Using QR codes to access food information: A behavioural study with European consumers

Using QR codes to access food information: A behavioural study with European consumers | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Using QR codes to access food information: A behavioural study with European consumers Alexia Gaudeul, Michal Krawczyk This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-8549197/v1 This work is licensed under a CC BY 4.0 License Status: Under Revision Version 1 posted 9 You are reading this latest preprint version Abstract We conducted an online experiment with 3420 participants in Bulgaria, Germany and Spain to evaluate the consequences of digital access to food information via QR codes. Participants made incentivized choices between food products with labels that were either “paper” or “hybrid”. Only the latter type featured a QR code; scanning it was necessary to access some of the information. Presence of QR codes did not affect probability of choosing given product. As many as 37% of participants never scanned QR codes, and only 4% scanned all of them. Hybrid labels led to longer decision times and less accurate knowledge, indicating a negative impact on consumers. consumer choice digital labelling food information discrete choice experiment food labels QR codes Figures Figure 1 Figure 2 Figure 3 1 Introduction Food consumers should be able to access comprehensive information concerning the products they buy in a quick and convenient manner. However, this can be particularly challenging for products with limited label space and in cases when customers may expect more extensive information, for example about the origin of certain foods (Lombardi et al, 2017 ), the manufacturing process (Bradford et al, 2022 ), or producer’s sustainability practices (Kim and Woo, 2016 ). Firms may also want the consumers to be able to access Internet resources allowing audio, video, interactive, and personalized content (including targeted offers). One way to do so involves the use of hybrid labels whereby only some information is written on the label while a Quick Response (QR) code directs to additional online resources. Those QR codes can be scanned while purchasing a product (van der Laan & Orcholska, 2022 ) or later on during preparation or consumption. The question we investigate in this paper is whether information available in such a way is accessed, perceived, processed, and retrieved in a similar fashion to that available directly. Our objective here is to pursue a systematic experimental investigation to address these issues. We do not only want to understand whether consumers prefer products where food information is provided on-label rather than products that offer the same information digitally, but also how often they access the digital information, whether it slows down their choice, and whether it affects the accuracy of their knowledge of food information. These questions are important for measuring the effects of regulations concerning provision of food information on the welfare of consumers. While corporations underline the benefits of digitalization of food labels, allowing them to cut the costs of packaging, attract consumers to their websites, strengthen brand loyalty and personalize offers (FoodDrinkEurope, 2024; QualityChain, 2024), the consequences for the consumers must be carefully examined. Our study thus contributes to the important discussion on the consequences of using digital means to convey food information, and more specifically, of the effects of splitting such information across communication channels. 1.1 Labels and food choice Labels are a key source of information about food, but consumers often find them confusing (Roberto and Khandpur, 2014 ; Temple and Fraser, 2014 ). Full and correct processing of the information they provide cannot be taken for granted. On average, consumers may be making up to 200 food decisions a day (Wansink and Sobel, 2007 ), so most of those decisions must be made very quickly. They cannot consider all information on labels at the time of purchase. Food choices thus typically follow the “direct heuristic route” corresponding to rough, error-prone “system 1”, as found in eye-tracking research (Ma and Zhuang, 2021 ). So-called top-down attention, driven by consciously formulated goals and plans, plays a limited role in food choice (Fenko et al., 2018 ). Bottom-up attention, which is driven by external stimuli, is a more important factor. Some studies find that labels have a limited effect on actual food choice (see Ma and Zhuang, 2021 for a review). However, they may still affect consumers’ willingness to try new foods (McFarlane and Pliner, 1997 ) and to pay for dietary products (Øvrum et al., 2012 ). Some studies (Orquin et al., 2020 ) find that the capacity to affect choices depends on such qualities as size, prominence, and salience of a label. We therefore must take account not only of the fact that providing food information via a QR code affects the likelihood to access this information, but also how this mode of delivery affects the choice process. For example, digital information may be accessed only after deliberation and consideration of other information on the label, rather than automatically. Furthermore, digital information that is shown separately (on one’s smartphone for example), may have a different impact than information that is shown along other information on the same support. Finally, consumers may prefer products with directly accessible information, or conversely prefer labels with less information. All this may result in different product choices. 1.2 Research questions and hypotheses This experimental behavioural study measured participants’ preferences as well as their access to, and knowledge of food information presented either via a QR code or directly on-label. To improve control over the information participants could access, the experiment featured and compared labels containing no QR code (henceforth “paper labels”) to those with one type of information (e.g. caloric contents) accessible via QR code and other types of information available directly (henceforth “hybrid labels”, with the part accessible via QR code referred to as Digital Food Information, DFI). This setup allowed us to address seven important inter-related research questions and associated hypotheses. Our first question is 1) whether products with paper labels are preferred to products with hybrid labels (or the other way round), other things being equal. In general, people tend to choose options that are easier to evaluate (Garbarino and Edell, 1997 ). As paper labels provide all the information in the same manner, they also probably make it easier to compare and thus evaluate products with respect to each other. However, empirical evidence on QR codes is very scarce in this respect. Oonk ( 2013 ) found that purchase intentions were independent of whether information was provided digitally or on-label. We test the following hypotheses: H1.1: Participants will be more likely to choose a product with a paper label when given a choice between a product with a paper label and one with a hybrid label than when given a choice between two products with paper labels. H1.2: Participants will be more likely to choose a given product when it is displayed with a paper label than the same product when it is displayed with a hybrid label. How QR codes affect choices clearly depends on whether 2) consumers perceive information they open as sufficiently valuable to access it . The pragmatic theory of information sees it as a means or resource to solve a problem by overcoming uncertainty. Consumers are thus only expected to access information if it is perceived as sufficiently likely to affect their choice of product, given the costs of accessing it. These costs involve, in the case of QR codes, the effort and time of picking up one’s smartphone, opening the camera or a dedicated app, scanning the QR code, waiting for the information to load, and reading it. Then again, there is ample empirical evidence that such an instrumental view of information is overly simplistic, as humans (and other animals) are often willing to acquire information also when it has no pragmatic value (Kang et al., 2009 ; Zentall and Stagner, 2012 ). It could be motivated by sheer curiosity or willingness to have one’s opinions confirmed. Sometimes people may actively avoid obtaining information despite its positive pragmatic value (Golman et al., 2017 ). In the context of food choice, for example, they may prefer not to know how many calories their favourite desert has. Empirically, most studies find spontaneous use of QR codes to be low (Li & Messer, 2019 ; Bashir, 2022 ) unless a dedicated device is readily available (Li & Messer, 2019 ). Relatedly, Bray et al., ( 2019 ) reported that consumers tend to prefer on-label information when given a choice. Li & Messer ( 2019 ) showed how dramatically willingness to use DFI depends on how it can be accessed. In the scenario most comparable to ours, involving a link on a tablet provided by the experimenters, 20.2% of participants accessed it. The following hypotheses emerge: H2.1: The prevalence of DFI access will be low. DFI will be accessed less than one fourth of the time in our experiment on average. H2.2: The prevalence of access to DFI will diminish over the course of the experiment. Another welfare-relevant dimension is 3) the speed of choice . Holding the quality of the decision constant, information channels that lead to quicker decisions should be preferred. We are not aware of directly relevant extant research, but it seems natural to expect choices involving products with hybrid labels to take more time than choices involving only products with paper labels, given the extra time (if short) necessary to access DFI. Moreover, the Elementary Information Processes perspective (Bettman et al, 1990 ) predicts that with more information being processed, the processing time becomes longer, so that we expect that choices in which consumer accessed DFI to be slower than those in which she could access it but did not. It should be noted, however that later literature nuanced this picture, showing that additional information may, in fact, speed up decisions if it improves coherence, i.e. additional information promotes the option the consumer would be inclined to choose without this information already (Glöckner and Betsch, 2012 ). The following hypotheses can thus be put forward. H3.1: Among choice situations with at least one product with a hybrid label, choices in which DFI is accessed will take more time than those in which it is not. H3.2: Choice involving one or two products with hybrid labels will take more time than choices involving only products with paper labels. We also investigate 4) product knowledge . Naturally, if DFI is rarely even accessed, it cannot be expected to be known, which would result in inferior product knowledge in case of hybrid labels. However, one may argue that participants who care about a given type of information would likely be willing to access it via a QR code, and those who do not, would not read it even if it was printed on the label, so that overall, DFI would not result in inferior product knowledge. One could even go so far as to say that by requiring more active pursuit of information, QR codes may make the information more memorable to those who sought it. There is also some evidence of small advantage of reading from paper rather than the screen (Neijens & Voorveld, 2018 ; Clinton, 2019 ) but the mechanism and, therefore, border conditions for this effect are poorly understood. We test the following hypotheses (in which, since pre-registration, we have slightly changed potentially misleading wording): H4.1: Responding to knowledge questions, participants will be more precise if they accessed information than if they did not. H4.2: Participants will be more precise in their knowledge of information that was provided on the label than of DFI (this is because most of them will not access DFI). H4.3: For a given piece of information, participants will know it better if it is shown as DFI and they access it than if it is shown on the label (this is because we know they accessed the information in one case, thus signalling they are interested in it, while we do not know this in the other case). For all our dependent variables, we are interested in 5) Individual differences. Based on existing studies (Ozkaya et al., 2015 ; Mendelson and Bergstrom, 2013; Francis et al, 2019 ; Riddle, 2007 ), there are some reasons to expect greater use of DFI and better product knowledge in younger, better educated, richer, urban-dwelling individuals. We hypothesise as follows. H5: The following factors will make accessing DFI more likely and product information knowledge more accurate: higher education, younger age, higher income, urban location. Naturally 6) the type of information shown as DFI may also matter, although existing literature is scarce. Indirect evidence comes from Dickinson and Kakoschke ( 2021 ) who observed confirmation bias, in that participants caring about taste (rather than health) of the product were more likely to choose to watch a "taste matters most" (rather than "health matters most") video. The same is true for the role of 7) food category and characteristics. Based on the distinction between “haptic” and “nonhaptic” products (Jha et al., 2020 ) we could expect comparatively little interest in DFI for “haptic” products, i.e. those that consumers would be inclined to touch before buying. In our case, that would likely be the vegetable category. These considerations give rise to the following hypotheses. H6: There will be differences in the rate of access to DFI depending on the type of information it contains. H7: Rates of access of DFI and precision of knowledge of food characteristics will depend on the food category. The hypotheses had been pre-registered under https://osf.io/nwv38/ . We reproduced them here with some trivial reordering and re-labelling to ensure consistency within the paper. 