Abstract
This research examines how price discounts—a classic marketing incentive—drive children's healthy choices in the understudied context of a developing economy. The authors partnered with UNICEF to launch three field experiments in Panamá among 2,418 children to examine four pillars of price discount promotions for children: what to discount (product selection), how to discount (message design), whom to target (children's age), and whether to discount again (repetition). They uncovered four previously undocumented insights. First, price discounts alone effectively increase demand among children 6–11 years of age, reconciling conflicting findings in prior literature. Second, product selection based on relative price—a particularly crucial factor in developing regions—drives opposing postpromotion effects: ironically, marketers should not discount expensive healthy options but rather moderately priced ones. Third, different from prior literature's practice of directly communicating final prices, discount messages that require older children to derive final prices are more effective. Fourth, repetition can amplify or undermine discounts’ efficacy depending on message complexity and children's age. This research offers concrete guidelines for researchers and practitioners, uncovering both positive and negative effects of price promotions on children, and shedding light on price promotion interventions that most powerfully nudge children of different ages to act.
Childhood obesity is one of the most serious global health challenges of the 21st century (Di Cesare et al. 2019); and it is not a problem only for developed countries. Childhood obesity is an increasingly pressing issue in developing countries as well, and the number of overweight children has even tripled in some developing countries in the past three decades (Gregg and Shaw 2017). The ultimate paradox in developing countries is that food insecurity not only causes malnutrition but also can lead to obesity because calories consumed may come from calorie-dense but nutrient-poor foods (Caballero 2005). This paradox occurs because foods high in fat and sugar are often much cheaper than nutrient-dense foods (Drewnowski 2004).
Field interventions aimed to address the grave challenge of childhood obesity have often done so by incentivizing young consumers to eat healthier foods, such as by offering gifts for healthy choices or offering a temporary discount on healthy items (Brown and Tammineni 2009; Gordon, Dynan, and Siegel 2018; List and Samek 2015). When it comes to offering price discounts—a classic marketing incentive—prior explorations have found positive effects on young consumers’ purchases during the promotion period and null effects after the discount promotion ended (Brown and Tammineni 2009; French et al. 1997; Hannan et al. 2002). However, these prior investigations have focused solely on developed countries as test markets, and studied teenagers or tracked the aggregate purchases of elementary school children (6 to 11 years of age) mixed with teenagers and other older consumers. In addition, these prior interventions have relied on multivehicle approaches that combined price discounts with other marketing strategies (e.g., changing the product mix or promotion environment; French et al. 1997; Kocken et al. 2012). This multivehicle and mixed-age approach confounds attempts to gain insight into how discounting healthy options can drive children's choices.
The current research fills this important gap by investigating how price discounts shape children's healthy choices in the unique and understudied context of a developing economy. We conducted three field experiments at three elementary schools in Panamá (totaling 2,418 children) and tracked these children's actual purchases of the promoted healthy products at their schools’ kiosks. To our knowledge, these are the first field experiments in any country to test the impact of price discounts in isolation on children as young as 6 to 11 years of age. The empirical context of a developing economy is equally critical, as developing and developed countries differ on many consequential dimensions, including their economic and health conditions (Koller et al. 2015; Mandava et al. 2012). As a result, many findings on interventions and health in developed countries manifest differently, or not at all, in developing countries (Balsa and Gandelman 2010; Belli 2020; Lavy 2015; Sivaraman, Nye, and Bowes 2019; Yang and Huffman 2013).
Leveraging discount coupons as the promotion vehicles, we homed in on four pillars that scholars and marketers should consider when designing price discount promotions for children: what to discount (product selection), how to discount (message design), whom to target (children's age), and whether to discount again (repetition). Investigating these variables together is imperative because we theorize (and find) that the effects of some of these variables hinge on—and can even reverse as a function of—the others. We captured two aggregate outcomes: purchases of the promoted healthy product during the promotion period and residual purchase effects after the promotion ends. We captured postpromotion purchases because discounts that depress healthy choices after the promotion may ultimately do more aggregate harm than good, whereas those that sustain or increase healthy choices after the promotion have a higher likelihood of yielding meaningful health benefits. In addition to aggregate purchases, we tracked each child's discount coupon redemptions to examine which discount messages were effective, for whom, and when (i.e., discount repetition).
Our investigation reveals four valuable insights: (1) Price discounts in isolation effectively increase demand among children. Our experiments provide the first unconfounded tests isolating the usefulness of price promotions as a marketing tool for motivating children's healthy choices. (2) Product selection based on relative price drives opposing postpromotion effects (i.e., both positive and negative effects) on purchases. This finding extends beyond the null or positive postintervention effects documented in related literature (i.e., discount incentives in Table 1 and nondiscount incentives in Web Appendix A). This finding also sounds a cautionary bell for marketers to select the most suitable products to discount: specifically, not expensive healthy options but moderately priced ones. (3) Different from the prior literature's practice of directly communicating the final price, discount messages that require older children to derive the final price can be even more effective (and the opposite is true for younger children). (4) Repetition can either amplify or undermine discounts’ efficacy, depending on children's age and the discount message's complexity—repeating discounts in most cases does not boost demand. Finally, this collaboration with UNICEF offers concrete directions and action plans for marketers, policy makers, and school administrators, which we delineate in the “General Discussion” section.
Discount-Based Healthy-Eating Field Interventions with Young Consumers.
In the following sections, we first review the incentive-based health intervention literature, discuss our unique approach, and theorize how the four pillars of price discount promotions (product selection, message design, target consumer, and repetition) will shape children's healthy choices. An overview of our field experiments then follows, providing details on the field context, product selection, and intervention design.
Conceptual Framework
Increasing Children's Healthy Choices Through Incentives in the Field
Field interventions conducted in developed economies often offer incentives to motivate young consumers to make healthier choices (e.g., Gordon, Dynan, and Siegel 2018). As many elementary school children in developed countries depend on their parents to pay their food expenses and/or participate in fixed lunch programs at school (School Nutrition Association 2024), a meta-analysis (An 2013) highlighted that most of the field interventions with children, compared with those conducted with older consumers, did not leverage price discounts but instead focused on using nonmonetary gifts (e.g., stickers, pens, and bracelets; Horne et al. 2011; List and Samek 2015) or nonincentive methods (e.g., role modeling; An 2013).
While relatively scant, six investigations (see Table 1) have tried to employ price discounts to shape young consumers’ healthy choices in the field. As illustrated in Table 1, these prior interventions focus solely on developed countries, incorporate additional intervention vehicles while offering discounts, and do not isolate children's purchase decisions. The current research advances this literature by isolating whether discounts alone can alter children's purchase decisions. Doing so is critical because elementary school children undergo two distinct developmental phases (John 1999) that we propose will guide them to react to discount messages in unique ways not yet explored in this intervention literature. Investigating the impact of discounts on children is also vital because unhealthy eating habits take root during childhood (Roberts 2019). Indeed, prior work suggests that interventions that prevent childhood obesity can be more effective than interventions that help adults with obesity later in life (Wofford 2008). Children's (vs. adults’) eating habits are also more malleable (Epstein et al. 1995, 1998; Wilson 1994); potentially as a result, the same healthy-eating interventions can be more effective among children than adults (Epstein et al. 1995; Wilson 1994).
