In-store food promotions increase sales as well as household food waste

Abstract

Food waste is a serious problem affecting both global food security and climate change. Using a large-scale field experiment with data from 43,246 perishable vegetable purchases from eight supermarkets, we show that in-store offers lead to over-purchasing of food items and result in an increase in household food waste. Shoppers exposed to multi-unit offers (e.g., “2 for $5” or two-for-one pricing) purchase greater food quantities compared to those exposed to a single-unit discounts (e.g., “$2.50, regular price $2.89” or “50% off”). A follow-up survey shows that these additional items are subsequently less likely to be consumed, leading to an increase in household food waste. A complementary online survey provides further support for these results. Importantly, as a part of the field experiment, we test two strategies that can address in-store over-purchasing. Taken together, the current findings provide important insights into factors that increase over-purchasing and household food waste, as well as strategies that can help overcome them. We thus demonstrate how retailers can negatively, but also positively, impact important societal and environmental causes.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-025-01800-x.

Introduction

A third of all food is lost or wasted¹. Over $1 trillion worth of food go to waste each year, while 783 million people worldwide struggle with hunger¹. Food waste also presents a pressing and avoidable climate problem². The production, transportation, and handling of food generate significant amounts of CO₂, while rotting food in landfills releases considerable quantities of an even more potent greenhouse gas—methane³.

Food waste, as defined by the U.N. Environment Program, comprises food and associated inedible parts (e.g., peel, stem) that are removed from the human food supply chain. Food is lost at many stages along the supply chain, with 13% of global food failing to reach the retail stage¹. In the current paper, we focus on household food waste, i.e. edible parts of the food intended for human consumption that were not consumed after being purchased at a retailer, a segment that represents 60% of the food waste from and after the retail stage¹.

Awareness of food waste as a societal problem is generally high among consumers, but intentions to reduce food waste often fail to translate into action^4,5. Household food waste is the result of a chain of decisions over several days or weeks, from planning and buying a product, to storage, cooking, eating and ultimately disposal of the product. A comprehensive conceptual overview lists 57 factors that may lead to food waste behaviors at each stage of this “squander sequence”, i.e., the stages along which consumers make food consumption decisions⁶. In more recent work, the lack of planning and “over-purchasing” as a potentially important driver for household food waste is highlighted⁷. The idea that unplanned “over-purchasing” drives food waste points to the role that retailers can play in creating, or potentially reducing, household food waste. Retailers can help reduce food waste by (1) upgrading inventory systems with the latest technology, (2) partnering with farmers in the supply chain, (3) teaming up with consumers, and (4) modifying or eliminating store practices that increase waste⁸. This last factor, i.e., store practices, is the focus of the current research. While all these sources highlight the urgency and importance of examining the interaction between retailers and consumers in creating food waste, causal insights into the impact of marketing practices on household food waste are still lacking^9–11.

The goal of the current research is to help close this gap by examining the role of in-store promotional offers on purchase decisions leading to household food waste. Using a large-scale field experiment and an online survey, we examine whether multi-unit offers lead to an increase in purchase quantity and whether this increase leads to household food waste. In collaboration with eight major supermarkets, we conducted a field experiment on 43,246 vegetables purchases by everyday shoppers to assess the real-world effects of multi-unit offers on purchase quantity. We find that multi-unit offers (i.e., price for two = x) increase sales (number of units sold) compared to equally-priced single-unit discounts (i.e., price for one = x/2) by 19.5%. To understand whether this increase in purchase quantity leads to increased household food waste, we surveyed participants of the field study by attaching small stickers with QR codes to 21,000 vegetables in stores during the experiment. We find that products were more likely to be consumed when shoppers randomly encountered and purchased products on single-unit discount in-store offers, compared to multi-unit offers, demonstrating the causal effect of promotional offers on reported household food waste. We then complemented the field experiment with a controlled, highly-powered online survey to further explore the role of multi-unit offers on household food waste. The results of the online survey are consistent with the findings from the field study and provide further support for the role of in-store multi-unit offers in increasing household food waste.