2 Design of the experiment 2.1 General methodology We designed an incentivized discrete choice experiment, whereby participants were presented with choices between pairs of food products. The products were selected to represent food that is widely consumed and represent a broad range of types of food. In each case, participants saw a picture of the package, the price and mandatory food information (MFI) items, i.e. those, which, according to current EU legislation, must always be printed on label. We varied the mode of delivery of MFI, whereby some MFI items could be read on the paper label, and others were accessible via QR code only. 1 We motivated participants by giving them a budget that covered the purchase of four units of one of the products they chose during the experiment. At the end of the experiment, four units of one of the products they chose would be sold to them, at the price shown to them. The amount due was subtracted from the budget they were given. The money remaining was sent to them along the purchased items. 2 In this way, unlike in most related studies, participants had incentives to pick, from each pair, the product they really preferred, given their prices. We thereby avoid the “hypothetical bias” of such studies, whereby participants are not motivated to consider their choice carefully. Instead, they may choose what they think the experimenter wants them to choose (“experimenter demand effect”), or products that make them feel socially approved or that correspond best to their own ideal image of themselves (“warm glow effect”). The experiment was run online, whereby participants received an e-mail invitation and, if they decided to participate, clicked on a link to the experiment that could be accessed from their PC or tablet. This means that even the “paper” labels had to be read from the screen. On the one hand, this may be seen as a limitation on the external validity of the experiment. This design choice made “scanning the QR code” as quick as possible, without the need to reach for the device, launch the QR code scanner etc. As a result, the fraction of cases in which the QR code was “scanned” in our experiment cannot be directly taken as a predictor for the prevalence of QR code use in naturally occurring food choices. On the other hand, this feature of the design allowed us to control for various factors identified as potentially affecting the willingness to make use of digital food information. For example, as mentioned in our brief review, people may be reluctant to use QR codes out of concern for cyber security, a concern that was very unlikely to play a role in our experiment. As a result, we could focus on the essence, namely how information processing and choice depend on whether all food information is displayed immediately (paper labels) or an additional step is necessary to unlock some bits of information (hybrid labels). 2.2 Implementation The experiment was programmed by Open Evidence in collaboration with Schlesinger Group Spain, based on a design provided by the authors. It was translated in the local languages of each country in which it took place (Bulgaria, Germany, and Spain). 2.2.1 Design of the choice menu Participants in this study had to make incentivized choices among 16 pairs of existing products (see incentives ). There were four pairs of products in each of four food categories (see selection of food categories and products ). Each pair of products featured two real products that were selected to be similar to each other. For example, a 500g pack of rice at 2€ was compared to a 450g pack of rice at 1€90, and the consumer had to make a choice between the two. Appendix A provides a list of pairs of products that were presented in each country. Information about each product (see food information ) was the same as shown on label on the actual physical product. We then varied whether all that information was provided on label in the traditional way (“paper label”) or part of it was provided as DFI (“hybrid label”); see food information conditions below. Participants were asked to choose one product in each of the pairs proposed. Their choice was recorded, as well as whether they accessed the DFI (if any was accessible). We also recorded how long they took to make their choice for each pair of products, and how long they looked at the DFI (if at all). For the last three choices, they were also asked to report information concerning the products they just chose from (see knowledge questions ). 2.2.1.1 Selection of food categories and products There were four food categories, with four sub-categories each: carbohydrates (pasta, bread, lentils, and rice), dairy (milk, butter, yogurt and cheese), proteins (fish, peas, meat and a vegetarian meat alternative) and vegetables (such as green beans, carrots, tomatoes and salad). Consumers had to choose between two products in each sub-category, selected to be close substitutes, also in terms of price. Consumers thus had to make 4x4 = 16 pairwise choices . Food categories were selected to be representative of the typical food basket of a European consumer in countries selected for this study (Bulgaria, Germany and Spain). Food categories were the same across the three countries. Furthermore, sub-categories were very similar across countries. We varied, however, the type of cheese, the type of fish, the type of meat, or the specific vegetables. The specific products in each sub-category were country-specific, so as to correspond to actual brands available in each country. Finally, products in each sub-category were selected to be similar in terms of brand reputation, popularity and availability, quality, and relative price across countries. 2.2.1.2 Display of food information As explained below, we varied whether a product was presented with all MFI items on the label (“paper”), or with one of the Mandatory Food Information items accessible only by clicking on a QR code that was printed on the label (“hybrid”). The experiment was run online, so that “scanning a QR code” meant clicking on the QR code on screen, which opened a popup window that showed the missing information. Figure 1 shows a screenshot of such choice situations. A participant can click at the bottom on “Option 1” or “Option 2” to indicate his or her choice. As stated, we chose products in each pair to be close substitutes. For example, they are both tomato sauce in this case. We show a picture of both products, their price, and a label containing all mandatory food information (see “food information” below). The screenshot in Fig. 1 is only one example of a choice situation. We varied the type of information that could be accessed via the QR code (cf. “Food information” below), and we also varied which products included a QR code. In the above example, the QR code is on the product on the left. It could also be on the product on the right, on none of them, or on both. (cf. “Information condition” below). The label for Option 1 includes a QR code, with a text above it saying that it gives access to the “best before” date (via a pop-up window). The label for Option 2 on the other hand does not have a QR code, the “best before” date being provided directly on the label. Participants know they only need to click on the QR code for a “popup” to appear, which shows the “missing” information. They are given an opportunity to train to do this before starting to make choices. 2.2.1.3 Food information For every product, we showed a picture of the product, its price, as well as the following Mandatory Food Information items: 1. Brand name of the food. 2. List of ingredients (including those causing allergies or intolerances). 3. Nutrition declaration. 4. Net quantity of the food. 5. Date marking (use by/best before). 6. Any special storage conditions and/or conditions of use (when relevant). 7. Name or business name and address of the food business operator. Of those, the brand name and allergens were always shown on label. Other ingredients, and other types of information could be shown either on label or as DFI. 2.2.1.4 Information conditions There were four information conditions (ICs) for each product pair. The ICs were as follows: ● Paper-Paper: both products have paper labels, meaning that all information are shown on-label for both products in a pair. ● Paper-Hybrid: only the label on the right is hybrid, meaning that all information is shown on-label for the product on the left, while one of the Mandatory Food Information items is shown only via the QR code for the product on the right. ● Hybrid-Paper: only the label on the left is hybrid, meaning that all information shown on-label for the product on the right, while one of the Mandatory Food Information items is shown only via the QR code for the product on the left. ● Hybrid-Hybrid: both products have hybrid labels, meaning that both products show the same Mandatory Food Information item only via the QR code. Those ICs were systematically varied across pairs for a given participant, and across participants for a given pair (Appendix B gives more details on the randomization). 2.2.2 Timeline of the experiment The timeline of the experiment is represented in Fig. 2 . The experiment started with a short introduction explaining that participants would have to make choices between food products, as well as answer some questions and that their responses were anonymous. We then explained to them the incentive system in the experiment, whereby there was a one in four chance that four units of one of their chosen products would be sent to them. We ensured they agreed to take part in the experiment and to give us their address in case they were chosen to receive one of their chosen products. 1. Food familiarity questionnaire. After explaining how to make choices and what information about products would be available to them, we asked participants to fill a food familiarity questionnaire. This consisted in a list of products that were going to be offered for choice, shown as pictures. Participants selected which products they already knew or had purchased in the past. 2. Food choices. Participants then had to make 16 choices between food products presented in pairs, as explained in the previous section on the design of food choices. Namely, in each choice situation, they saw two food products, whereby both, one or none of the products had a hybrid label (see “Information condition”). They decided whether to click on the QR code(s), if available, and then chose one of the two products. They then went on to the next choice situation. 3. Knowledge questions. Immediately after each of the last three choice situations the participants faced, we asked them to report their best estimate of the weight, best before date, or number of calories, for both products in the last three pairs. 3 4. Final survey. Participants were asked a range of questions covering socio-demographic variables (age, gender, education …), use of Internet and digital tools, familiarity with QR codes, dietary and food related habits, as well as questions about their health, welfare, and ability to process choice information. Appendix C lists all questions asked. 5. Incentives . A lottery was played at the end of the experiment, whereby each participant had a one in four chances to get a budget of 20€ to pay for purchasing 4 units of one of the 16 products they had chosen, at the price shown to them. This was sent to them at their address with postage paid by us. They also received any money remaining from the 20€ budget after taking out the cost of the products. The lottery itself was verifiably random, whereby we told participants how to access the page source with the code of the program that drew a random number between one and four. 4 3 Sample and data collection The study was run in Bulgaria, Germany, and Spain. This country selection provides good variation in terms of geographic location, GDP level, dietary habits, and propensity to buy groceries online, among other factors. Data collection was done by Open Evidence in collaboration with Schlesinger Group Spain. The experiment took place between the 8th of February and the 15th of April 2023. The target population was EU residents above 18 years of age, with a target sample size of 1000 participants for each of the three countries. In each country, the sample was stratified to ensure a good representation of the population in terms of age, gender, regions, place of residence (rural/urban) and education level. 