In addition to examining whether price discounts can alter children's healthy choices, the current research also provides novel insights to optimize discount-based interventions by exploring four factors fundamental to their architecture: product selection, message design, children's age, and repetition. Different from prior work's regular practice of directly communicating the discounted final price and launching those discounts in only one promotion period, we investigate whether discount messages that invite children to derive the final price can be even more effective (and among which age groups of children), whether this efficacy evolves when the discount message is repeated, and whether the key product attribute altered by the discount (i.e., its relative price) reverses these effects.
Our empirical context of elementary schools in Panamá offers a unique opportunity to conduct this investigation. Children in many developing countries (e.g., Argentina, Chile, Costa Rica, Guatemala, and Panamá) often carry money and make independent food and beverage purchases daily at their schools’ kiosks (Busse and Diaz 2016; Busse and Piotrowski 2017; Cornejo et al. 2022; Jensen et al. 2021; Prochownik, Vera-Vergara, and Cheskin 2015; Truchi and Navarrete 2016). This field context enables us to distribute discount coupons to children in an understudied economic environment, and track not only their aggregate purchases but also each child's coupon redemption decisions. Doing so further answers the call in this prior literature (French et al. 2001, p. 115) to track food choices at an individual level in order to address a critical constraint highlighted in Table 1: whether any resulting changes in aggregate purchases originate from many young consumers, each purchasing an individual serving, or from a subset of young consumers purchasing large quantities, the latter of which could suggest that an intervention prompts a few individuals to overeat and thus backfires.
In the following sections, we detail our theorizing regarding the impact of discount coupons on three core outcomes: purchases during the promotion, purchases after the promotion, and individual coupon redemptions.
Outcome 1: Purchases During the Promotion
Will price discounts drive children to purchase the promoted product? Some research suggests that they will not. For example, Webley (2005) argues that consumers younger than 12 years of age do not have mental accounts when spending their money (and thus do not budget or effectively save), and that children younger than 9 years of age do not even understand the value of saving money. Research also shows that children often do not include prices when thinking about shopping or searching for purchase-related information (McNeal 1992; Ward, Wackman, and Wartella 1977). Interestingly, other research points to the opposite prediction, that price discounts might drive children's choices, arguing that children as young as age three may be able to understand the concept of money (Gasiorowska et al. 2016) and that children as young as age five can distinguish spending from saving (Smith et al. 2018).
Although children may not regularly pay attention to price or prioritize saving goals (McNeal 1992; Ward, Wackman, and Wartella 1977; Webley 2005), we hypothesize that they nonetheless will respond to the price promotion vehicles in our field experiments—discount coupons. This is because receiving coupons at school can be a novel experience (An 2013), externally directing children's attention to the promoted product and its temporarily lowered price. Indeed, novelty can heighten children's attention to a range of stimuli (e.g., pictures and objects; Bonmassar, Widmann, and Wetzel 2020; Horst et al. 2011; Hutt 1975). In addition, children in developing economies may be particularly likely to attend to price discounts because they are used to bringing money to school and making purchases at school (Busse and Diaz 2016; Jensen et al. 2021; Prochownik, Vera-Vergara, and Cheskin 2015). Accordingly, we propose that once the introduction of novel discount coupons draws children's attention to the temporarily lowered price, they likely will act on the incentive.
Testing this hypothesis is important because extant literature (Table 1) has yet to conduct an unconfounded test of these conflicting predictions due to the use of multivehicle approaches and mixed-age samples.
Outcome 2: Postpromotion Purchases
If discount coupons successfully increase children's healthy choices, will this effect last beyond the promotion period? Investigating this question is valuable because discounts that depress healthy choices after the promotion will ultimately do more aggregate harm than good.
Most prior discount-based healthy-eating interventions on young consumers do not capture postpromotion effects (Table 1). The only exception is French et al. (1997), who tested the impact of price discounts coupled with additional signs and announcements at two high schools and observed a null postpromotion effect. Interestingly, a broader literature on nonmonetary incentives (e.g., using gifts of toys, pens, and stickers to incentivize healthy choices) consistently finds a positive effect among children after a gift incentive is removed (e.g., De Marchi et al. 2020; Laureati, Bergamaschi, and Pagliarini 2014; List and Samek 2015; Loewenstein, Price, and Volpp 2016), with speculations that these incentives fostered liking and social norms. However, the developmentally distinct population (high school students) examined in French et al. (1997), and the different incentives employed in the nonmonetary incentives research, render these findings unfitting predictors of our postpromotion effects. We propose that discount coupons can result in an entirely different postpromotion outcome on children in a developing country, and that this outcome hinges on a core element of architecting price discount interventions: product selection.
Specifically, we theorize that the selected healthy product's original price relative to other options can drive opposing postpromotion effects. Examining this variable in a developing economy offers important insights for two reasons: First, healthy food is often more expensive for consumers in developing (vs. developed) regions; for instance, to purchase food that satisfies the basic nutrition guidelines, consumers in the developing region of Latin America need to pay 3.5 times more as a percentage of their total household expenditure than do North Americans (Bai et al. 2021). Second, healthy foods are pricier than unhealthy foods by a larger margin in developing (vs. developed) regions, and the percentage of people who cannot afford a nutritious diet (relative to a nonnutritious but calorie-sufficient diet) is thus many times greater in Latin America than in the United States (Drewnowski, Gupta, and Darmon 2020; Headey and Alderman 2019; Herforth et al. 2020).
Research conducted among adults indicates that consumers are less likely to purchase healthy foods that are more (vs. less) expensive when they are attuned to price (Darmon and Drewnowski 2015; Kern et al. 2017). Although some prior literature has suggested that children are not regularly attuned to price (McNeal 1992; Ward, Wackman, and Wartella 1977; Webley 2005), we propose that this is not the case in our context because our discount coupons introduce a novel and salient shift in price. As changes can dramatically and immediately grab children's attention (sometimes even more so than adults’ attention; Lempert and Kinsbourne 1983; Mitchell and Smith 2019; Young 1982), we propose that changes in price resulting from the introduction and then removal of discount coupons will draw children's attention to the prevailing pricing gap after the promotion.
If this is true, children's continued attention to price after the promotion ends will have important consequences: If the promoted product is more expensive than most of the other options—as is frequently the case in developing regions like Panamá—this prevailing pricing gap can lead children to become even less likely to purchase the product after the promotion, resulting in a demand deficit that has not been documented in prior incentive-based healthy-eating interventions on young consumers (Table 1 and Web Appendix A). By contrast, if the promoted product is not more expensive than most of the other options, we theorize that children's continued attention to price will prompt them to remain interested in purchasing the healthy product, yielding a positive postpromotion effect. This latter result would provide the first evidence that a positive postpromotion effect can occur among children after a discount incentive ends.
If these hypothesized patterns are observed, then the promoted product needs to be carefully selected to minimize potentially negative postpromotion effects. We further discuss this theorizing's conceptual contributions and practical implications for marketers, policy makers, and school administrators in the “General Discussion” section.
Outcome 3: Individual Coupon Redemptions
As detailed previously, the purchase data enable us to compare the total number of times children purchase the discounted healthy product before, during, and after the promotion, providing direct insight into H1–H2b. The collection of an additional outcome variable—individual coupon redemptions—enables us to examine which discount messages are effective, for whom, and when (i.e., repetition), as detailed subsequently in H3–H4c.
Notably, the use of discount coupons stands in contrast to the methodology employed in prior work (Table 1): except for the study in Cash and McAlister (2017), all of these prior interventions introduced price discounts without leveraging physical coupons. Using physical coupons provides two benefits: (1) we can track each child's purchases instead of relying only on aggregate purchase movements, the latter of which can result from a small subset of children purchasing more rather than a broad shift in behavior (French et al. 2001), and (2) we can alter discount messages while minimizing the possibility of cross-condition contamination.