Finally, our field experiment also tested two in-store interventions intended to reduce over-purchasing: making the comparison price of a single item salient and using a prompt to remind people to consider how much they really need to purchase. Both interventions significantly reduced in-store purchasing compared to multi-unit offers. We conclude the article with a discussion of key implication of our results for consumer decision making, retailing and promotional offers, food waste, and sustainable business practices.

Related literature and theoretical reasoning

A number of conceptual papers have raised open questions about the role of retailers in creating household food waste. For example, research in sustainability has recognized the potential relationship between retailer price discounts and household food waste and has asked the question “To what extent do in-store stimuli and offers contribute to food waste?”^10,12. Multi-unit offers (such as two items at a reduced price or two-for-one), are especially highlighted as a potential factor leading to increased household food waste through increased purchase quantity^6,9,13,14. In surveys, households report their belief that they are affected by in-store marketing prompts to buy more which then leads to more waste at home^15,16. Yet, these propositions mostly went untested because examining the effects of in-store promotions on household food waste is a difficult task. To do so, a researcher must manipulate which offers shoppers are exposed to and track not only purchases but also which food is consumed and/or thrown away. As a result, researchers have so far mostly shown correlations rather than providing causal evidence^10,17. To address these shortcomings, in the current research, we use a hybrid approach combining a large-scale field experiment that randomly assigns shoppers to target offers with online surveys to empirically examine whether multi-unit offers lead not only to higher purchase quantity but to household food waste as well.

Why might multi-unit offers increase household food waste compared to single-unit discounts when the price for one unit is the same across both offers (or the savings from buying more are minimal)? Rational economic theory would predict that, given the identical price, consumers should buy only as much as they plan to consume—whatever the format of the promotional discount. There is mounting evidence, however, that consumers have limited attention and that firms can exploit this limited attention for profit¹⁸. Consumer “mistakes” due to inattention have been documented, among others, within consumer goods^19,20, taxes and insurance choices^21,22, financial markets²³, and utility markets²⁴. Further, there are many striking examples of consumers resorting to a simple heuristic instead of engaging in effortful information processing when making decisions. For example, in a used car market, people show a left digit bias, which leads to overspending²⁵; consumers misperceive effects of percentage changes greater than 100%, which leads to foregoing higher pay²⁶; while in the grocery setting, across almost half of all purchases, consumers purchase an item closer to expiration date, even when a fresher one is available at the same price²⁷.These examples demonstrate how consumers’ limited attention influences their choices.

One of the most powerful decision heuristics that occurs due to limited attention is the default effect. A recent meta-analysis of 58 studies showed that defaults, particularly in the consumer domain, have a strong effect on choices²⁸. A simple framework explains why defaults affect choices²⁹. It puts forth three explanations: (1) endorsement, (2) ease, and (3) endowment. The endorsement explanation suggests that consumers use the default to infer what the choice architect would recommend. The ease explanation implies that, especially in low-stakes situations, it is easier to stay with the default compared to exerting cognitive effort in deciding on another option. And lastly, the endowment explanation pertains to the idea that moving away from the status quo can be perceived as a loss by the decision maker. In the context of the current study, this framework would suggest that consumers are likely to purchase higher quantities of food when exposed to multi-unit offers because (1) they believe that the suggested promotion is beneficial to them (endorsement), (2) they need to exert less cognitive effort when staying with the suggested amount (ease), and/or (3) they perceive the multi-unit promotion as the status quo (endowment). More specifically, when an option is pre-selected or suggested (i.e., made a default option), consumers may not evaluate alternatives but rather only assess whether this default option satisfies them³⁰. Of interest to the current paper, we propose that when consumers are exposed to a single-unit discount, they will perceive one unit as the default; however, when they are exposed to a multi-unit offer, they will perceive two or more units as the default and will be more likely to (unintentionally) purchase higher quantities. This would explain why, when exposed to multi-unit offers, consumers over-purchase, which may ultimately result in some of that food ending up as waste. This theorizing informs debiasing strategies that can counteract over-purchasing stemming from the default effect. First, highlighting the unit price to show how much they can save has been shown to decrease consumer spending³¹. Second, reminding consumers via a gentle prompt to consider how much they really need to purchase can activate reflective and future-focused thinking and avoid mindless over-purchasing³². We test these the two debiasing interventions in the current paper.