3.1 Mode of recruitment and administration of the survey The experiment was self-administered, online, using a computer or a tablet (for optimal display of images), with no time limits. Participants were recruited using a panel-blending approach, meaning they were drawn from multiple panels simultaneously to improve representativeness of the population of reference and reach specific representativeness targets. Potential participants from panels in each country were notified about the experiment's purpose and incentives. They were assured of anonymity and data privacy, with the option to freely and voluntarily consent or decline participation. The experiment was accessed through a one-time access link provided in the invitation, which also included links to the SGS Research Privacy Policy, a removal link, and contact information for the project manager. 3.2 Pre-registration, data and analytic code The following exclusion criteria, as well our hypotheses and the analysis of results, were pre-registered at the Open Source Framework’s open registries network ( https://osf.io/nwv38/ ). Pre-registration guarantees that none of our results are affected by hindsight bias or subject to fishing for significance (Nosek et al., 2018 ). The wording of all the stimuli, the data, analytic code, and further supporting materials are also available at https://osf.io/nwv38/ . 3.3 Exclusion criteria In total, 3835 participants completed the experiment. In line with the pre-registered procedures ( https://osf.io/nwv38/ ) , we excluded 35 participants who failed a basic attention question and then the fastest 5% and the slowest 5%. This meant excluding participants who spent less than 7 minutes or more than 47 minutes on the experiment. The effective sample size was then 3420 participants . The average duration of the experiment for that sample was 19 minutes. 3.4 Evaluation of the experiment by the participants In general, participants considered the survey to be easy (mean score of 2.7 on a 1–10 difficulty scale), and interesting (mean score of 8.1 on a 1–10 interest scale). 3.5 Socio-demographics There were 1039 participants from Bulgaria, 1203 from Germany and 1178 from Spain. There was a nearly equal representation of different age groups from 18 to 65. The sample was also balanced in terms of gender. All education levels and location types were also well represented. Samples in each country differed slightly in terms of their composition. For example, compared to Bulgaria and Spain, German participants were slightly less educated and more likely to live in a village. Participants from Bulgaria were more likely to live in big cities. In the following analyses, we systematically test the robustness of our results to socio-demographic heterogeneity in our sample. In particular, we focus on vulnerable participants, defined as those who are older (above 55), less educated (primary and high school education), with lower income 5 or who report being in fairly difficult or very difficult financial situation, and who live in rural settings (in villages or the countryside). 3.6 Experience with digital devices, shopping online and product labels About 97% of participants reported owning a smartphone, and 96% using it to access the Internet. 73% of participants had scanned a QR code on a food product; the latter number was higher in Spain (83%) than in Germany (69%) or Bulgaria (66%). 80% said they would be either very likely or quite likely to scan a QR code on a food product in the future. This percentage was highest in Spain and lowest in Germany. 82% of participants said they really liked or liked the idea of having QR code on food products. Again, this percentage was lowest in Germany. We also asked participants what they thought would be the main benefits and drawbacks of having QR codes on food products. In terms of benefits, 57% of participants mentioned being able to get more information on food products, and 51% mentioned easier access to that additional information. In terms of drawbacks, 34% of participants mentioned the time and effort to scan QR codes, and 33% mentioned the issue of having to go on the Internet to access that information. Spanish participants were more likely than others to shop for groceries online at least sometimes (63%, compared to 43% in Bulgaria and 42% in Germany). Most participants claimed they “always” or “frequently” read food labels. They also expressed relatively high trust in those labels (mean of 7 on a 1–10 trustworthiness scale). 4 Main results In this section, we report the analyses of the choices, in particular related to the use of DFI. 4.1 Preference for or against DFI In this subsection, we test whether participants preferred products with DFI or products without DFI, when both types were available. For each participant, we computed the average frequency with which they chose the product on the left of the screen in the four information conditions, namely: ● Hybrid-Hybrid: both products have hybrid labels. ● Hybrid-Paper: only the label on the left is hybrid. ● Paper-Hybrid: only the label on the right is hybrid. ● Paper-Paper: both products have paper labels. Figure 3 shows the average and 95% confidence interval for this statistic depending on the information condition across all participants. The graph shows average and 95% confidence intervals. We found that in the “Hybrid-Paper” condition, the Hybrid product was chosen 50.3% of the time, while in the “Paper-Hybrid” condition, the Paper product was chosen 49.3% of the time. The one percentage point difference was not statistically significant. Similarly, in the “Hybrid-Hybrid” condition the Hybrid product on the left was chosen 49.3% of the time, while it was chosen 50.3% of the time in the “Hybrid-Paper” condition. Again, this small difference was not significant. Other comparisons did not yield significant differences either. H1.1 is not verified. Overall, the likelihood of choosing a product did not depend on whether some information was only available via a QR code. We further investigated whether at least some participants always avoided products with DFI. However, the number of participants who never chose products with DFI was not significantly higher than what one would expect to happen by chance if they were in fact indifferent (for an indifferent participant facing k choices featuring one product with DFI each, the probability of not choosing any of them is \(\:{2}^{-k}\) ). The same was true of the number of participants who always chose products with DFI. Product differences We further ran fixed-effects (within-individual) panel regressions, whereby we considered how labels affect choice while controlling for characteristics of the product on the left and on the right, for product category and for the type of information in the QR code. Column 1 in Table D1, Appendix D, confirms that choice is not affected by the type of label on the product on the right. Column 2 controls for the order in which products were presented, and for differences between products other than their labels. This includes not only the product characteristics but also whether the consumer knew or purchased either product before. We found that consumers were more likely to choose the Hybrid product in later choices (“Choice order”, + 0.001 per choice period, p < 0.05). A product was more likely to be chosen if it had been known by the consumer while the other product had not (“diff_known”, + 0.181, p < 0.001), if it had a higher weight (“diff_weight”, + 0.0003 per gram, p < 0.001), or if it had a later expiry date (“diff_date”, + 0.001 per day, p < 0.001). Similarly, it was less likely to be chosen if it was more expensive (“diff_price”), -0.109, p < 0.001), or more calorific (“diff in kcal/100g”, -0.0002 per kcal, p < 0.001). Results in terms of preference for the product with a QR code are not affected by controlling for those variables. H1.2 is thus not verified. We furthermore test if the likelihood to choose Hybrid products depended on the category the product was in, and on the type of food information accessible through the QR code (column 3). We found no such dependence on the category and the food information. Finally, we tested if differences in weight, calories and best before date mattered less if that information was shown via a QR code for the hybrid product (column 4). This would be the case if consumers did not access the QR codes and therefore could not judge the difference between the two products. We find that impact of differences in weight and calories are similar whether those differences are shown digitally or on label, but the impact of differences in the best before date is lower if it is shown digitally. Individual differences We also test whether there were differences in the likelihood to choose DFI products across different socio-demographic groups, focusing on vulnerable groups (column 5). We found that ● Germans were more likely to favour hybrid labels, though this was only marginally significant (+ 0.014, p 55 years) were less likely to choose hybrid labels (-0.017, p < 0.05). ● Less educated individuals were also less likely to choose hybrid labels, but again this was marginally significant (-0.012, p < 0.1). ● Place of residence, income or financial situation do not affect the likelihood to choose a product with DFI. Discussion We expected that consumers would be less likely to choose a product with a Hybrid label. Indeed, as we will see in the following sections, QR code scanning frequency is relatively low, choice takes longer if there is a QR code, and consumers end up knowing less about products that have a Hybrid label. This should have resulted in a lower likelihood to choose a product with a Hybrid label. However, this is not the case in our experiment. Some specificities of our design can explain this result. First, paper label and hybrid labels contained the same type of information, the only difference was how easy it was to access it Second, consumers could infer form the experimental set-up that the decision to adopt a Hybrid label was not an attempt by the firm to hide information about the product. Results could be different, therefore, if information provided via QR codes was additional, not available for other products, or if firms were free to choose which information to show on label. 4.2 Frequency of access to DFI On average, participants scanned QR codes only 23% of the time. If both products in a pair had a QR code, then participants scanned both 20% of the time, and only one 5% of the time (Table 1 ). H2.1 is verified. Table 1 Number of QR codes scanned, by information condition. Information condition Number of QR codes scanned One hybrid label Both hybrid labels 0 76.86% 75.26% 1 23.14% 4.83% 2 NA 19.90% N 27360 13680 Participants were more likely to scan a QR code the first time they encountered one. Indeed, as many as 46% scanned the first QR code they encountered. The QR scanning frequency then dropped and stabilized at 20%. H2.2 is verified. Product differences We ran panel regressions to find determinants of the likelihood to open a QR code (Table D2, Appendix D). We found that, compared to hybrid-paper and paper-hybrid situations, participants were more likely to scan at least one QR code if both products had hybrid labels (variable “both hybrid”, + 1.6%, p < 0.01) (column 1). As mentioned previously, we also found they were becoming less likely to scan a QR code over time (variable “order”, column 2). They were also less likely to open a QR code if there was a large difference in price or in weight between the two products in the pair, possibly because their choice could then be made primarily on that basis (variables “diff_weight”, “diff_price”, column 2). We also found that the likelihood to scan a QR code did not depend on the food category (column 3). H7 is thus not verified in terms of DFI access. However, the likelihood depended on the type of information in the QR code. Participants were most likely to scan a QR code giving access to nutrition information (32%), followed by quantity information (25%), and best before date information (22%). Those differences are statistically significant (column 3). H6 is thus verified. Individual differences When considering differences in behaviour across participants, we found that 37% of the participants never scanned any QR code across all the 16 choices they had to make. An additional 17% scanned a QR code only once. Only 4% of the total scanned QR codes across all choices they made. Finally, we considered the impact of socio-demographic characteristics of participants (column 4, Table D2, Appendix D). We found that Spanish participants were more likely to scan QR codes (+ 3.6%, p < 0.01) compared to the base category of Bulgarians. Vulnerable participants were not statistically different from the rest of the sample, regardless of whether vulnerability was defined in terms of