How to discount: Message design
Prior discount-based healthy-eating interventions on young consumers have used messages that directly communicate the promoted product's final price (Table 1). Advancing this literature, we compare these previously examined direct messages to “derived messages,” which are categorically distinct types of discount messages that we propose would require greater effort for children to comprehend.
Consider coupon messages that require computation (e.g., “$2 off the original price”) or abstract connection making (e.g., “buy A and pay the price of B”) to derive the final price. Whereas consumers who have completed elementary school (i.e., consumers age 12 and older) process these messages relatively easily and quickly (which is why price discounts have long been categorized as a “System 1” rather than “System 2” intervention; Gordon, Dynan, and Siegel 2018), elementary school children will need to exert greater cognitive effort to process these messages to derive the final discounted price. We thus refer to these messages as derived messages. In contrast, and as verified in our pretests and posttests (Web Appendix B), coupon messages that directly specify the final discounted price (e.g., “pay $8” when the original price is $10) require little cognitive effort for children to process. We refer to these messages as direct messages. We hypothesize that derived messages offer a unique opportunity to maximize the effectiveness of price discounts for some (but not all) children, depending on their age.
Whom to target: Children’s age
The distinction between direct and derived discount messages is important because children undergo distinct cognitive developmental phases, and we predict that these phases can reverse the relative efficacy of these two message types. Although children develop basic numeric abilities like counting spontaneously (i.e., without formal instruction) at ages two and three, higher-order conceptualization and operations such as deduction and multiplication (i.e., abilities required to understand derived messages) need to be formally taught and learned (Flavell, Miller, and Miller 2002; John 1999; Wynn 1998). For elementary school children, this learning process spans over six years (John 1999).
Children ages 6 to 7—whom we refer to as “younger children”—tend to rely on one observable dimension, such as size or shape, when making decisions. In contrast, children ages 8 to 11—whom we refer to as “older children”—can consider multiple dimensions and engage in abstract thinking (Boland, Connell, and Erickson 2012; Moses and Baldwin 2005; Ward, Wackman, and Wartella 1977). Older children ages 8 to 11 can also move through a line of reasoning more quickly than can younger children ages 6 to 7 (Hale 1990; Kail 2000). Furthermore, older children ages 8 to 11 tend to use more complex information processing strategies than do younger children ages 6 to 7 (Davidson 1991; Gregan-Paxton and John 1997). As a result of these key distinctions between children ages 6 to 7 and children ages 8 to 11, this age demarcation is the most prevalent classification for children in the prior literature (Bereby-Meyer, Assor, and Katz 2004; Davidson 1991; Gregan-Paxton and John 1997; John 1999; Moses and Baldwin 2005; Ward, Wackman, and Wartella 1977).
This developmental demarcation has important implications for discount messages’ effectiveness. First, we hypothesize that younger children ages 6 to 7 (first and second graders) will likely encounter difficulty processing discount messages that require computation or making abstract connections to derive the final price (e.g., “15 cents less” or “buy A, pay the price of B”). Importantly, such difficulty often prompts disengagement from tasks that require this effortful processing (Hurst and Cordes 2018; John 1999; Patall et al. 2018); hence, we predict that derived (vs. direct) discount messages will be less effective for younger children.
By contrast, older children 8 to 11 years of age (third to sixth graders) are more capable of considering multiple dimensions at once and engaging in abstract thinking (Boland, Connell, and Erickson 2012; Kail 2000; Moses and Baldwin 2005; Ward, Wackman, and Wartella 1977); they can also move through a line of reasoning more quickly than younger children (Kail 2000). Given that older children are more capable of processing both types of discount messages, which message will be more effective?
The extant literature offers competing predictions. The fluency literature would predict that the greater ease of processing simpler information (e.g., direct messages) can make those messages more persuasive, increasing liking of the messages and thus of the promoted products (Chang 2013; Leonhardt, Catlin, and Pirouz 2015; Storme et al. 2015). Such fluency effects can also increase children's preferences and choices (Bernard, Proust, and Clément 2014; Bowker and Sawyers 1988; Büttner, Florack, and Serfas 2014). By contrast, we theorize that the cognitive effort required for children 8 to 11 years of age to process derived messages can have an opposite effect in the context of discount coupons. This theorizing draws on the persuasion literature, which has shown that greater processing of information can heighten that information's persuasive impact, particularly when the information contains strong arguments for a focal proposition (Karmarkar and Tormala 2010; Miniard et al. 1991; Newman and Dolich 1979; Petty and Cacioppo 1979). A novel opportunity to save money, we hypothesize, can constitute a strong argument for purchasing healthy products. Accordingly, we propose that the cognitive effort required for children 8 to 11 years of age to process derived messages can conversely make these messages even more engaging and persuasive than direct messages.
Testing this hypothesis critically expands the discount-based healthy-eating intervention literature (which has thus far relied on direct messages to communicate discounts to young consumers; Table 1). In so doing, we also enrich the health intervention framework that has broadly classified all monetary incentives as System 1 interventions that consumers can easily and quickly comprehend (Gordon, Dynan, and Siegel 2018); our findings suggest that price discounts can be System 2 interventions for some consumers, and that these interventions can be even more effective for specific age groups of children.
Whether to discount again: Repetition
In the real world, discount coupons are rarely used just once; the same promotions are often repeatedly deployed to give consumers multiple opportunities to take advantage of them. In other words, marketers often distribute a series of the same sets of coupons at distinct intervals. However, the literature examining the effect of discounts on young consumers has yet to investigate the role of repetition (Table 1). Testing the effect of repetition not only offers novel insight into the mechanisms through which children's healthy choices can be repeatedly improved, but also is practically beneficial because repeated improvement of healthy choices can fuel habit formation and solidify long-term behavior change (Lally, Chipperfield, and Wardle 2008). In the same vein, if repeating a discount dampens its efficacy, this insight would also provide valuable guidance and illuminate the importance of incentive variety for young consumers.
One prediction is that repeating a message amplifies its persuasive impact. Repetition of marketing messages increases the ease with which they are processed (Nordhielm 2002; Reber, Schwarz, and Winkielman 2004; Shapiro and Nielsen 2013), and this ease increases liking of the messages and the purchase likelihood of the advertised products (Chang 2013; Leonhardt, Catlin, and Pirouz 2015; Storme et al. 2015). Such repetition effects also occur among children (Getz 1963; Taras et al. 1989); for example, children 8 to 10 years of age who saw a dessert advertisement three times (vs. one time) preferred the advertised dessert more (Gorn and Goldberg 1980).
However, novelty plays a bigger role in driving the behavior of children than adults (Montoya et al. 2017). Compared with adults, children attend to novel items for a longer duration (Wetzel, Widmann, and Schröger 2009, 2011) and more often choose, try, and explore novel items (Bradbury and Ross 1990; Sehl et al. 2022; Sumner et al. 2019). Importantly, repetition reduces novelty (Nunes, Ordanini, and Valsesia 2015). When novelty dwindles due to repetition, children's engagement with the discount incentive may decrease. In this light, and in contrast to the classic repetition-induced fluency account, we theorize that for children, the effectiveness of a discount coupon will decline upon repetition.