Field experiment and follow-up survey

Field experiment

We conducted the natural field experiment in collaboration with eight large Swedish supermarkets from two store chains. Ethics approval was obtained from the University of Copenhagen, Department of Economics, Ethics board and all relevant guidelines were adhered to. For the survey part, informed consent was collected at the beginning of the survey. We collected sales data over 2 weeks (8–21 March 2021). Holding prices constant, we randomly exposed shoppers to one of four experimental conditions when buying fresh vegetables as shown in Fig. 1. Besides the Multi-Unit Promotion condition (1A; e.g., “2 for 30 kr”) and the Single-Unit Discount condition (1B; e.g., “1 for 15 kr”), we also included two additional conditions to test potential interventions for reducing over-purchasing. In the Salience condition (1C), the comparison price for purchasing one unit (e.g., “Reg. price: 15.95 kr”) was made salient by being displayed in a larger font. Doing so makes it salient just how small the savings from the offer are: buying two items instead of one saves 0.95kr or around 6% per unit. The Prompt condition (1D) included a small speech bubble stating: “I am happy to come home with you if you will eat me.” The idea here is that the speech bubble prompts consumers to consider whether they will consume both products in the following days. It encourages shoppers to question the status quo and the implicit recommendation by the store to buy two.

Fig. 1.

The treatments used in field experiment. Treatment A is the Multi-Unit Offer baseline with the price for when the customer purchases two products for price X, B is the Single-Unit Discount which keeps the price equal to X/2., C is the Salient treatment and D is the Prompt, which both have the same price X, as in the baseline, when the customer purchases two.

We focused on fresh produce because about half of all wasted food comes from fruit and vegetable categories³³. In six stores we ran the experiment only on cucumbers, in two stores we ran two experiments, both on cucumbers and on broccoli. We chose these two products for the following reasons. A food waste diary study conducted in Sweden a year prior to our study showed that both broccoli and cucumbers are among the top five most wasted fruits and vegetables among Swedish households, with cruciferous vegetables such as broccoli being number one³⁴. Our data shows that, on average, 309 units (370 cucumbers and 64 units of broccoli) were sold per day in each of the stores, for a total of 43,246 products sold. In addition, both items are sold per unit and not by weight to allow us to manipulate single-unit discount versus multi-unit offers. Both items are individually wrapped making it possible to attach stickers to invite shoppers to complete the follow-up survey. Finally, both vegetables expire within a week so that we can measure food waste and are always (cucumber) or often (broccoli) purchased fresh (and not frozen). Another benefit of using these products is that because of their high sales volume, supermarkets could quickly adjust their stock based on daily sales and thus did not face the risk of in-store food waste. Across all conditions, we held the promotion price constant (price for two = x; price for one = x/2). Since the supermarkets were in different areas of the country, the prices of the cucumbers and broccoli could not be identical across stores. We only required the stores to keep the price identical over the 2-week period and their assigned treatments. In seven out of eight stores prices were either 28 kr (four stores) for two cucumbers or 30 kr (three stores). One store used 24 kr for two cucumbers. The price for broccoli was either 15 kr for two or 22 kr for two. We do not find any meaningful differences across these two categories, so we report aggregate results.

