Our findings related to sex and weight status-specific effects of the interventions apply only to purchases made by adults and children alone, and should therefore be regarded as a preliminary indication of the need for additional study of effect modifiers in this context. We collected observations of patrons who purchased items in the target concession, however these individuals do not necessarily represent those who contributed to the food purchasing decision or those who consumed the items. It is also not clear whether our findings have implications for dietary intake and body weight outcomes. If the changes observed are contextually specific, are not sustained over time, or do not lead to displacement of energy-dense foods in the diet, then these interventions may have little to no real impact. Given that the interventions were additive it was not possible to isolate their individual effects. Moreover, findings may not be generalizable to other settings and populations, or to sales of other healthy foods.
There is no precise, operational definition of nudging . To date, nudging has principally been used in an ad hoc manner and there is a need for a more robust theoretical underpinning to inform development and implementation of interventions . A variety of data will be needed to achieve this outcome. Future studies should compare the relative efficacy of nudges implemented in different populations and settings, alone or in combination, and at multiple decision points such as when selecting a restaurant, at the point of ordering, during meal consumption, and at subsequent meals. The current literature suffers from heterogeneity in study outcomes, intervention sites, types of interventions, participants, outcome measures and types of meals , and it will therefore be important that future studies be designed in a manner that facilitates cross-study comparisons. Studies should also incorporate process measures to assist in understanding why some nudges work in some settings and others do not. Longer-term studies are needed, as the efficacy of nudges implemented in the same manner for the same foods might wane over time.