Yet there may be a unique instance in which the repetition-induced decrease of novelty matters less than the increase of fluency, leading to an overall positive repetition effect. As detailed previously in H3a and H3b, we propose that when first launched, derived discount messages can be uniquely persuasive among older children by evoking greater processing of a strong value proposition; by contrast, these messages’ complexity can prompt disengagement among younger children 6 to 7 years of age. Importantly, repetition aids children's learning (Peracchio 1992; Schwab and Lew-Williams 2016). Since children 6 to 7 years of age can understand concrete observable dimensions and build on them to learn (John 1999), repeated exposure may be what is needed to facilitate their processing of derived messages. Furthermore, repeated exposure increases children's confidence in engaging with the reexposed stimulus (Finn and Tauber 2015), which can also increase their willingness to process an originally complex message. We propose that these processes culminate in a unique exception: the efficacy of derived messages can improve, instead of decline, among younger children 6 to 7 years of age. In sum, we propose that the effectiveness of repetition will vary based on message design and children's age:
Overview of the Current Research
Because the effectiveness of price promotions manifests through actual purchases when consumers have finite resources and face a variety of options, we tested our hypotheses in a real, noisy purchase environment in which children regularly carry cash and purchase food and beverages by themselves. To this end, in collaboration with UNICEF, we conducted three field experiments among elementary school children in Panamá. Next, we briefly discuss Panamá's economic, nutritional, and marketing environment.
Panamá
Panamá is a developing country that lags behind developed countries on numerous dimensions, such as lower gross domestic product per capita (World Bank 2020), lower education quality (World Economic Forum 2018c), and lower buyer sophistication (World Economic Forum 2018a). Panamá also lags behind developed countries on critical health dimensions. For example, it has a higher stunting prevalence, a higher rate of infant mortality, and a higher rate of death from nutrition conditions and communicable diseases (World Bank 2020; World Economic Forum 2018b). Also in contrast to children in developed regions, Panamanian children's average height has not increased in recent years, which indicates that nutrition conditions in Panamá have not improved (Berlinski and Schady 2015). A core problem underpinning these health challenges is that the nutrient-poor foods that many Panamanian children regularly consume are also calorie-dense, which fuels rapidly rising obesity rates (30% and rising; TeleSUR 2019). For instance, studies suggest that 52% of Panamanian children eat fast food weekly, and 81% drink sugary drinks five or more times per week (Kakkad et al. 2020). Panamanian children consume on average 511.2 grams of sugar weekly, which is approximately three times the recommended limit (American Heart Association 2016).
Since there is no validated database in Panamá on food marketing that leverages price promotions, we surveyed 203 Panamanian and U.S. parents and found that Panamanian (vs. U.S.) parents see significantly fewer discount coupons in stores, online, in schools, and on television (Web Appendix C). Additional observational data collected by research assistants suggest that in the limited instances in which price discounts and coupons are used in Panamá, they predominantly promote unhealthy (vs. healthy) options (Web Appendices D1 and D2). This food marketing environment in Panamá is fueling rapidly rising obesity rates: the rates of obesity in Panamá are projected to outpace those in the majority of other developing countries in Latin America by 2050 (Webber et al. 2012). We examined our hypotheses in this developing economy to shed light on potential solutions and their constraints.
School Selection
With help from UNICEF, we approached the authorities of the Ministry of Education of the Republic of Panamá to recruit elementary schools. At the time of the study, there were 3,260 elementary schools enrolling approximately 418,000 students. We focused our consideration set on schools that had a preexisting kiosk that sold healthy options which met UNICEF's nutritional standards (Web Appendix E), and where children regularly purchased food and drinks at the kiosk during school hours. Furthermore, to ensure proper sample size, we limited the final consideration set to schools enrolling at least approximately 100 students in each grade. UNICEF reached out to 12 schools that met the criteria and were in contact with the organization; we obtained agreement from three schools.
These three schools are in urban areas (with Gran Bretaña and La Siesta in the district of Panamá, and Estado de Israel in the district of San Miguelito) and enroll students in grades 1 through 6 (Web Appendix F). These schools are in districts that have a representative poverty rate (the poverty rate in Panamá is 12.6%, and the poverty rates in the districts of Panamá and San Miguelito are 13.0% and 12.7% [Ministerio de Economía y Finanzas 2015; World Bank 2020]) and a representative gender distribution (48% of students in Panamá are female, and 49% and 48% of students in the districts of Panamá and San Miguelito are female [Instituto Nacional de Estadística y Censo 2021]).
Product Selection
Along with the UNICEF nutrition team, we interviewed the local research groups and teachers at these schools to select one healthy product to promote at each school. The selection criteria included having been certified by UNICEF as a healthy option for the target population (Web Appendix E) and being stocked at the kiosks. Field Experiments 1 and 2 investigated the common instances in Panamá in which healthier products are pricier than less healthy products; we promoted two different products in these two schools to ensure generalizability. Field Experiment 3 investigated the proposed postpromotion reversal (H2b) that occurs when the promoted healthy product is not pricier than most other options in the set; to hold constant the healthy product with a product promoted in the previous experiments, Field Experiment 3 promoted a product that was similar to that promoted in Field Experiment 2. The final products selected were a healthy cereal shake in Field Experiment 1 and a healthy natural fruit beverage in Field Experiments 2 and 3 (see Table 2). These products were freshly made every day using local ingredients. The list of all products at each school is in Web Appendix G.
Overview of the Experimental Methodology.
Design of Discount Messages
Each field experiment tested two price discount messages: a direct message and a derived message. In all experiments, the direct messages directly specified the final price. Across experiments, we tested two types of derived messages to ensure generalizability: (1) a derived message that required children to deduct the discount amount from the original price to derive a final price (e.g., “15 cents less”) and (2) a derived message that required children to make connections between products to derive a final price (i.e., “buy A, pay the price of B”; see Table 2). A pretest and posttest found converging evidence that these derived (vs. direct) messages required greater cognitive effort for Panamanian children to process (Web Appendix B).
We hired an illustrator at Stanford University to design the visual for these coupons. Each coupon was 3.5 by 2 inches in size (close to a typical business card, for ease of carrying in a pocket or wallet). To keep the visual impact constant, we used the same color and design for the direct and derived messages at each school. For quality control, we produced the coupons in the United States and then shipped the coupons in clearly labeled individual packages (by school and by class) to Panamá.
Design of Promotion and Purchase-Tracking Periods
In all schools we tracked purchases across five periods: (1) baseline period: approximately a month before the discount coupons were first distributed; (2) Phase 1: the three days during which the first set of coupons could be redeemed; (3) Rest 1: approximately three weeks between the first and second coupon promotions; (4) Phase 2: the three days during which the second set of coupons could be redeemed; and (5) Rest 2: approximately three weeks after the second coupon promotion. These promotion and purchase-tracking periods refer exclusively to school days (and not weekends), during which children were at school and thus had the opportunity to purchase items from the kiosk.
These durations are the longest that the schools’ administrations allowed and are within the ranges tested in prior literature. For instance, incentive-based healthy-eating interventions on young consumers have examined postintervention periods lasting as short as one day (Jones, Madden, and Wengreen 2014), two days (Lai, List, and Samek 2020; List and Samek 2017), three days (Grubliauskiene, Verhoeven, and Dewitte 2012; List and Samek 2015), seven days (Toossi 2017), and as long as three weeks (French et al. 1997) to months (Loewenstein, Price, and Volpp 2016).