Each of the eight participating stores was randomly assigned two conditions for our main product cucumber (and two for broccoli, if relevant). Thus, each intervention was tested in four stores. In total, we have ten store-level experiments, each testing two interventions. To control for variation in the number of shoppers, and differences in shopping behavior between weekdays and weekends, the 14-day test period is divided into three periods using an A-B-A design, where each treatment was implemented for 7 days in total. The stores’ first intervention is active during the first 3 days of the experiment. On the fourth day, the price displays are changed to the second intervention. During the final 4 days of the experiment, the first intervention is used once more. This treatment schedule was also randomized. It was chosen so as to balance statistical power with the feasibility of implementation for the store employees. Before randomizing the supermarkets to treatments, we stratified on size of the supermarket. The supermarkets are spread out across the country and vary in size. The experiment included four large supermarkets and four smaller ones. After the randomization, we sent the supermarkets their treatments and asked them to create signs that fit their corporate design. We did this to ensure external validity by making the experiment as natural as possible and to minimize the likelihood of any potential experimenter demand effects. Shoppers were, thus, not aware that there was an experiment taking place. We vetted all signs before they were used to make sure that they were in line with the treatments. This approach allows us to compare the causal effect of the offers on purchase quantity and control for store fixed effects. It is possible, even likely, that the same customers shopped twice in the same store within the experimental period. However, it is impossible, given that the eight stores were spread out across the country, that they shopped in two participating stores. Close to 350,000 customers shopped in the participating stores during the experiment, so any repeated purchases are very unlikely to have an impact on our results. Furthermore, since the treatments were randomized across time, it is unclear how any within-shopper spillovers should affect our results.

We examined the in-store sales data, at the daily level, for each store and condition; that is, we analyzed the data for 43,246 units purchased during the experiment. We estimate the average treatment effects in a random effects model, conditioning on baseline covariates:

1

where UnitsSold(log)_jt is the outcome variable by store (j) and day (t), SingleUnit Discount, Salience and Prompt are dummy variables that equal 1 when supermarket j is in the respective treatment group and 0 otherwise, and X_jt is a vector of observables for each store that are potentially predictive of sales. Multi-Unit Offer condition serves as the baseline. We control for the number of shoppers and sales data for 2 weeks prior to the experiment. We do not control for price, as the price did not vary within a store over the time of the experiment by design and would thus pick up variation across store locations (stores in urban areas have higher prices than stores in rural areas). To control for unobserved differences, we conduct a random effects regression (µ_j is the random effect and ε_jt is the error term). We mainly chose the logarithmic form to ease interpretation of effect sizes in percentage terms, as sales numbers vary at different stores, making absolute numbers less informative. The log form also allows for easier comparison with the survey data effect sizes. Standard errors are clustered at the product-in-store level. We treat the broccoli and cucumber experiments as independent observations, as they were independently randomized. Since there were thousands of products and various offers displayed in the stores, it is highly improbable that they would influence each other given the natural field experiment setting. We have 140 day and store observations (10 stores/products for 14 days each). By design, there is a nearly equal number of observations per treatment (Multi-Unit N = 35, Salience N = 34, Prompt N = 35, Single-Unit Discount N = 36).

As shown in the “Units sold in store” column of Table 1, as expected, sales (measured in units sold) were 19.5% higher in the two-item Multi-Unit Offer (i.e., baseline condition) compared to the Single-Unit Discount condition (p < 0.001). This indicates that consumers are influenced by retail offers and purchase 19.5% more units of food when they are presented as a multiple-unit offer (i.e., “2 for x” compared to a “1 for x/2”). We next assessed the effects of the two interventions for reducing over-purchasing. We find that compared to the Multi-Unit Offer condition, sales were 10.8% lower in the Prompt condition (p < 0.001), and 9.1% lower in the Salience condition (p < 0.001).

Table 1.

In-store sales in log form (column 1) and survey-reported purchases in log form (column 2).

	Units sold in store relative to multi-unit	Follow-up survey reported purchases relative to multi-unit
Single-Unit Discount	− 0.178***	− 0.176*
	(0.020)	(0.084)
Salience	− 0.095***	− 0.088
	(0.018)	(0.070)
Prompt	− 0.114***	0.071
	(0.027)	(0.064)
Shoppers	0.0005***
	(0.000)
Week-2	− 0.0002***
	(0.000)
Week-1	0.0004***
	(0.000)
Store size		− 0.113
		(0.060)
Constant	3.774***	0.637***
	(0.224)	(0.259)
Observations	140	176
R2	0.837	0.067

The baseline is sales in multi-unit offer condition.We have 140 store/day observations (column 1). We have 176 survey responses (column 2). In column 1, we control for number of shoppers in the stores and 2 weeks of pre-sales. In column 2, we control for store size. Percentage changes for binary predictors were calculated as [exp(β) − 1] 100 for transitions from 0 to 1 and [exp(− β) − 1]100 for transitions from 1 to 0. Standard errors in parentheses. *p < 0.10; **p < 0.05; *** p < 0.01.