age, education, income, or place of residence. H5 is thus not verified in terms of DFI access frequency. Discussion : Rates of QR code scanning in our experiment were quite high compared to rates observed in previous studies we discussed before. This is what we aimed for: we made QR code scanning as easy as possible to be able to identify the effect of scanning QR code on behaviour. Our setting was indeed a best-case scenario for the use of QR code, whereby QR code scanning meant only clicking on the screen, and participants only had 16 choices to make, of which 12 where at least one of the products had a QR code. Yet, participants made that effort only 24% of the time. This shows that delivering food information by that channel cannot possibly ensure participants are exposed to it. Our experiment allowed us to identify interesting differences between participants. The 37% who never scanned any QR codes were also those who made the fastest choices; in other words, not scanning QR codes saved time. This was at the expense of knowing less about products, as we will see. 4.3 Speed of choice We measured how long participants took to make choices for each of the 16 choice situations they faced. We also measured how long they took to access information via the QR code. We report statistics after eliminating outliers, namely the 1% of participants who were the slowest (taking more than 2 minutes and 46 seconds to make choices). This makes our comparisons of statistics across situations more robust, as such outliers may not be equally distributed in our sample and may thus unduly influence our statistics. We also disregarded the last three choices, as participants also had to report information about their chosen product for those choices, which was included in their choice time. We found that participants took 16.9 seconds on average to make a choice and spent 3.6 seconds accessing DFI if they clicked on the QR code. Choice became faster as participants progressed through choice: the first choice took 26.8 seconds on average, while the 13th choice took only 14.8 seconds on average. We found that mean decision times were longer if there were some QR codes to be opened. Choice took 15.9 seconds on average when there were no QR codes (“Both Paper”), 16.9 seconds if one product in the pair had a QR code (“One Hybrid”), and 17.5 seconds if both products in the pair had QR codes (“Both Hybrid”) (Table 2 ). Those differences are statistically significant (Table D3, Appendix D). H3.2 is thus verified. Table 2 Average choice speed, in seconds, depending on the information condition. Information condition Both Paper One Hybrid Both Hybrid mean 15.9 16.9 17.5 sd 12.3 11.8 13.6 N 3420 3420 3420 These differences in total time spent making a choice are consistent with the likelihood to access DFIs and time spent accessing them. Indeed, participants opened about 25% of the QR codes and spent about 4 seconds reading the information in them, so this would explain why time spent making a choice is about 25% times 4 seconds = 1 seconds more when there is a QR code. We now consider how time spent making a choice varied depending on whether a QR code was opened or not. Choices took 14.6 seconds if one product had a QR code and it was not opened, compared to 26.4 seconds if it was opened. Similarly, choices took 14.3 seconds if two products had a QR code and they were not opened, vs. 21.6 seconds if one QR code was opened, and 31.1 seconds if both were opened. Regressions (not shown) confirm those differences are significant. H3.1 is thus verified. The increase in time spent making a choice when there were QR codes, and they were opened was greater than could be explained simply by the time spent accessing the DFIs (we saw this was about 4 seconds for each QR code). There is therefore likely to be a selection effect, whereby people not opening QR codes were also those who made fast choices. Later in this section we show regressions intended to isolate the selection effects from the loss in speed that is due to QR codes. We also tested for a selection effect by relating average individual time spent making a choice in menus without QR codes and the individual likelihood to open QR codes when some are available. We found such a relation, whereby for example participants who never opened QR codes also were the fastest in making choices when there were no QR codes, taking 12.6 seconds, while those who always opened QR codes took an average of 24.5 seconds in the same situation. This points to the existence of different types of participants: those who try to make fast choices at the expense of taking less information into account, and those who try to make more considered choices at the expense of spending more time gathering information (Kahneman, 2011 ). Individual differences Given how differences in choice speed depending on information conditions may also be linked to individual differences, we run fixed effect (“within”) panel regressions to take out variability in choice speed across individuals and focus on differences in choice speed across choice situation for the same individual. Indeed, in fixed effects regressions, we computed the mean choice speed of each individual, and then considered deviations from that speed depending on the information condition (Table D3, Appendix D).2F 6 We found that hybrid labels did slow choice down even when controlling for individual differences. Compared to the paper-paper condition, choices were 0.882 seconds slower on average when there was one hybrid label and 1.625 seconds slower when there were two (column 1, Table D3, Appendix D). This confirms H3.2. We further controlled for choice order (column 2) and found that individuals made faster choice over time. Choice was also faster if only one product was known (“abs_diff_known”), when there was a larger price difference (“abs_diff_price”), or when there was a large difference in best before dates (“abs_diff_date”). This maybe be due to easier application of simple heuristics in those cases, such as going for the known product, the cheaper one or the one that will last longer. Indeed, if differences on one dimension are large, then the buyer can more easily spare time considering other aspects since they are not likely to change the decision. In terms of differences in product categories and type of information in QR codes (column 3), we found that choice for vegetable products was fastest, and that choice was slowest when the QR code contained nutritional information. This may result from consumers’ previously mentioned greater propensity to access the DFI in the choices in which it contained this type of information. Finally, we consider the impact of individual difference on choice speed (column 4), and how the presence of a QR code on one or both products impacted this speed differently for vulnerable participants (column 5). We found that German consumers were the fastest in making choices, followed by the Spanish, whereas Bulgarians were the slowest. Older participants (those above 55) took 5.6 seconds more to make a choice, while those with low income took 1.3 seconds less (column 4). We also see that hybrid labels impacted older participants more than the rest, whereby they slowed their choice by an additional 0.953 second (“DFI & Age > 55”, column 5). Discussion This part shows that products with a QR code slowed choice down by about one second. We also saw that participants who were slow in making choices were also more likely to open QR codes, which slowed them even more. Slow decision makers were thus most affected by QR codes. This may be because they were those who tried to make considered choices, which involves getting all information even if behind a QR code. Unlike what could have been expected, slow decision makers did not refrain from opening QR codes (to avoid being slowed down even more). Rather, they proceed to open QR codes even at the cost of further slowdown. Even considering differences in choice speed across participants, we found that participants were slower in making choices when there was a QR code on one or both of the products. Their choice was 0.886 second slower when one product had a Hybrid label than when both products had a paper label, and 1.625 seconds slower when both products have a hybrid label. Consumers did not compensate for time spent scanning QR codes and accessing DFIs by spending less time considering other information on food labels. Rather, the time spent scanning QR codes resulted directly in an increase in time spent making choices. This finding helps predict whether choices would be slower if Hybrid labels were widely introduced in the field. We would expect QR code scanning to be much less prevalent than in our experiment, where scanning QR codes was much easier and faster than it would be in the outside world. However, time spent scanning the QR code and accessing information would also be much higher than in our experiment, for the same reasons. It is hard to put a number for the net effect, but it would likely also result in slower choice overall. The introduction of QR codes would particularly affect slow decision makers, who as we saw before, were also the most likely to want to access DFIs. They would thus lose the most time. 4.4 Product knowledge We asked participants to report their best estimate of information about products they had chosen from: the weight, the number of kilocalories per 100g, and the “best before” dates. They answered these questions for the last three choices they had to make, right after each of those three choices, for both products in the pair they had to choose from. We varied whether the information they were asked about was shown on the product label or accessible only via a QR code. We measured errors in their responses in two ways: For the weight and the number of calories, we defined the error as abs(ln(stated value / true value)). In words, we took the absolute value of the logarithm of the ratio between the value reported by the participant and the true value. The error is 0 if the participant reported the true value. This is a good way to normalise the error because for example, the error is the same, equal to ln(2) = 0.69 whether the participant reported twice the true value, or half of the true value. Those are indeed errors of analogous magnitude. We compute the error in a slightly different way when considering best before dates, by considering the number of days left before the best before date, as estimated by the participant vs. as shown to them with the best before date. 7 The error is then abs(ln(stated days left / true days left)). Average errors We found that the average error for weight was 0.39 (Table 3 , rightmost column), which is equivalent to overestimating weight by 48% or underestimating it by 32%. Average error for nutrition information (number of kilocalories per 100g) was 0.67, which is equivalent to overestimating kilocalories by 95% or underestimating it by 49%. Participants were thus significantly worse at reporting caloric content of the food than in reporting weight. This is probably because most consumers are less experienced estimating caloric content per 100g. Average error for days left was still higher, at 0.89, which is equivalent to overestimating the number of days left by 143% or underestimating it by 59%. We can speculate on the reason for this low precision; it may be that participants expected in any case that the products would not be out of date and did not care for length of conservation beyond this. Impact of the means of food information delivery Table 3 also shows how label types affected error rates. Table 3 Errors for weight, calories and best before date, depending on information condition Paper-Paper IC Paper label in Paper-Hybrid IC Hybrid label in Paper-Hybrid IC Hybrid-Hybrid IC Total Weight Mean (sd) 0.38 (0.43) 0.39 (0.40) 0.39 (0.40) 0.42 (0.43) 0.39 (0.42) N 4205 856 852 855 6768 Calories Mean (sd) 0.68 (0.63) 0.65 (0.61) 0.67 (0.62) 0.68 (0.61) 0.67 (0.63) N 5050 429 430 862 6771 Best before date Mean (sd) 0.84 (0.70) 0.92 (0.72) 0.94 (0.72) 1.05 (0.80) 0.89 (0.72 N 4170 839 841 809 6659 We show average error depending on if none, one or both products in the pair had a hybrid label. In case only one had a paper label, we show error for the product with a paper label and error for the product with a hybrid label. For weight, the mean error was higher if information was available via the QR code, increasing from 0.38 if weight was shown on paper for both products, to 0.42 if weight was shown as DFI for both products (Table 6). For calories, the error was the same if calories were both shown on paper or both shown as DFI. For the best before date, the error increased from 0.84 if “best before” dates were shown on the paper label for both products to 1.05 if “best before” dates were accessible only as DFI for both products. Regressions presented in Table D4, Appendix D, confirm H4.2. Impact of clicking the QR code Disaggregating further the statistic for the case of weight information being accessible via the QR code, we found that the error for weight was 0.44 if the person did not access the weight information in the DFI, vs. 0.26 if the person did access the DFI. This is to be compared with the error of 0.38 if the weight information was on the label. We thereby see a self-selection effect at play, whereby those who are motivated to access the DFI do make lower error than the average of those who get this information on the label. However, conversely, those who do not access the QR code make higher errors. Overall, too few consumers access the DFI, so the overall effect of putting information behind the QR code is to increase the error. For calorie information, we found that the error was 0.69 if the person did not access the DFI with the calorie information, vs. 0.62 if she did. This is to be compared with the error of 0.67 when information was on the label. The pattern is thus the same as for weights, but weaker and since more people accessed digital nutrition information than other types of information, the total effect is null. Finally, we found that the error for best before date was 1.08 if the person did not access the DFI with that information, vs. 0.85 if the person did access the DFI. This is to be compared with the error of 0.67 if the information is on the label. Regressions (not shown) confirm those differences are significant. H4.1 and H4.3 are thus verified. Overall, therefore, we found that consumers did gain in precision by accessing the QR code with the information, but also lose in precision when not accessing the QR code. The net effect of QR codes is thereby negative because too few of them click on QR codes. Individual differences We run fixed effect (“within”) panel regressions to control for differences across individuals. We normalize errors in terms of dates, weight and calories with the mean and standard deviation of the error in each of those categories. 