Panamá's Institutional Review Board and Government Approval
We sent a formal letter and presented our plans to the authorities of the Ministry of Education of the Republic of Panamá. After obtaining their permission, we sent an official letter to these three schools’ principals detailing the field experiments’ procedures. The research team trained the school kiosk managers to (1) track purchases of the promoted products using prepared sheets and (2) ensure sufficient inventory of the promoted products for the full duration of the experiment. The research protocols for all three schools were submitted to the National Secretariat of Science, Technology, and Innovation in Panamá (SENACYT), which manages logistics for its parent organization, the National Research Bioethics Committee of Panamá (CNBI). The approval was granted by a member of CNBI that specializes in children's health (i.e., Panamá's Hospital del Niño). The recruitment, negotiation, preparation, and approval process from all parties took two years.
Field Experiment 1: Leveraging Direct and Derived Discount Messages to Promote a Healthy Product
The School
The first school was Gran Bretaña. The average monthly income of parents at this school was U.S. $350, with 10% formally working, 40% informally working, and 50% unemployed. The gender distribution of children was 47% boys and 53% girls. Sixty students (i.e., 11% of students) had disabilities (50% boys and 50% girls). The ethnic breakdown of the children was 20% Guna and 80% Emberá.
Coupon Stimuli
In collaboration with UNICEF and the school, we selected a healthy cereal shake to promote. This product was more expensive than approximately two-thirds (i.e., 66.67%) of all other products at the school's kiosk (Web Appendix G).
We manipulated whether the coupons required effortful processing by changing whether the coupons directly communicated the final price or required children to make a connection between two numbers to derive the final price. Specifically, the derived coupon read “A cereal shake usually costs 1 dollar.
Coupons for Field Experiment 1.
Method
Because it was crucial for our experiment that children could not easily compare their coupons across conditions, we assigned one type of discount message to each class, and randomly assigned the two types of messages across classes within each grade (Table 3). Thus, about half of the first grade classes received the direct message while the other half received the derived message, half of the second grade classes received the direct message while the other half received the derived message, and so on. Since classes within each grade are created based only on numbers (i.e., to enroll about the same number of children per class) and not on any other variables (e.g., academic performance), the randomization across classes ensured the experiment's internal validity while minimizing the possibility of cross-condition comparison.
Assignment of Classrooms to Coupons in Field Experiment 1.
Each coupon was marked with a serial number to indicate the grade and class it was distributed to, followed by a unique count number, such that the coupons could not be duplicated or shared with other children. Research assistants, who were unaware of the hypotheses, distributed the coupons on a Tuesday morning. The research assistants were accompanied by the teacher of the class to ensure credibility. The teachers were not involved in the communication regarding the coupons and were informed not to discuss these coupons in their class; in addition, teachers were not aware of the different discount messages between conditions. The teachers were simply present to reduce children's anxiety or confusion about having a stranger in their class.
All coupons in all classes were distributed on the same morning to keep the starting time consistent across classes and grades. Each student received one coupon on that Tuesday morning. Research assistants visited the school every day that week to validate that there was no issue with redeeming the coupons and that the kiosk manager accurately tracked purchases of the focal product. As previously noted, children in many developing regions—including Panamá—regularly carry a small amount of cash to purchase food and beverages at school kiosks during their multiple school breaks by themselves, autonomously and unsupervised by adults. If the children in our field experiment wished to use the coupon, they had until Thursday of that week to redeem it. The promotion period was specified on the coupon.
A month after the distribution of the first-phase coupons, on a Tuesday morning, the research assistants visited the school again to distribute the second-phase coupons. All children who were present received a coupon (see Web Appendix H1 for further details). As in the first phase, the children could redeem these coupons until Thursday of that week. These second-phase coupons were identical to the ones distributed in the first phase, enabling us to examine the effect of repetition. To ensure that all coupons redeemed in the second phase were the ones freshly distributed (and not unused coupons from the first phase), we created different serial numbers for second-phase coupons, and alerted the kiosk managers to pay attention to the serial number when redeeming coupons. All reported redemptions were cross-checked with the coupons collected at each kiosk to ensure the accuracy of the dependent measure. The school kiosks were the only place to purchase food and beverages in each experiment.
Results and Discussion
Immediate increase in purchases (H1)
A Poisson regression of the five periods (baseline vs. Phase 1 vs. Rest 1 vs. Phase 2 vs. Rest 2) on daily purchases during each of the 61 school days of data collection revealed a significant effect (χ2(4) = 53.03, p < .001; Figure 2). Planned contrasts found that purchases increased from 23.75 (SD = 7.13) shakes per day at baseline to 32.50 (SD = 3.54) shakes per day during the first promotion (β = .31, χ2(1) = 5.76, p = .017). The first promotion thus significantly increased Panamanian children's purchase of the promoted healthy product. We observed a similar, though nonsignificant, spike when the coupons were introduced a second time. The average daily purchases between the two promotions were 20.00 (SD = 5.83) shakes per day, compared with the (nonsignificantly) higher purchases of 24.33 (SD = 4.51) shakes per day during the second promotion period (β = .20, χ2(1) = 2.28, p = .131). Figure 2 details the number of school days in each phase, and Web Appendix H1 reports additional methodological details (e.g., the number of coupons not distributed in each phase).
Average Daily Purchases (Field Experiment 1).
Postpromotion decrease in purchases (H2a)
However, purchases declined after the promotion. Further planned contrasts revealed that after the first promotion, purchases declined from 32.50 (SD = 3.53) shakes per day to 20.00 (SD = 5.83) shakes per day (β = −.49, χ2(1) = 12.74, p < .001). There was also a decline after the second promotion, such that purchases declined from 24.33 (SD = 4.51) shakes per day to 14.88 (SD = 4.51) shakes per day (β = −.49, χ2(1) = 13.54, p < .001). Of note, there were fewer purchases after the first promotion than at baseline (20.00 vs. 23.75; β = −.17, χ2(1) = 6.05, p = .014), and there were also fewer purchases after the second promotion than at baseline (14.88 vs. 23.75; β = −.47, χ2(1) = 36.72, p < .001).
Discount message (H3a and H3b)
To cleanly examine the immediate impact of the discount message on children of different age groups without repetition effects, we next conducted a logistic regression with message (direct vs. derived), age (younger vs. older), and their interaction as the independent variables, and redemption rate in Phase 1 as the dependent variable; we explore repetition (and the three-way interaction with repetition) in the next section of results. As detailed in the introduction, we followed the procedures in prior literature (e.g., Bereby-Meyer, Assor, and Katz 2004; Davidson 1991; Ginsburg and Opper 1988; John 1999) to collapse grades 1 and 2 into the younger age group, as children in these grades are in the same early developmental stages (ages 6–7), and grades 3–6 into the older age group, as children in these grades are in the same later developmental stages (ages 8–11).
The logistic regression revealed a main effect of message (Wald χ2 = 5.19, p = .023) and age (Wald χ2 = 8.89, p = .003), qualified by the hypothesized message × age interaction (Wald χ2 = 17.96, p < .001; Figure 3). Simple effect analyses revealed that when the coupons were first distributed, the derived message led to a higher redemption rate than the direct message among older children (19.1% vs. 10.5%; χ2 = 5.32, p = .021), whereas the reverse occurred among younger children (direct = 26.0% vs. derived = 5.1%; χ2 = 15.56, p = .001). Hence, the derived (vs. direct) message was more effective among older children, whereas the direct (vs. derived) message was more effective among younger children.
Coupon Redemption (Field Experiment 1).