Our field experiment, thus, confirms that multi-unit offers lead to higher purchase quantities. At the same time, we demonstrate two easy-to-implement strategies for reducing over-purchasing, although these interventions may go against profit-maximizing objectives of retailers, as discussed in more detail in general discussion. We next examine the impact of multi-unit offers on household food waste.

Follow-up at-home survey

We hypothesized that since consumers purchased more food in the Multi-Unit Offer condition compared to the other conditions, this would lead to an increase in household food waste. We test this via a follow-up survey. We asked the supermarkets to attach small stickers with QR codes to 21,000 vegetables, which corresponds to 150 stickers per condition, day, store, and product type (cucumber or broccoli). Using a different QR code for each condition, store and product type, we were able to track under which condition the product was purchased without needing to ask respondents and thus draw their attention to the treatment. The stickers invited consumers to take part in a 2 min incentivized online survey. To avoid self-selection into buying the vegetables, the stickers were kept as small as possible in line with the store corporate design, making them inconspicuous, and did not mention food waste. We verified their placement using photos of the displays. When scanning the QR code consumers were asked to enter their email address to receive the survey. Requesting the email address serves two purposes. First, it allows us to control whether a person tried to participate twice (no one did). Second, since fresh produce, such as broccoli, will keep in a refrigerator for 3 to 5 days³⁵ we could send participants the survey link 1 week after they registered to take the survey to allow enough time for the food to either be consumed or thrown away. Participants who completed the survey received an equivalent of a $10 (i.e., 100 kr) coupon for the store chain in which they purchased the items (i.e., they were emailed a code they could scan at the register). Participants reported how many items they purchased and whether these items had been consumed (a binary variable: Yes or No). They also answered questions about their shopping habits and household demographics (see Supplementary Information A for all survey items). Survey measures can potentially suffer from desirability bias, inattention or misunderstanding of the question, i.e. what is meant with “waste” or “fully consumed”³⁶. We argue that while this is a problem in traditional consumer surveys and could impact the level effects, our randomized controlled trial design nevertheless allows us to estimate treatment effects, as there is no reason to assume that these biases should vary systematically by condition.

One hundred and seventy-eight supermarket shoppers completed the survey, corresponding roughly to a 1% response rate, which is typical for marketing surveys of this type. Two participants reported purchasing zero items and were removed from the study. A similar number of shoppers responded in each treatment (Multi-Unit Offer = 37, Salience = 51, Prompt = 38, Single-Unit Discount = 50), which rules out self-selection into the survey based on treatment (χ²(3) = 1.96, p = 0.580).

First, we examined purchase quantity, i.e., how many units of cucumbers or broccoli participants reported purchasing. We estimate an OLS regression with standard errors clustered at store level:

2

where y is the log of products purchased as stated in the survey, Single-Unit Discount, Salience and Prompt are the treatment dummies and StoreSize is a dummy variable equal to 1 if the store was a large store. The subscripts (i) stand for individual observations. Overall, we find that the follow-up survey data is representative of the in-store data. As shown in the “Follow-up survey reported purchases” column of Table 1, participants in the Multi-Unit Offer condition reported purchasing 19.2% more units compared to the Single-Unit Discount condition. This number is nearly identical to the actual increase observed from the in-store sales data and, thus, demonstrates that survey respondents are representative of the overall group of shoppers. At the same time, compared to the Multi-Unit Offer condition, we observe a decrease in purchase quantity of 8.4% in the Salience condition and 7.4% in the Prompt condition. While the point estimates are similar between the in-store data and the at-home follow up survey, the estimates for the Salience and Prompt condition are not statistically significant at conventional levels, likely due to the limited sample size.