8 Those regressions allow us to check how differences in label presentations affect the error while controlling for a possible selection effect whereby for example those who do not click on QR codes are those who already know the relevant food information. Results are shown in Table D4, Appendix D. We find that providing the information via a QR code consistently increases the error (columns 1–2). German and Spanish participants appear to make higher errors than Bulgarians. Older participants make lower errors, but those with low income make higher errors (column 3). H5 is thus verified in terms of the error. Providing information via a QR code increases the error a lot more in Bulgaria (baseline, + 0.21) than for participants in Germany or Spain, although the effect is still significant in each of those countries. The impact does not appear to differ consistently for vulnerable participants, except for those who report being financially constrained (+ 0.11) (column 4). Discussion This part on errors confirms that a consequence of participants not scanning QR codes is that they know less about the product they choose. Hybrid labels result in them knowing less about product characteristics than if all information is on a paper label. 5 Conclusions We found that most consumers did not show a preference for or against products that displayed some information through a QR code compared to products that displayed all information on printed paper labels. However, this was not the case for older and less educated consumers, who were less likely to choose products with QR codes. Consumers showed little interest in accessing information through QR codes, even though our experiment was designed to make it comparatively easy to access digital food information. As many as 37% of the participants in the experiment never scanned any QR code over all the choices they made in our experiment. Only 4% scanned all of them. Overall, participants scanned QR codes in only 24% of the cases when one was shown on product labels. Providing food information via QR codes had a negative impact on consumers in two respects. First, scanning QR codes slowed choice down. Participants took longer making choices when a product had a QR code. This was especially the case among older participants. This slowdown was proportional to the time spent accessing the digital food information and to the likelihood of accessing it. This means that consumers did not compensate for the time they spent scanning QR codes by accelerating other aspects of their choice process. Second, participants were less precise in their knowledge of product information if it was shown via a QR code than if it was shown printed on the labels. Indeed, consumers made more mistakes when reporting information about products with digital food information. They were not able to correctly report product information when they did not scan the QR codes. This means that consumers who did not scan QR codes did not do so because they already knew the information therein. Rather, they simply chose to forgo knowledge about the product. In summary, we found that consumers were unlikely to access information provided through QR codes. Giving access to information via QR codes slowed choice down. Providing food information digitally also resulted in consumers knowing less about the food products they had chosen from. However, consumers did not respond to this by avoiding products with QR codes. This study illustrates that printed paper labels are more effective than QR codes in giving access to food information. Labels that require digital means to access food information are not a good substitute for printed paper labels that give direct access to food information. This is due to the time spent scanning QR codes, the low likelihood to scan them, and the resulting lack of knowledge about the characteristics of the products. These negative effects of the use of QR codes emerged even though our setup made it particularly easy to access digital information. We note however that our study did not aim to investigate possible benefits of QR codes when used as a complement, rather than as a substitute for information printed on labels. For example, QR codes could provide more in-depth information than could fit when printed on a label. Furthermore, changes in technology and in consumers’ habits could affect the potential of digital means of access to food information. Funding Declaration: This work was funded by the European Commission. Clinical trial number: not applicable. Human Ethics and Consent to Participate declarations: see attached ethics committee approval Declarations Funding Declaration: This work was funded by the European Commission. Clinical trial number: not applicable. Human Ethics and Consent to Participate declarations: see attached ethics committee approval Author Contribution AG took part in all the stages of this project.MK consulted the design and data analysis, supervised development of stimuli, co-edited the text. Data Availability The wording of all the stimuli, the data, analytic code, and further supporting materials are also available at https://osf.io/nwv38/. References Bashir, H. (2022). Leveraging technology to communicate sustainability-related product information: Evidence from the field. Journal of Cleaner Production , 362 , 132508. Bettman, J. R., Johnson, E. J., & Payne, J. W. (1990). A componential analysis of cognitive effort in choice. Organizational behavior and human decision processes , 45 (1), 111–139. Bradford, H., McKernan, C., Elliott, C., & Dean, M. (2022). Consumer purchase intention towards a quick response (QR) code for antibiotic information: an exploratory study. npj Science of Food , 6 (1), 23. Bray, J., Hartwell, H., Price, S., Viglia, G., Kapuściński, G., Appleton, K., & Mavridis, I. (2019). Food information presentation: consumer preferences when eating out. British Food Journal , 121 (8), 1744–1762. Clinton, V. (2019). Reading from paper compared to screens: A systematic review and meta-analysis. Journal of research in reading , 42 (2), 288–325. Dickinson, D. L., & Kakoschke, N. (2021). Seeking confirmation? Biased information search and deliberation in the food domain. Food quality and preference , 91 , 104189. Fenko, A., Nicolaas, I., & Galetzka, M. (2018). Does attention to health labels predict a healthy food choice? An eye-tracking study. Food quality and preference , 69 , 57–65. Francis, J., Ball, C., Kadylak, T., & Cotten, S. R. (2019). Aging in the digital age: Conceptualizing technology adoption and digital inequalities. In Ageing and digital technology: Designing and evaluating emerging technologies for older adults, 35–49. FoodDrinkEurope (2024, January 17). Making consumer information accessible to all . https://www.linkedin.com/pulse/making-consumer-information-accessible-all-fooddrinkeurope-nmwde/ Garbarino, E. C., & Edell, J. A. (1997). Cognitive Effort, Affect, and Choice. Journal of Consumer Research , 24 (2), 147–158. Glöckner, A., & Betsch, T. (2012). Decisions beyond boundaries: When more information is processed faster than less. Acta psychologica , 139 (3), 532–542. Golman, R., Hagmann, D., & Loewenstein, G. (2017). Information avoidance. Journal of economic literature , 55 (1), 96–135. Jha, S., Balaji, M. S., Stafford, M. B. R., & Spears, N. (2020). Haptic information processing and need for touch in product evaluation. Journal of Consumer Marketing , 37 (1), 55–64. Kahneman, D. (2011). Thinking, Fast and Slow . Macmillan. Kang, M. J., Hsu, M., Krajbich, I. M., Loewenstein, G., McClure, S. M., Wang, J. T. Y., & Camerer, C. F. (2009). The wick in the candle of learning: Epistemic curiosity activates reward circuitry and enhances memory. Psychological science , 20 (8), 963–973. Kim, Y. G., & Woo, E. (2016). Consumer acceptance of a quick response (QR) code for the food traceability system: Application of an extended technology acceptance model (TAM). Food Research International , 85 , 266–272. Lacroix, A., Muller, L., & Ruffieux, B. (2019). Labeling for Sustainable Food: The Consumer's Point of View. In V. Clavier, & De J. P. Oliveira (Eds.), Food and Health . Actor Strategies in Information and Communication, Wiley. Li, T., & Messer, K. D. (2019). To Scan or Not to Scan: The Question of Consumer Behavior and QR Codes on Food Packages. Journal of Agricultural and Resource Economics , 44 (2), 311–327. Lombardi, A., Carfora, V., Cicia, G., Del Giudice, T., Lombardi, P., & Panico, T. (2017). Exploring willingness to pay for QR code labeled extra-virgin olive oil: An application of the theory of planned behavior. International Journal on Food System Dynamics , 8 (1), 14–31. Ma, G., & Zhuang, X. (2021). Nutrition label processing in the past 10 years: Contributions from eye tracking approach. Appetite , 156 , 104859. McFarlane, T., & Pliner, P. (1997). Increasing willingness to taste novel foods: effects of nutrition and taste information. Appetite , 28 (3), 227–238. Mendelson, J., & Romano Bergstrom, J. C. (2013). Age differences in the knowledge and usage of QR codes. In Universal Access in Human-Computer Interaction. User and Context Diversity: 7th International Conference, UAHCI 2013, Held as Part of HCI International 2013, Las Vegas, NV, USA, July 21–26, 2013, Proceedings, Part II 7 (pp. 156–161). Springer Berlin Heidelberg. Neijens, P. C., & Voorveld, H. A. (2018). Digital replica editions versus printed newspapers: Different reading styles? Different recall? New media & society , 20 (2), 760–776. Nosek, B. A., Ebersole, C. R., DeHaven, A. C., & Mellor, D. T. (2018). The preregistration revolution. Proceedings of the National Academy of Sciences, 115(11), 2600–2606. Oonk, L. (2013). QR codes, quick response or quick rejection: A study about the contribution of the phenomenon QR codes on food products, on the intention to seek information and the purchase intention. Master’s thesis, University of Twente. Orquin, J. L., Bagger, M. P., Lahm, E. S., Grunert, K. G., & Scholderer, J. (2020). The visual ecology of product packaging and its effects on consumer attention. Journal of Business Research , 111 , 187–195. Øvrum, A., Alfnes, F., Almli, V. L., & Rickertsen, K. (2012). Health information and diet choices: Results from a cheese experiment. Food Policy , 37 (5), 520–529. Ozkaya, E., Ozkaya, H. E., Roxas, J., Bryant, F., & Whitson, D. (2015). Factors affecting consumer usage of QR codes. Journal of Direct Data and Digital Marketing Practice , 16 , 209–224. QualityChain (2024, January 16). We just can’t put everything on the label. https://www.linkedin.com/pulse/we-just-cant-put-everything-label-qualitychain-qhysf Riddle, D. R. (Ed.). (2007). Brain Aging Models, Methods, and Mechanisms . Routledge. Roberto, C. A., & Khandpur, N. (2014). Improving the design of nutrition labels to promote healthy food choices and reasonable portion sizes. International Journal of Obesity , 38 (1), S25–S33. Temple, N. J., & Fraser, J. (2014). Food labels: a critical assessment. Nutrition (Burbank, Los Angeles County, Calif.) , 30 (3), 257–260. van der Laan, L. N., & Orcholska, O. (2022). Effects of digital Just-In-Time nudges on healthy food choice – A field experiment. Food Quality and Preference , 98 , 104535. https://doi.org/10.1016/j.foodqual.2022.104535 Wansink, B., & Sobel, J. (2007). Hidden persuaders and 200 daily decisions. Environ Behav , 39 (1), 