Repetition (H4a–H4c)
We then explored the effect of repetition by conducting a 2 (Phase: 1 vs. 2) × 2 (message: direct vs. derived) × 2 (age: younger vs. older) logistic regression on redemption rate. As predicted, we observed a three-way interaction between message, age, and phase (Wald χ2 = 17.19, p < .001; Figure 3); to conserve the space needed to detail the predicted simple effects resulting from this hypothesized interaction, any effects outside of this interaction are detailed in Web Appendix H2 for comprehensiveness. We decomposed the three-way interaction by age to test how repetition of the different messages affected children of different age groups.
Among older children, we observed a main effect of message (Wald χ2 = 5.19, p = .023) and a main effect of phase (Wald χ2 = 6.35, p = .012); there was no interaction. Overall, older children lost interest in both types of coupons from Phase 1 to Phase 2 (14.4% vs. 8.1%; χ2 = 7.11, p = .008). Among younger children, we found a main effect of phase (Wald χ2 = 8.66, p = .003) and message (Wald χ2 = 12.79, p < .001), qualified by an interaction between phase and message (Wald χ2 = 19.70, p < .001). Simple effect analyses revealed that for younger children, the direct message was less effective in Phase 2 than in Phase 1 (5.7% vs. 26.0%; χ2 = 13.15, p < .001). Conversely and as hypothesized, the derived message was more effective in Phase 2 than in Phase 1 (20.0% vs. 5.1%; χ2 = 9.99, p = .002). Put together, repetition reduced the effectiveness of all discount messages except for the derived message among younger Panamanian children.
In sum, Field Experiment 1 found that discount coupons were immediately effective (H1) but can also prompt a negative postpromotion effect (H2a). In addition, the derived (vs. direct) message was more effective among older children (H3b), while the reverse occurred among younger children (H3a). Notably, repetition reduced children's interest in using the coupons (H4a and H4b), except for conversely increasing the effectiveness of derived discount messages among younger children (H4c).
Field Experiment 2: Launching a Different Derived Discount Message to Promote a Different Healthy Product
In Field Experiment 2, we further triangulated the effect of direct versus derived discount messages by manipulating this construct in a different way: instead of altering the need for computation to derive the final price, we designed a derived message that required children to make connections between products of different sizes to derive the final price. The pretest and posttest (Web Appendix B) suggested that this message similarly required children to devote more cognitive effort to process. Testing this manipulation in Field Experiment 2 enhanced the generalizability of our findings in Field Experiment 1 and provided convergent evidence for the proposed theory of direct versus derived messages. It also ruled out potential alternative accounts that might result from employing a math-based message or differentially highlighting the discount amount versus the product price. Also of note, the direct and derived messages in this experiment specified the same two reference points: both the promoted product's original price (65 cents) and the discounted price (40 cents).
The School
To ensure generalizability, Field Experiment 2 tested our hypotheses at a different elementary school, Estado de Israel. The average monthly income of parents at this school was U.S. $600, with 60% formally working and 40% unemployed. The ethnic breakdown of children was approximately evenly distributed among Guna, Emberá, and Ngobe.
Coupon Stimuli
To further ensure generalizability, we worked with UNICEF and the school to select a different healthy product to promote—a healthy fruit beverage. This product was again more expensive than approximately two-thirds (i.e., 67.57%) of all other products at the school's kiosk (Web Appendix G), enabling us to document converging evidence for H2a.
We manipulated message type by changing whether the children had to make connections between products of different sizes to derive a final price. As in Field Experiment 1, the direct message directly specified the final price (40 cents; see Figure 4). The derived message, in contrast, required children to make a connection between the healthy product's large and small sizes to derive its final price, and read “A large natural fruit beverage usually costs 65 cents and a small natural fruit beverage usually costs 40 cents.
Coupons for Field Experiment 2.
Assignment of Classrooms to Coupons in Field Experiment 2.
Results and Discussion
Immediate increase in purchases (H1)
Following the same analysis procedures as in Field Experiment 1, a Poisson regression of the five periods (baseline vs. Phase 1 vs. Rest 1 vs. Phase 2 vs. Rest 2; see Web Appendix H for further details) on daily purchases during each of the 51 school days of data collection revealed a significant effect (χ2(4) = 262.95, p < .001; Figure 5). Planned contrasts again showed that average daily purchases significantly increased from 4.00 (SD = 2.00) cups per day at baseline to 33.67 (SD = 13.50) cups per day during the first promotion period (β = 2.13, χ2(1) = 223.20, p < .001). We observed a similar spike in purchases when the coupons were introduced a second time; the average daily purchases increased from 2.33 (SD = 2.79) cups per day between the two promotions to 8.00 (SD = 7.00) cups per day during the second promotion period (β = 1.23, χ2(1) = 21.62, p < .001).
Average Daily Purchases (Field Experiment 2).
Postpromotion decrease in purchases (H2a)
As in Field Experiment 1, purchases declined after the promotion. Planned contrasts revealed that after the first promotion, purchases declined from 33.67 (SD = 13.50) cups per day to 2.33 (SD = 2.79) cups per day (β = −2.67, χ2(1) = 185.23, p < .001). Purchases also declined after the second promotion from 8.00 (SD = 7.00) to 2.00 (SD = 1.09) cups per day (β = −1.39, χ2(1) = 15.37, p < .001). Of note, there were fewer purchases after the first promotion than at baseline (2.33 vs. 4.00; β = −.54, χ2(1) = 7.45, p = .006), and there were fewer purchases after the second promotion than at baseline (2.00 vs. 4.00; β = −.69, χ2(1) = 5.11, p = .024).
Discount message (H3a and H3b)
The 2 (message: direct vs. derived) × 2 (age: younger vs. older) logistic regression on redemption rate in Phase 1 revealed a main effect of message (Wald χ2 = 36.07, p < .001), and a main effect of age (Wald χ2 = 24.04, p < .001), qualified again by a message × age interaction (Wald χ2 = 24.19, p < .001; Figure 6). Simple effect analyses revealed that when the coupons were first distributed, the derived message led to a higher redemption rate than the direct message among older children (40.9% vs. 13.5%; χ2 = 39.13, p < .001), whereas the reverse again occurred among younger children (direct = 19.4% vs. derived = 3.4%; χ2 = 11.04, p = .001).
Coupon Redemption (Field Experiment 2).
Repetition (H4a–H4c)
The 2 (Phase: 1 vs. 2) × 2 (message: direct vs. derived) × 2 (age: younger vs. older) logistic regression on redemption rates again yielded a three-way interaction (Wald χ2 = 31.46, p < .001; Figure 6). Decomposition of the three-way interaction by age replicated the results in Field Experiment 1. Among older children, we again observed a main effect of phase (Wald χ2 = 39.14, p < .001) and a main effect of message (Wald χ2 = 36.07, p < .001); we also observed a two-way interaction (Wald χ2 = 17.53, p < .001). Simple effect analyses revealed that among older children, the direct message again was less effective in Phase 2 than in Phase 1 (6.5% vs. 13.5%; χ2 = 5.36, p = .021), and this decrease in effectiveness was even more drastic for the derived message (1.6% vs. 40.9%; χ2 = 87.66, p < .001).