Next, we examined whether shoppers consumed all of the purchased food. One hundred and twenty-one supermarket shoppers reported purchasing two or more food items and we focus on these participants. Importantly, we find that only in the Single-Unit Discount condition, all participants reported consuming all the food, not letting any of it go to waste. In the Multi-Unit Offer condition, 12% of participants reported not consuming all the food compared to 0% in the Single-Unit Discount condition (a two-proportion z-test, 12% vs. 0%; z = 1.85, p = 0.064, Cohen’s h = 0.71). Further, 15% of participants in the Prompt condition and 6% of participants in the Salience condition reported not consuming all the food, neither of which were statistically different from the Multi-Unit Offer condition (Prompt: z = 0.345, p = 0.730, Cohen’s h = 0.09; Salience: z = 0.902, p = 0.367, Cohen’s h = 0.23). These results are influenced by limited sample size, which is why we conducted a complementary controlled online survey, as discussed in the next section, and observed consistent results.

Finally, we examine how demographics and shopping behavior influence our outcome variables, as shown in Table 2. When it comes to purchases (see the “Units purchased” column in Table 2), unsurprisingly, the main predictor of the quantity purchased was household size: 15.8% more items were purchased per each additional household member. When controlling for household size, none of the other factors had an effect on the amount purchased. When it comes to food waste (see the “Fully consumed” column in Table 2), the having a larger household size seems to have a slightly negative effect on consuming all of the products, although the effect is small and only marginally significant. Interestingly, whether shoppers expect that they are affected by marketing practices, whether they frequently buy too much or buy unplanned, has no effect on the quantity purchased or the amount of waste. These results then support the argument that consumer perceptions are not a reliable indicator of the effects of offers on consumption and food waste, underscoring the need for the natural field experiment carried out in the current research.

Table 2.

Predictors of units purchased, consumption, and consumption expectations.

	Units purchased	Fully consumed	Consumption expectations
Household size	0.147*	− 0.054*	− 1.807
	(0.054)	(0.021)	(1.122)
Nr. of kids	− 0.048	0.055*	1.966
	(0.049)	(0.062)	(0.852)
Shopping frequency	− 0.023	0.009	− 0.097
	(0.054)	(0.026)	(6.331)
Buy too much	− 0.007	− 0.069	− 2.626
	(0.009)	(0.030)	(1.533)
Buy unplanned	0.010	0.003	− 0.077
	(0.016)	(0.025)	(0.768)
Influenced by marketing	− 0.018	− 0.016	0.709
	(0.010)	(0.017)	(0.967)
Constant	0.276*	1.215***	95.642**
	(0.095)	(0.132)	(26.960)
Observations	175	175	175
R2	0.116	0.087	0.027

Units purchased is the reported amount of cucumber or broccoli purchased. Fully consumed is a binary variable which is 0 if not all of the purchased products were consumed at least 1 week after purchase. Consumption expectations is a scale from 0 to 100% measuring shoppers’ expectations at time of purchase about consuming all products. Household size and Nr. of kids are continuous variables. Shopping frequency is measured on a three-point scale from daily to weekly, Buy too much, Buy unplanned and Influenced by marketing are measured on a scale from 1 to 7 with 7 showing highest agreement. Standard errors are clustered by treatment and presented in parentheses. *p < 0.10; **p < 0.05; ***p < 0.01.

In sum, the results of the field experiment and the follow-up survey suggest that multi-unit in-store promotions affect purchase quantity and lead to an increase in household food waste compared to single-unit discounts.

There are a couple of limitations to the field experiment. While it ensured realism and generalizability, and supermarkets placed 21,000 stickers on vegetables to recruit survey participants, our resulting limited sample size affected statistical power, even though we observe a medium effect size for our main treatment. In addition, as mentioned above, it is difficult to measure actual amounts of household food waste as it is not possible to measure actual waste in shoppers’ homes. We resorted to asking participants whether they have consumed all of the purchased food. We refrained from asking more detailed questions such as “what percentage of the edible part of the cucumber have you consumed?” as participants’ perceptions of whether the peel or the end of a cucumber, or broccoli stem, are edible or not might vary. A randomized controlled trial, as presented here, has the benefit that any biases or misunderstandings should be equivalent across treatment groups, allowing us to focus on the causal effect of the treatments on our outcome variables. We selected two food items that either cannot be frozen (cucumbers) or are not conventionally purchased fresh and then frozen (broccoli) so that if shoppers buy more than necessary, their only options would be to increase consumption or throw away the food. Consistent with this logic, research showed that consumers perceive food they have stored for a while as less fresh, even if it has not expired or gone bad, and are more likely to throw it out³⁶. Nonetheless, our measure does not definitively tell us that shoppers have actually thrown the food away. To address some of these shortcomings, we next conducted a complementary controlled online survey. The survey is highly powered to test our main hypothesis that multi-unit offers increase the likelihood of household food waste. While such a controlled experiment sacrifices some realism, when complemented with the field experiment, the two studies provide a robust test of the effects of multi-unit offers on household food waste.