106–123. Zentall, T. R., & Stagner, J. P. (2012). Do pigeons prefer information in the absence of differential reinforcement? Learning & behavior , 40 , 465–475. Footnotes An alternative to our choice to vary presentation of MFI would have been to consider digital delivery of voluntary food information, such as measures of a product’s Environmental, Social and Governance dimensions, or nutritional aspects of the food that are not yet generally provided (such as their Nutriscore and/or alternatives and variations thereof). We decided not to do so because that type of information is not available for all products, and consumers are not yet knowledgeable about their meaning and importance. For practical and budgetary reasons, this was implemented for 25% of randomly suggested participants only (and they were aware of this). We asked this only for the last three choices to make the first 14 choices as close to normal choice as possible. In the last two choices, participants may anticipate they will be asked to recall information they consulted, which may affect their choice and information search. Postal addresses were collected at the end of the experiment from those participants who won the lottery. However, we made sure at the beginning of the experiment that all participants were ready to give their address in case of a win. This guarantees that all participants participated with the knowledge that one of their choices would be implemented. Monthly household income less than 1400 LEV in Bulgaria, less than 2400€ in Germany, less than 1500€ in Spain. Those thresholds correspond to the bottom 40% in terms of household income in our sample in each country. We need to do so because as we saw, slower individuals are also more likely to open QR codes, which would explain why choice is slower when the QR code is opened. Unlike weight and calories, which have a lowest value of 0, dates do not have such a clear and meaningful lowest value. We therefore use the date at which a participant made their choice as the reference point. For example, the normalized error for weight is (error - mean error for weight)/std dev. of the error for weight. Additional Declarations No competing interests reported. Supplementary Files DFIappendices.docx Cite Share Download PDF Status: Under Revision Version 1 posted Editorial decision: Revision requested 03 Mar, 2026 Reviews received at journal 02 Mar, 2026 Reviews received at journal 02 Mar, 2026 Reviewers agreed at journal 03 Feb, 2026 Reviewers agreed at journal 02 Feb, 2026 Reviewers invited by journal 29 Jan, 2026 Editor assigned by journal 27 Jan, 2026 Submission checks completed at journal 15 Jan, 2026 First submitted to journal 08 Jan, 2026 You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-8549197","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":583246830,"identity":"a9765941-050c-41f8-ba5b-e21f8a398ce8","order_by":0,"name":"Alexia Gaudeul","email":"","orcid":"","institution":"Joint Research Center","correspondingAuthor":false,"prefix":"","firstName":"Alexia","middleName":"","lastName":"Gaudeul","suffix":""},{"id":583246831,"identity":"bb4a48f9-8b31-486e-8714-df4bd822532b","order_by":1,"name":"Michal Krawczyk","email":"data:image/png;base64,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","orcid":"","institution":"Joint Research Center","correspondingAuthor":true,"prefix":"","firstName":"Michal","middleName":"","lastName":"Krawczyk","suffix":""}],"badges":[],"createdAt":"2026-01-08 08:54:04","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-8549197/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-8549197/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":101880650,"identity":"55e6f57b-7474-48c5-942c-7d3e555daab2","added_by":"auto","created_at":"2026-02-04 15:04:42","extension":"jpg","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":128650,"visible":true,"origin":"","legend":"\u003cp\u003eAn example of a choice screen presented to participants.\u003c/p\u003e","description":"","filename":"1.jpg","url":"https://assets-eu.researchsquare.com/files/rs-8549197/v1/c526c09c40ac2319e0a5adc2.jpg"},{"id":101787020,"identity":"0a3f065a-c1e5-4184-907d-dde036f71312","added_by":"auto","created_at":"2026-02-03 15:45:16","extension":"jpg","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":83435,"visible":true,"origin":"","legend":"\u003cp\u003eTimeline of the experiment\u003c/p\u003e","description":"","filename":"2.jpg","url":"https://assets-eu.researchsquare.com/files/rs-8549197/v1/27f3c9fe3afae797d48ff538.jpg"},{"id":101787018,"identity":"2145265d-43e3-4c5f-914a-ea840ea02e2b","added_by":"auto","created_at":"2026-02-03 15:45:14","extension":"jpg","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":69099,"visible":true,"origin":"","legend":"\u003cp\u003eChoice of product on the left in the pair, labelled by information condition.\u003c/p\u003e\n\u003cp\u003eThe graph shows average and 95% confidence intervals.\u003c/p\u003e","description":"","filename":"3.jpg","url":"https://assets-eu.researchsquare.com/files/rs-8549197/v1/5c391bce13a9aac999eb0c94.jpg"},{"id":102298654,"identity":"9c4dcf6f-97b2-4ae7-8710-e878cfc94ab2","added_by":"auto","created_at":"2026-02-10 10:56:41","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1611943,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8549197/v1/1a1dc19a-913c-44ad-9f7a-b473e111ecad.pdf"},{"id":101787016,"identity":"4ad60950-c42f-4a28-8d62-71a8e482cd86","added_by":"auto","created_at":"2026-02-03 15:45:14","extension":"docx","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":67798,"visible":true,"origin":"","legend":"","description":"","filename":"DFIappendices.docx","url":"https://assets-eu.researchsquare.com/files/rs-8549197/v1/d5f185e7ceb5b25d973b21e2.docx"}],"financialInterests":"No competing interests reported.","formattedTitle":"\u003cp\u003eUsing QR codes to access food information: A behavioural study with European consumers\u003c/p\u003e","fulltext":[{"header":"1 Introduction","content":"\u003cp\u003eFood consumers should be able to access comprehensive information concerning the products they buy in a quick and convenient manner. However, this can be particularly challenging for products with limited label space and in cases when customers may expect more extensive information, for example about the origin of certain foods (Lombardi et al, \u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e2017\u003c/span\u003e), the manufacturing process (Bradford et al, \u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e2022\u003c/span\u003e), or producer\u0026rsquo;s sustainability practices (Kim and Woo, \u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e2016\u003c/span\u003e). Firms may also want the consumers to be able to access Internet resources allowing audio, video, interactive, and personalized content (including targeted offers). One way to do so involves the use of hybrid labels whereby only some information is written on the label while a Quick Response (QR) code directs to additional online resources. Those QR codes can be scanned while purchasing a product (van der Laan \u0026amp; Orcholska, \u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e2022\u003c/span\u003e) or later on during preparation or consumption. The question we investigate in this paper is whether information available in such a way is accessed, perceived, processed, and retrieved in a similar fashion to that available directly.\u003c/p\u003e \u003cp\u003eOur objective here is to pursue a systematic experimental investigation to address these issues. We do not only want to understand whether consumers prefer products where food information is provided on-label rather than products that offer the same information digitally, but also how often they \u003cem\u003eaccess\u003c/em\u003e the digital information, whether it slows down their choice, and whether it affects the accuracy of their knowledge of food information.\u003c/p\u003e \u003cp\u003eThese questions are important for measuring the effects of regulations concerning provision of food information on the welfare of consumers. While corporations underline the benefits of digitalization of food labels, allowing them to cut the costs of packaging, attract consumers to their websites, strengthen brand loyalty and personalize offers (FoodDrinkEurope, 2024; QualityChain, 2024), the consequences for the consumers must be carefully examined. Our study thus contributes to the important discussion on the consequences of using digital means to convey food information, and more specifically, of the effects of splitting such information across communication channels.\u003c/p\u003e \u003cdiv id=\"Sec2\" class=\"Section2\"\u003e \u003ch2\u003e1.1 Labels and food choice\u003c/h2\u003e \u003cp\u003eLabels are a key source of information about food, but consumers often find them confusing (Roberto and Khandpur, \u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e2014\u003c/span\u003e; Temple and Fraser, \u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e2014\u003c/span\u003e). Full and correct processing of the information they provide cannot be taken for granted. On average, consumers may be making up to 200 food decisions a day (Wansink and Sobel, \u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e2007\u003c/span\u003e), so most of those decisions must be made very quickly. They cannot consider all information on labels at the time of purchase.\u003c/p\u003e \u003cp\u003eFood choices thus typically follow the \u0026ldquo;direct heuristic route\u0026rdquo; corresponding to rough, error-prone \u0026ldquo;system 1\u0026rdquo;, as found in eye-tracking research (Ma and Zhuang, \u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e2021\u003c/span\u003e). So-called top-down attention, driven by consciously formulated goals and plans, plays a limited role in food choice (Fenko et al., \u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e2018\u003c/span\u003e). Bottom-up attention, which is driven by external stimuli, is a more important factor.\u003c/p\u003e \u003cp\u003eSome studies find that labels have a limited effect on actual food choice (see Ma and Zhuang, \u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e2021\u003c/span\u003e for a review). However, they may still affect consumers\u0026rsquo; willingness to try new foods (McFarlane and Pliner, \u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e1997\u003c/span\u003e) and to pay for dietary products (\u0026Oslash;vrum et al., \u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e2012\u003c/span\u003e). Some studies (Orquin et al., \u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e2020\u003c/span\u003e) find that the capacity to affect choices depends on such qualities as size, prominence, and salience of a label.\u003c/p\u003e \u003cp\u003eWe therefore must take account not only of the fact that providing food information via a QR code affects the likelihood to access this information, but also how this mode of delivery affects the choice process. For example, digital information may be accessed only after deliberation and consideration of other information on the label, rather than automatically. Furthermore, digital information that is shown separately (on one\u0026rsquo;s smartphone for example), may have a different impact than information that is shown along other information on the same support. Finally, consumers may prefer products with directly accessible information, or conversely prefer labels with less information. All this may result in different product choices.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003e1.2 Research questions and hypotheses\u003c/h2\u003e \u003cp\u003eThis experimental behavioural study measured participants\u0026rsquo; preferences as well as their access to, and knowledge of food information presented either via a QR code or directly on-label. To improve control over the information participants could access, the experiment featured and compared labels containing no QR code (henceforth \u0026ldquo;paper labels\u0026rdquo;) to those with one type of information (e.g. caloric contents) accessible via QR code and other types of information available directly (henceforth \u0026ldquo;hybrid labels\u0026rdquo;, with the part accessible via QR code referred to as Digital Food Information, DFI).\u003c/p\u003e \u003cp\u003eThis setup allowed us to address seven important inter-related research questions and associated hypotheses.\u003c/p\u003e \u003cp\u003eOur first question is \u003cb\u003e1) whether products with paper labels are preferred to products with hybrid labels\u003c/b\u003e (or the other way round), other things being equal. In general, people tend to choose options that are easier to evaluate (Garbarino and Edell, \u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e1997\u003c/span\u003e). As paper labels provide all the information in the same manner, they also probably make it easier to compare and thus evaluate products with respect to each other. However, empirical evidence on QR codes is very scarce in this respect. Oonk (\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e2013\u003c/span\u003e) found that purchase intentions were independent of whether information was provided digitally or on-label. We test the following hypotheses:\u003c/p\u003e \u003cp\u003e \u003cul\u003e \u003cli\u003e \u003cp\u003eH1.1: Participants will be more likely to choose a product with a paper label when given a choice between a product with a paper label and one with a hybrid label than when given a choice between two products with paper labels.