Among younger children, we found a main effect of phase (Wald χ2 = 9.13, p = .003) and message (Wald χ2 = 8.76, p < .003), qualified by the predicted two-way interaction between message and phase (Wald χ2 = 14.36, p < .001). Simple effect analyses revealed that for younger children, the direct message was less effective in Phase 2 than in Phase 1 (7.5% vs. 19.4%; χ2 = 5.59, p = .018). Also as in the prior experiment, the derived message was more effective in Phase 2 than in Phase 1 (20.0% vs. 3.4%; χ2 = 11.59, p = .001). Thus, repetition again reduced the efficacy of all messages except for the derived message among younger children.
Field Experiment 2 provided robustness checks for the results in Field Experiment 1 by replicating the same patterns in the promotion of a different healthy product at a different school. Field Experiment 2 also triangulated the effect of direct versus derived messages by employing a different operationalization to manipulate this construct. In so doing, it ruled out alternative accounts that may result from specifying the discount amount or requiring computation.
Field Experiments 1 and 2 also documented a troubling pattern: price discounts on healthy products produced a unique postpromotion decline in purchases of those products (H2a). This pattern has not been documented in any prior discount incentive or nondiscount incentive healthy-eating interventions on young consumers (Table 1 and Web Appendix A). Investing in price discounts to promote healthy options among children seems to ultimately produce less good than doing nothing at all. Field Experiment 3 explores a critical boundary of this postpromotion backlash by carefully selecting an inexpensive healthy product to promote (H2b). We propose that doing so can fuel a unique postpromotion increase in purchases.
Field Experiment 3: Reversing the Postpromotion Decrease in Purchases
In Field Experiment 3, we promoted a healthy product that was relatively inexpensive (i.e., the promoted product was more expensive than only 27.3% of all the products in the school's kiosk). The procedure and design of this experiment were otherwise identical to those in Field Experiments 1 and 2 to provide further robustness checks for H1, H3, and H4.
The School
The third school was La Siesta. Although the average income was not tracked at this school, about 50% of the parents were formally employed and 50% were unemployed. The ethnic breakdown of children was 80% nationals, most of them Guna, and 20% foreigners.
Coupon Stimuli
In this field experiment, we promoted the same type of product—a healthy fruit beverage—as in Field Experiment 2. Different from the first two experiments, however, the promoted product was not among the most expensive products sold at the school kiosk. In Field Experiment 3, the promoted product was more expensive than only 27.3% of all the products in the school kiosk, compared with 66.67% in Field Experiment 1 (χ2 = 11.87, p < .001) and 67.57% in Field Experiment 2 (χ2 = 13.15, p < .001). We used a similar manipulation of message type as in Field Experiment 1, altering whether the coupons required children to deduct the discount amount from the original price to derive the final price (Figure 7). All coupon distribution and purchase recording procedures followed the first two field experiments. See Table 5 for assignment of classrooms to coupons in this experiment.
Coupons for Field Experiment 3.
Assignment of Classrooms to Coupons in Field Experiment 3.
Results and Discussion
Immediate increase in purchases (H1)
Consistent with the prior experiments, the distribution of coupons again immediately increased purchases of the promoted healthy fruit beverage. A Poisson regression of the five periods on daily purchases during each of the 55 school days of data collection revealed a significant effect (χ2(4) = 479.95, p < .001; Figure 8). Planned contrasts revealed that the average daily purchases increased from .27 (SD = .88) cups per day at baseline to 34.33 (SD = 12.34) cups per day during the first promotion period (β = 4.84, χ2(1) = 132.48, p < .001). The second coupon distribution again boosted purchases, increasing from 4.93 (SD = 10.58) cups per day between the two promotions to 18.00 (SD = 12.12) cups per day during the second promotion period (β = 1.29, χ2(1) = 52.27, p < .001).
Average Daily Purchases (Field Experiment 3).
Postpromotion increase in purchases (H2b)
Importantly, purchases increased after the promotion. Although the same decrease in purchases occurred right after the first promotion (from 34.33 [SD = 12.34] to 4.93 [SD = 10.58] cups per day; β = −1.94, χ2(1) = 162.05, p < .001) and the second promotion (from 18.00 [SD = 12.12] to 2.08 [SD = 1.97] cups per day; β = −2.16, χ2(1) = 79.39, p < .001), we observed the hypothesized boost in purchases relative to the baseline (H2b). Unlike in Field Experiments 1 and 2, there were more purchases after the first promotion than at baseline (4.93 vs. .27; β = 2.89, χ2(1) = 46.51, p < .001); similarly, there were more purchases after the second promotion than at baseline (2.08 vs. .27; β = 2.03, χ2(1) = 19.98, p < .001). Hence, coupons promoting the less-expensive healthy fruit beverage led to a postpromotion increase in purchases (H2b; Figure 8).
Discount message (H3a and H3b)
The 2 (message: direct vs. derived) × 2 (age: younger vs. older) logistic regression on redemption rate in Phase 1 revealed a main effect of message (Wald χ2 = 6.21, p = .013) and age (Wald χ2 = 4.47, p = .034), qualified again by a message × age interaction (Wald χ2 = 10.08, p = .002; Figure 9). Replicating the prior field experiments, simple effect analyses showed that when the coupons were first distributed, the derived message led to a higher redemption rate than the direct message among older children (14.7% vs. 9.0%; χ2 = 6.32, p = .012), whereas the reverse occurred among younger children (direct = 16.3% vs. derived = 7.5%; χ2 = 5.21, p = .022).
Coupon Redemption (Field Experiment 3).
Repetition (H4a–H4c)
The 2 (Phase: 1 vs. 2) × 2 (message: direct vs. derived) × 2 (age: younger vs. older) logistic regression on redemption rates again yielded a three-way interaction (Wald χ2 = 25.89, p < .001; Figure 9). Among older children, we observed a main effect of message (Wald χ2 = 6.21, p = .013) and a main effect of phase (Wald χ2 = 18.07, p < .001), qualified by a two-way interaction between message and phase (Wald χ2 = 4.15, p = .042). Simple effect analyses revealed that the derived message was less effective in Phase 2 than in Phase 1 (5.4% vs. 14.7%; χ2 = 19.49, p < .001), whereas the direct message's efficacy did not change significantly (6.6% vs. 9.0%; p = .230).
Once again, among younger children, we found main effects of phase (Wald χ2 = 12.93, p < .001) and message (Wald χ2 = 4.98, p = .026), qualified by a two-way interaction between message and phase (Wald χ2 = 22.53, p < .001). Each simple effect replicated both prior experiments: for younger children, the direct message was less effective in Phase 2 than in Phase 1 (5.0% vs. 16.3%; χ2 = 10.91, p < .001) whereas the derived message was more effective in Phase 2 than in Phase 1 (25.5% vs. 7.5%; χ2 = 14.37, p < .001). Put together, repetition only increased the effectiveness of the derived message among younger Panamanian children.
In sum, during the two phases in which the coupons were distributed, redemption of the coupons followed the same patterns as in Field Experiments 1 and 2. Further consistent with our theorizing, it was only when the promotion ended—and thus when the cost of the focal healthy product returned to its original price—that a different pattern emerged in Field Experiment 3. These findings are consistent with the theorizing that when a promoted healthy product's original price is relatively inexpensive (vs. expensive) compared with other options, the removal of that price discount produces a postpromotion increase (vs. decrease) in demand among children.
General Discussion
Three field experiments in Panamá, each with two rounds of price promotions, real coupon redemptions, and weeks of purchase tracking, uncovered four core insights: (1) price discounts promoting healthy options (alone) can increase immediate demand among consumers as young as elementary school children; (2) the original relative price of the promoted product—a particularly critical factor to consider in developing regions like Panamá—fuels opposing postpromotion effects of price discounts; (3) derived (vs. direct) discount messages offer a unique opportunity to increase older children's healthy choices; and (4) repeating derived discount messages uniquely increases younger children's healthy choices.