Online experiment

We conducted an online survey to directly test the hypothesis that multi-unit offers increase household food waste. The study was preregistered (AsPredicted #127914). This study was approved by the Southern Methodist University Institutional Review Board and all relevant guidelines were adhered to. We planned to collect 100 participants per experimental condition, which resulted in the statistical power of 99% (above the recommended threshold of 80%). A total of 397 participants recruited from United States completed the survey (CloudResearch; M_Age = 42.58 years, SD = 12.22; 51% female). The design of the study was adapted from van Lin et al.³⁷. Participants were told to imagine that they go to the supermarket to buy ingredients for multiple meals and that they plan to buy one bunch of fresh broccoli. Participants were randomly assigned to one of four conditions (see Supplementary Information B for stimuli). They were told that they end up purchasing one bunch of broccoli (Regular Price One Item)/ two bunches of broccoli (Regular Price Two Items)/ one bunch of broccoli at a discount of 50% (Discount)/ or two bunches of broccoli as part of a two-for-one promotion (Multi-Unit Offer). They also saw a picture of the food and, in the two conditions with a price promotion, a sign displaying the corresponding deal. After reading the scenario, participants rated Waste, i.e., the likelihood that some of the broccoli may end up going to waste over the next few days (0 = “not at all likely” to 100 = “extremely likely”). Finally, as an attention check, participants were asked to drag a slider scale all the way to the right; eight participants failed to do so and, per preregistration, were removed from the analyses, resulting in the final sample of 389 participants (64 participants failed a pre-survey 2-question attention check and did not enter the study).

We conducted a one-way ANOVA on Waste. Figure 2 shows the means for all conditions. We find that Waste differs significantly across conditions (F(3, 385) = 8.01, p < 0.001, ηp² = 0.59). Tukey post-hoc test results revealed that participants who purchased on multi-unit offer reported higher likelihood of food waste (M = 54.73, SD = 28.75) compared to those who purchased one product on 50% off promotion (M = 42.07, SD = 32.50; p = 0.021) or those who purchased one product at regular price (M = 34.19, SD = 28.68; p < 0.001). Further, those who purchased two products at the regular price (M = 48.26, SD = 32.45) also reported higher likelihood of wasting food compared to those who purchased one product at regular price (M = 34.19, SD = 28.68; p = 0.009). This served as a simple check to show that purchasing larger food quantity, indeed, leads to more household food waste. No other differences were significant. These results are consistent with the findings from our field experiment, providing further evidence for the hypothesis that multi-unit offers increase the likelihood of household food waste.

Fig. 2.

The effects of multi-unit offers on household food waste. Bars show SD. p < 0.10; **p < 0.05; ***p < 0.01 No other differences were significant.

Discussion

Food waste is a serious problem affecting both global food security and climate change. Across a field study, a follow-up at-home survey, and a complementary online survey, we find that in-store multi-unit offers (i.e., two items at a reduced price or two-for-one promotions) lead to both higher food purchases and more household food waste compared to single-unit discounts.

Our findings contribute to the literatures on consumer decision making, in-store promotional offers, and food waste. Deciding how much to buy is one of the key decisions consumers make^14,38–41. Because everyday shopping is an “automatic”, low-involvement behavior (e.g., purchasing a cucumber or broccoli), the use of “default” offers to present larger-than-typical quantities can lead shoppers to unintentionally purchase larger quantities⁴². While previous research showed that an increase in purchase quantity can lead consumers to consume at a faster rate^42–45, in the current research, we document another important consequence of increased purchase quantity—an increase in household food waste.