\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003eH1.2: Participants will be more likely to choose a given product when it is displayed with a paper label than the same product when it is displayed with a hybrid label.\u003c/p\u003e \u003c/li\u003e \u003c/ul\u003e \u003c/p\u003e \u003cp\u003eHow QR codes affect choices clearly depends on whether \u003cb\u003e2) consumers perceive information they open as sufficiently valuable to access it\u003c/b\u003e. The pragmatic theory of information sees it as a means or resource to solve a problem by overcoming uncertainty. Consumers are thus only expected to access information if it is perceived as sufficiently likely to affect their choice of product, given the costs of accessing it. These costs involve, in the case of QR codes, the effort and time of picking up one\u0026rsquo;s smartphone, opening the camera or a dedicated app, scanning the QR code, waiting for the information to load, and reading it.\u003c/p\u003e \u003cp\u003eThen again, there is ample empirical evidence that such an instrumental view of information is overly simplistic, as humans (and other animals) are often willing to acquire information also when it has no pragmatic value (Kang et al., \u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e2009\u003c/span\u003e; Zentall and Stagner, \u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e2012\u003c/span\u003e). It could be motivated by sheer curiosity or willingness to have one\u0026rsquo;s opinions confirmed. Sometimes people may actively \u003cem\u003eavoid\u003c/em\u003e obtaining information despite its positive pragmatic value (Golman et al., \u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e2017\u003c/span\u003e). In the context of food choice, for example, they may prefer not to know how many calories their favourite desert has.\u003c/p\u003e \u003cp\u003eEmpirically, most studies find spontaneous use of QR codes to be low (Li \u0026amp; Messer, \u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e2019\u003c/span\u003e; Bashir, \u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e2022\u003c/span\u003e) unless a dedicated device is readily available (Li \u0026amp; Messer, \u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e2019\u003c/span\u003e). Relatedly, Bray et al., (\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e2019\u003c/span\u003e) reported that consumers tend to prefer on-label information when given a choice. Li \u0026amp; Messer (\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e2019\u003c/span\u003e) showed how dramatically willingness to use DFI depends on how it can be accessed. In the scenario most comparable to ours, involving a link on a tablet provided by the experimenters, 20.2% of participants accessed it. The following hypotheses emerge:\u003c/p\u003e \u003cp\u003e \u003cul\u003e \u003cli\u003e \u003cp\u003eH2.1: The prevalence of DFI access will be low. DFI will be accessed less than one fourth of the time in our experiment on average.\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003eH2.2: The prevalence of access to DFI will diminish over the course of the experiment.\u003c/p\u003e \u003c/li\u003e \u003c/ul\u003e \u003c/p\u003e \u003cp\u003eAnother welfare-relevant dimension is \u003cb\u003e3) the speed of choice\u003c/b\u003e. Holding the quality of the decision constant, information channels that lead to quicker decisions should be preferred. We are not aware of directly relevant extant research, but it seems natural to expect choices involving products with hybrid labels to take more time than choices involving only products with paper labels, given the extra time (if short) necessary to access DFI. Moreover, the Elementary Information Processes perspective (Bettman et al, \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e1990\u003c/span\u003e) predicts that with more information being processed, the processing time becomes longer, so that we expect that choices in which consumer accessed DFI to be slower than those in which she could access it but did not. It should be noted, however that later literature nuanced this picture, showing that additional information may, in fact, speed up decisions if it improves coherence, i.e. additional information promotes the option the consumer would be inclined to choose without this information already (Gl\u0026ouml;ckner and Betsch, \u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e2012\u003c/span\u003e). The following hypotheses can thus be put forward.\u003c/p\u003e \u003cp\u003e \u003cul\u003e \u003cli\u003e \u003cp\u003eH3.1: Among choice situations with at least one product with a hybrid label, choices in which DFI is accessed will take more time than those in which it is not.\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003eH3.2: Choice involving one or two products with hybrid labels will take more time than choices involving only products with paper labels.\u003c/p\u003e \u003c/li\u003e \u003c/ul\u003e \u003c/p\u003e \u003cp\u003eWe also investigate \u003cb\u003e4) product knowledge\u003c/b\u003e. Naturally, if DFI is rarely even accessed, it cannot be expected to be known, which would result in inferior product knowledge in case of hybrid labels. However, one may argue that participants who care about a given type of information would likely be willing to access it via a QR code, and those who do not, would not read it even if it was printed on the label, so that overall, DFI would not result in inferior product knowledge. One could even go so far as to say that by requiring more active pursuit of information, QR codes may make the information more memorable to those who sought it. There is also some evidence of small advantage of reading from paper rather than the screen (Neijens \u0026amp; Voorveld, \u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e2018\u003c/span\u003e; Clinton, \u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e2019\u003c/span\u003e) but the mechanism and, therefore, border conditions for this effect are poorly understood. We test the following hypotheses (in which, since pre-registration, we have slightly changed potentially misleading wording):\u003c/p\u003e \u003cp\u003e \u003cul\u003e \u003cli\u003e \u003cp\u003eH4.1: Responding to knowledge questions, participants will be more precise if they accessed information than if they did not.\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003eH4.2: Participants will be more precise in their knowledge of information that was provided on the label than of DFI (this is because most of them will not access DFI).\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003eH4.3: For a given piece of information, participants will know it better if it is shown as DFI and they access it than if it is shown on the label (this is because we know they accessed the information in one case, thus signalling they are interested in it, while we do not know this in the other case).\u003c/p\u003e \u003c/li\u003e \u003c/ul\u003e \u003c/p\u003e \u003cp\u003eFor all our dependent variables, we are interested in \u003cb\u003e5) Individual differences.\u003c/b\u003e Based on existing studies (Ozkaya et al., \u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e2015\u003c/span\u003e; Mendelson and Bergstrom, 2013; Francis et al, \u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e2019\u003c/span\u003e; Riddle, \u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e2007\u003c/span\u003e), there are some reasons to expect greater use of DFI and better product knowledge in younger, better educated, richer, urban-dwelling individuals. We hypothesise as follows.\u003c/p\u003e \u003cp\u003e \u003cul\u003e \u003cli\u003e \u003cp\u003eH5: The following factors will make accessing DFI more likely and product information knowledge more accurate: higher education, younger age, higher income, urban location.\u003c/p\u003e \u003c/li\u003e \u003c/ul\u003e \u003c/p\u003e \u003cp\u003eNaturally \u003cb\u003e6) the type of information shown as DFI\u003c/b\u003e may also matter, although existing literature is scarce. Indirect evidence comes from Dickinson and Kakoschke (\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e2021\u003c/span\u003e) who observed confirmation bias, in that participants caring about taste (rather than health) of the product were more likely to choose to watch a \"taste matters most\" (rather than \"health matters most\") video. The same is true for the role of \u003cb\u003e7) food category and characteristics.\u003c/b\u003e Based on the distinction between \u0026ldquo;haptic\u0026rdquo; and \u0026ldquo;nonhaptic\u0026rdquo; products (Jha et al., \u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e2020\u003c/span\u003e) we could expect comparatively little interest in DFI for \u0026ldquo;haptic\u0026rdquo; products, i.e. those that consumers would be inclined to touch before buying. In our case, that would likely be the vegetable category. These considerations give rise to the following hypotheses.\u003c/p\u003e \u003cp\u003e \u003cul\u003e \u003cli\u003e \u003cp\u003eH6: There will be differences in the rate of access to DFI depending on the type of information it contains.\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003eH7: Rates of access of DFI and precision of knowledge of food characteristics will depend on the food category.\u003c/p\u003e \u003c/li\u003e \u003c/ul\u003e \u003c/p\u003e \u003cp\u003eThe hypotheses had been pre-registered under \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://osf.io/nwv38/\u003c/span\u003e\u003cspan address=\"https://osf.io/nwv38/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. We reproduced them here with some trivial reordering and re-labelling to ensure consistency within the paper.\u003c/p\u003e \u003c/div\u003e"},{"header":"2 Design of the experiment","content":"\u003cdiv id=\"Sec5\" class=\"Section2\"\u003e \u003ch2\u003e2.1 General methodology\u003c/h2\u003e \u003cp\u003eWe designed an incentivized discrete choice experiment, whereby participants were presented with choices between pairs of food products. The products were selected to represent food that is widely consumed and represent a broad range of types of food.\u003c/p\u003e \u003cp\u003eIn each case, participants saw a picture of the package, the price and \u003cem\u003emandatory food information\u003c/em\u003e (MFI) items, i.e. those, which, according to current EU legislation, must always be printed on label. We varied the mode of delivery of MFI, whereby some MFI items could be read on the paper label, and others were accessible via QR code only.\u003csup\u003e1\u003c/sup\u003e\u003c/p\u003e \u003cp\u003eWe motivated participants by giving them a budget that covered the purchase of four units of one of the products they chose during the experiment. At the end of the experiment, four units of one of the products they chose would be sold to them, at the price shown to them. The amount due was subtracted from the budget they were given. The money remaining was sent to them along the purchased items.\u003csup\u003e2\u003c/sup\u003e In this way, unlike in most related studies, participants had incentives to pick, from each pair, the product they really preferred, given their prices.\u003c/p\u003e \u003cp\u003eWe thereby avoid the “hypothetical bias” of such studies, whereby participants are not motivated to consider their choice carefully. Instead, they may choose what they think the experimenter wants them to choose (“experimenter demand effect”), or products that make them feel socially approved or that correspond best to their own ideal image of themselves (“warm glow effect”).\u003c/p\u003e \u003cp\u003eThe experiment was run online, whereby participants received an e-mail invitation and, if they decided to participate, clicked on a link to the experiment that could be accessed from their PC or tablet. This means that even the “paper” labels had to be read from the screen. On the one hand, this may be seen as a limitation on the external validity of the experiment. This design choice made “scanning the QR code” as quick as possible, without the need to reach for the device, launch the QR code scanner etc. As a result, the fraction of cases in which the QR code was “scanned” in our experiment cannot be directly taken as a predictor for the prevalence of QR code use in naturally occurring food choices.\u003c/p\u003e \u003cp\u003eOn the other hand, this feature of the design allowed us to control for various factors identified as potentially affecting the willingness to make use of digital food information. For example, as mentioned in our brief review, people may be reluctant to use QR codes out of concern for cyber security, a concern that was very unlikely to play a role in our experiment. As a result, we could focus on the essence, namely how information processing and choice depend on whether all food information is displayed immediately (paper labels) or an additional step is necessary to unlock some bits of information (hybrid labels).