Theoretical and Substantive Takeaways
This research contributes to the price promotion literature by providing unconfounded insight into whether price discounts alone—delivered through physical coupons—can boost the healthy food choices of children as young as 6 to 11 years of age. Our field data provide important evidence that price discounts can indeed be a useful short-term marketing incentive for children. In addition, by tracking individual coupon redemption in addition to aggregate purchases, we uncovered valuable insights regarding how to best design price promotions for children. Because unhealthy eating habits often take root in childhood (Roberts 2019), shaping children's behaviors in an environment in which they make independent food choices can be consequential.
This research is also the first to observe opposing postpromotion effects in children's healthy choices in response to price discounts. These findings provide valuable guidance for selecting healthy options to promote to children. As prior discount interventions on young consumers have investigated only developed countries in which the pricing gap between healthy and unhealthy options is smaller (Headey and Alderman 2019), our research sheds light on a hidden factor that can be especially imperative for designing discount interventions in developing countries. These findings thus enrich the dialogue on how to promote children's health beyond the borders of developed economies, and respond to the call-to-action to increase behavioral science's sampling diversity (Henrich, Heine, and Norenzayan 2010). Also important, prior research has argued that because price is less salient to children than other features (e.g., packaging), children often do not take price into consideration (McNeal 1992; Ward, Wackman, and Wartella 1977). Our findings expand this literature by documenting situations in which price can become a salient feature that drives children's choices (i.e., when a coupon is introduced and then removed).
Our research also advances the understanding of the relative efficacy of direct versus derived messages. Because derived messages prompt effortful processing (Web Appendix B)—in other words, System 2 processing (Kahneman 2011)—this insight expands our understanding of what classically constitutes System 1 versus System 2 interventions: Whereas prior healthy food discounts for young consumers featured direct messages (categorized as System 1 messages given the ease with which they can be processed; Gordon, Dynan, and Siegel 2018), our research suggests a new dimension that can be at play—and can be even more effective—when shaping children's healthy choices. Notably, repeated exposure can reverse the initial inefficacy of derived discount messages among younger children. These findings are consistent with our theorizing that these children's different developmental stages guide them to react to discount messages in ways undocumented in the healthy-eating intervention literature.
All of these findings have invaluable implications for policy makers, marketers, and school administrators. We note several implications in Table 6, and we believe that further research will be instrumental in validating these ideas and recommendations, amplifying their impact, and delineating meaningful boundaries.
Implications for Policy Makers, Marketing Practitioners, and School Administrators.
Notes: Many of these suggestions are either incorporated into UNICEF’s recommendations to stakeholders or are under consideration to be included in UNICEF’s upcoming recommendations to stakeholders.
Limitations and Future Research
The current research examined price promotions by distributing three-day coupons to elementary school children to promote healthy options. We encourage future research to examine how shifting these parameters alters the impact of price discounts on healthy choices. For example, whereas children in the current field experiments had a few days to redeem their coupons, a longer redemption period may alter coupons’ impact as urgency may fade over time. Another question is whether the observed immediate boost in purchases generalizes to less tasty healthy options. Furthermore, if the immediate boost in purchases also generalizes to unhealthy options—and if policy makers do not limit use of the current intervention to the promotion of healthy options—manufacturers of unhealthy foods could leverage this intervention to harm children's health.
The healthy products promoted in Field Experiments 1 and 2 were pricier than most products in the set, a price difference that is common in developing regions such as Panamá (Drewnowski, Gupta, and Darmon 2020). To demonstrate robustness within this common price context, Field Experiments 1 and 2 promoted products with different absolute baseline prices ($1 and $.65, respectively, both of which were pricier than two-thirds of the other options in the set). By contrast, Field Experiment 3 promoted a $.50 product that was not pricier than most other options in the set (i.e., it was pricier than only 27.3% of the other options). We encourage future research to further test the generalizability of the current findings across other absolute and relative price points that may render an objectively expensive product relatively inexpensive, or vice versa.
Another factor to consider is the number of substitute products. Field Experiments 1, 2, and 3 had 13, 10, and 9 substitute products, respectively. Because Field Experiments 1 and 2 revealed an identical postpromotion decrease in purchases despite a difference of three substitutes, it is unlikely that a difference of just one substitute (e.g., in Field Experiments 2 and 3, where we see the postpromotion reversal) is sufficient to reverse the results. However, future research can test how bigger changes in set sizes may affect perceptions of healthy products’ prices; in situations where the set size is much larger, the current findings may be more prominent (e.g., a discount may be one of the few comparative attributes among numerous products) or may be weaker (e.g., the discount's efficacy may be diluted by other brands’ marketing efforts).
Also of note, no prior research except for that of French et al. (1997) has examined the effect of discount-based interventions on young consumers’ purchases after the promotion. We examined postpromotion periods of similar duration to those examined by French et al. The postpromotion durations in our research are also within the range tested in prior incentive-based healthy-eating literature, and longer than several influential field interventions on children's healthy eating (e.g., Grubliauskiene, Verhoeven, and Dewitte 2012; Jones, Madden, and Wengreen 2014; Lai, List, and Samek 2020; List and Samek 2015, 2017). Given that novelty fades even more over a longer period, yet robust habits can be formed with repetition over a longer period (especially during early childhood; Albuquerque et al. 2018; Connell, Brucks, and Nielsen 2014), we encourage future research to study the effects of discounts on children across varying durations.
Although the three field experiments provided robust and converging evidence, we were unable to collect additional data beyond purchases and coupon redemptions. Future research could benefit from collecting additional data that could offer further insight into the underlying mechanism, such as data connecting each child's traits to their purchase behavior. For example, children's gender, household norms, socioeconomic status, liking of the promoted product, experience with the promoted product, and price sensitivity could affect their responses to price promotions as well as to the removal of price promotions. In addition, just like any research that recruits subjects from specific pools, the generalizability of the present findings across different cultures requires further testing. We enthusiastically encourage future research to explore how children around the world may respond to different discount messages promoting healthy options.
In sum, our findings offer theoretical insight for devising practical strategies to encourage children to choose healthy products. We see this work as a starting point to examine how price discounts—a classic marketing incentive—can increase young consumers’ healthy choices in a developing economy. Our findings suggest that harnessing these incentives can powerfully increase children's healthy choices, but a careful architecture of these incentives (e.g., considering product selection, message design, children's age, and repetition) is key. We hope our initiative paves the way for more research that can inform the development (and ethical trade-offs) of discount policies and practices on young, vulnerable consumers around the world.
Supplemental Material
sj-pdf-1-mrj-10.1177_00222437241237483 - Supplemental material for Using Price Promotions to Drive Children's Healthy Choices in a Developing Economy
Supplemental material, sj-pdf-1-mrj-10.1177_00222437241237483 for Using Price Promotions to Drive Children's Healthy Choices in a Developing Economy by Szu-chi Huang, Michal Maimaran and Daniella Kupor in Journal of Marketing Research
Footnotes
Authors Contributions
The first two authors share equal authorship.
Coeditor
Rebecca Hamilton
Associate Editor
Nailya Ordabayeva
Declaration of Conflicting Interests
The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The authors received no financial support for the research, authorship, and/or publication of this article.
References
Supplementary Material
Please find the following supplemental material available below.
For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.
For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.