Two recent empirical papers found that food purchased on discount is wasted less than food purchased at regular prices^17,37. What can explain the difference in results between these two studies and the current research? In both studies, consumers were able to deliberately search for multi-unit discounts, that is, they were able to self-select into stores with advertised multi-unit discounts for target products. This opens a possibility that consumers who are price sensitive may intentionally seek out multi-unit offers in order to conserve (i.e., freeze) food for later consumption, which would be in line with earlier studies that found that consumers who identify as price conscious throw away less food^46,47. In contrast, our field experiment randomly exposes shoppers to offers in stores, thus precluding the possibility for price-conscious consumers seeking out promotional offers. In our field experiment, the average savings stemming from making purchases on multi-unit offers were only around 6% of the purchase price—less than 2 kr or $0.18. Hence, the popular argument that these types of discounts are beneficial for low-income consumers¹⁷ seems not to be overly strong, when considering the likely monetary losses due to food waste. Rather, to help price-sensitive shoppers stores can implement single-unit discounts which, as we show in our online experiment, result in lower likelihood of food waste compared to a multi-unit discount.

We further demonstrate two easy-to-implement in-store interventions that can reduce over-purchasing. While reducing over-purchasing should mechanically also reduce household food waste by leading shoppers to bring less unplanned produce into their homes, these results failed to reach statistical significance, likely due to limited sample size. Scarce research has demonstrated the possibility of reducing food waste through behavioral interventions in restaurants^48,49, but we are not aware of any randomized controlled trials examining the effectiveness of interventions to reduce household food waste. Our findings are also in line with the literature on behavioral interventions, especially in the food domain, where most nudges consistently produce small-to-medium increases in, for example, healthier meal choices⁵⁰.

The current research offers several important implications for policymakers. The U.N. Sustainable Development Goal 12.3 aims to halve per-capita global food waste at the retail and household levels. Many policy suggestions, such as teaching consumers about food waste with information campaigns⁵¹ or reducing packaging size⁴⁷ take significant time to implement and offer uncertain outcomes. In contrast, a policy abandoning multi-unit offers for highly perishable vegetables could be implemented immediately by supermarkets. Ideally, such a voluntary industry initiative would also have positive spillovers toward more stringent governmental regulation on food waste such as governmental regulation of multi-unit offers⁵². Such a systemic change would also reduce the attentional burden on consumers trying to process the informational onslaught of in-store offers^53,54. To the extent that a ban on multi-unit pricing is deemed impractical or politically infeasible, we demonstrate two easy-to-implement in-store interventions that can reduce over-purchasing. Both interventions aimed to snap consumers out of the automatic “default" mode so that they may consider their purchases more carefully. Our policy recommendations are not in conflict with the suggestions to use dynamic pricing on products that are close to their expiration date and would otherwise lead to store food waste⁵⁵. These types of discounts are rarely implemented as multi-unit offers and it is normally salient why they are on discount and that they should be consumed immediately.

Of course, the recommendations discussed here go against the short-term profit-maximizing objectives of most retailers, which may explain why multi-unit offers are so common. However, regulation of multi-unit offers could start by focusing on highly perishable products such as produce, bread and dairy, which make up only a fraction of the products sold by most retailers but account for a large share of household food waste. The current research, thus, also demonstrates the challenges retailers face today as they try to balance profit-maximizing objectives with corporate social responsibility, including environmental, sustainability, and governance goals, here specifically related to sustainability goals.

Electronic supplementary material

Below is the link to the electronic supplementary material.

Author contributions

C.G. designed, conducted and analyzed the field experiment and the follow-up survey. C.G. designed Fig. 1. M.M. designed, conducted and analyzed the online experiment. Both authors wrote and revised the manuscript.

Funding

Funding was provided by Livsmedelsverket, Sweden.

Data availability

All of the data and code used in the paper are available on OSF: DOI 10.17605/OSF.IO/7NSBP.

Code availability

The code to run the analyses and reproduce the figures is available on: DOI 10.17605/OSF.IO/7NSBP.

Declarations

Competing interests

The authors declare no competing interests.