\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec6\" class=\"Section2\"\u003e \u003ch2\u003e2.2 Implementation\u003c/h2\u003e \u003cp\u003eThe experiment was programmed by Open Evidence in collaboration with Schlesinger Group Spain, based on a design provided by the authors. It was translated in the local languages of each country in which it took place (Bulgaria, Germany, and Spain).\u003c/p\u003e \u003cdiv id=\"Sec7\" class=\"Section3\"\u003e \u003ch2\u003e2.2.1 Design of the choice menu\u003c/h2\u003e \u003cp\u003eParticipants in this study had to make incentivized choices among 16 pairs of existing products (see \u003cb\u003eincentives\u003c/b\u003e). There were four pairs of products in each of four food categories (see \u003cb\u003eselection of food categories and products\u003c/b\u003e). Each pair of products featured two real products that were selected to be similar to each other. For example, a 500g pack of rice at 2€ was compared to a 450g pack of rice at 1€90, and the consumer had to make a choice between the two. Appendix A provides a list of pairs of products that were presented in each country.\u003c/p\u003e \u003cp\u003eInformation about each product (see \u003cb\u003efood information\u003c/b\u003e) was the same as shown on label on the actual physical product. We then varied whether all that information was provided on label in the traditional way (“paper label”) or part of it was provided as DFI (“hybrid label”); see \u003cb\u003efood information conditions\u003c/b\u003e below.\u003c/p\u003e \u003cp\u003eParticipants were asked to \u003cb\u003echoose\u003c/b\u003e one product in each of the pairs proposed. Their choice was recorded, as well as whether they \u003cb\u003eaccessed\u003c/b\u003e the DFI (if any was accessible). We also recorded how long they took to make their choice for each pair of products, and how long they looked at the DFI (if at all). For the last three choices, they were also asked to \u003cb\u003ereport\u003c/b\u003e information concerning the products they just chose from (see \u003cb\u003eknowledge questions\u003c/b\u003e).\u003c/p\u003e \u003cdiv id=\"Sec8\" class=\"Section4\"\u003e \u003ch2\u003e2.2.1.1 Selection of food categories and products\u003c/h2\u003e \u003cp\u003eThere were \u003cb\u003efour\u003c/b\u003e food categories, with four sub-categories each: \u003cb\u003ecarbohydrates\u003c/b\u003e (pasta, bread, lentils, and rice), \u003cb\u003edairy\u003c/b\u003e (milk, butter, yogurt and cheese), \u003cb\u003eproteins\u003c/b\u003e (fish, peas, meat and a vegetarian meat alternative) and \u003cb\u003evegetables\u003c/b\u003e (such as green beans, carrots, tomatoes and salad). Consumers had to choose between two products in each sub-category, selected to be close substitutes, also in terms of price. Consumers thus had to make 4x4 = \u003cb\u003e16 pairwise choices\u003c/b\u003e.\u003c/p\u003e \u003cp\u003eFood categories were selected to be representative of the typical food basket of a European consumer in countries selected for this study (Bulgaria, Germany and Spain). Food categories were the same across the three countries. Furthermore, sub-categories were very similar across countries. We varied, however, the type of cheese, the type of fish, the type of meat, or the specific vegetables. The specific products in each sub-category were country-specific, so as to correspond to actual brands available in each country. Finally, products in each sub-category were selected to be \u003cb\u003esimilar\u003c/b\u003e in terms of brand reputation, popularity and availability, quality, and relative price across countries.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec9\" class=\"Section4\"\u003e \u003ch2\u003e2.2.1.2 Display of food information\u003c/h2\u003e \u003cp\u003eAs explained below, we varied whether a product was presented with all MFI items on the label (“paper”), or with one of the Mandatory Food Information items accessible only by clicking on a QR code that was printed on the label (“hybrid”). The experiment was run online, so that “scanning a QR code” meant clicking on the QR code on screen, which opened a popup window that showed the missing information.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eFigure \u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e shows a screenshot of such choice situations. A participant can click at the bottom on “Option 1” or “Option 2” to indicate his or her choice. As stated, we chose products in each pair to be close substitutes. For example, they are both tomato sauce in this case. We show a picture of both products, their price, and a label containing all mandatory food information (see “food information” below). The screenshot in Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e is only one example of a choice situation. We varied the type of information that could be accessed via the QR code (cf. “Food information” below), and we also varied which products included a QR code. In the above example, the QR code is on the product on the left. It could also be on the product on the right, on none of them, or on both. (cf. “Information condition” below).\u003c/p\u003e \u003cp\u003eThe label for Option 1 includes a QR code, with a text above it saying that it gives access to the “best before” date (via a pop-up window). The label for Option 2 on the other hand does not have a QR code, the “best before” date being provided directly on the label. Participants know they only need to click on the QR code for a “popup” to appear, which shows the “missing” information. They are given an opportunity to train to do this before starting to make choices.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec10\" class=\"Section4\"\u003e \u003ch2\u003e2.2.1.3 Food information\u003c/h2\u003e \u003cp\u003eFor every product, we showed a picture of the product, its price, as well as the following Mandatory Food Information items:\u003c/p\u003e \u003cp\u003e1. Brand name of the food.\u003c/p\u003e \u003cp\u003e2. List of ingredients (including those causing allergies or intolerances).\u003c/p\u003e \u003cp\u003e3. Nutrition declaration.\u003c/p\u003e \u003cp\u003e4. Net quantity of the food.\u003c/p\u003e \u003cp\u003e5. Date marking (use by/best before).\u003c/p\u003e \u003cp\u003e6. Any special storage conditions and/or conditions of use (when relevant).\u003c/p\u003e \u003cp\u003e7. Name or business name and address of the food business operator.\u003c/p\u003e \u003cp\u003eOf those, the brand name and allergens were always shown on label. Other ingredients, and other types of information could be shown either on label or as DFI.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec11\" class=\"Section4\"\u003e \u003ch2\u003e2.2.1.4 Information conditions\u003c/h2\u003e \u003cp\u003eThere were four information conditions (ICs) for each product pair.\u003c/p\u003e \u003cp\u003eThe ICs were as follows:\u003c/p\u003e \u003cp\u003e● Paper-Paper: both products have paper labels, meaning that all information are shown on-label for both products in a pair.\u003c/p\u003e \u003cp\u003e● Paper-Hybrid: only the label on the right is hybrid, meaning that all information is shown on-label for the product on the left, while one of the Mandatory Food Information items is shown only via the QR code for the product on the right.\u003c/p\u003e \u003cp\u003e● Hybrid-Paper: only the label on the left is hybrid, meaning that all information shown on-label for the product on the right, while one of the Mandatory Food Information items is shown only via the QR code for the product on the left.\u003c/p\u003e \u003cp\u003e● Hybrid-Hybrid: both products have hybrid labels, meaning that both products show the same Mandatory Food Information item only via the QR code.\u003c/p\u003e \u003cp\u003eThose ICs were systematically varied across pairs for a given participant, and across participants for a given pair (Appendix B gives more details on the randomization).\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv id=\"Sec12\" class=\"Section3\"\u003e \u003ch2\u003e2.2.2 Timeline of the experiment\u003c/h2\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eThe timeline of the experiment is represented in Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e. The experiment started with a short introduction explaining that participants would have to make choices between food products, as well as answer some questions and that their responses were anonymous. We then explained to them the incentive system in the experiment, whereby there was a one in four chance that four units of one of their chosen products would be sent to them. We ensured they agreed to take part in the experiment and to give us their address in case they were chosen to receive one of their chosen products.\u003c/p\u003e \u003cp\u003e1. \u003cb\u003eFood familiarity questionnaire.\u003c/b\u003e After explaining how to make choices and what information about products would be available to them, we asked participants to fill a \u003cb\u003efood familiarity questionnaire.\u003c/b\u003e This consisted in a list of products that were going to be offered for choice, shown as pictures. Participants selected which products they already knew or had purchased in the past.\u003c/p\u003e \u003cp\u003e2. \u003cb\u003eFood choices.\u003c/b\u003e Participants then had to make 16 choices between food products presented in pairs, as explained in the previous section on the design of food choices. Namely, in each choice situation, they saw two food products, whereby both, one or none of the products had a hybrid label (see “Information condition”). They decided whether to click on the QR code(s), if available, and then chose one of the two products. They then went on to the next choice situation.\u003c/p\u003e \u003cp\u003e3. \u003cb\u003eKnowledge questions.\u003c/b\u003e Immediately after each of the last three choice situations the participants faced, we asked them to report their best estimate of the weight, best before date, or number of calories, for both products in the last three pairs.\u003csup\u003e3\u003c/sup\u003e\u003c/p\u003e \u003cp\u003e4. \u003cb\u003eFinal survey.\u003c/b\u003e Participants were asked a range of questions covering socio-demographic variables (age, gender, education …), use of Internet and digital tools, familiarity with QR codes, dietary and food related habits, as well as questions about their health, welfare, and ability to process choice information. Appendix C lists all questions asked.\u003c/p\u003e \u003cp\u003e5. \u003cb\u003eIncentives\u003c/b\u003e. A lottery was played at the end of the experiment, whereby each participant had a one in four chances to get a budget of 20€ to pay for purchasing 4 units of one of the 16 products they had chosen, at the price shown to them. This was sent to them at their address with postage paid by us. They also received any money remaining from the 20€ budget after taking out the cost of the products. The lottery itself was verifiably random, whereby we told participants how to access the page source with the code of the program that drew a random number between one and four.\u003csup\u003e4\u003c/sup\u003e\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e "},{"header":"3 Sample and data collection","content":"\u003cp\u003eThe study was run in Bulgaria, Germany, and Spain. This country selection provides good variation in terms of geographic location, GDP level, dietary habits, and propensity to buy groceries online, among other factors. Data collection was done by Open Evidence in collaboration with Schlesinger Group Spain. The experiment took place between the 8th of February and the 15th of April 2023.\u003c/p\u003e\u003cp\u003eThe target population was EU residents above 18 years of age, with a target sample size of 1000 participants for each of the three countries. In each country, the sample was stratified to ensure a good representation of the population in terms of age, gender, regions, place of residence (rural/urban) and education level.\u003c/p\u003e\u003ch2\u003e3.1 Mode of recruitment and administration of the survey\u003c/h2\u003e\u003cp\u003eThe experiment was self-administered, online, using a computer or a tablet (for optimal display of images), with no time limits.\u003c/p\u003e\u003cp\u003eParticipants were recruited using a panel-blending approach, meaning they were drawn from multiple panels simultaneously to improve representativeness of the population of reference and reach specific representativeness targets. Potential participants from panels in each country were notified about the experiment's purpose and incentives. They were assured of anonymity and data privacy, with the option to freely and voluntarily consent or decline participation. The experiment was accessed through a one-time access link provided in the invitation, which also included links to the SGS Research Privacy Policy, a removal link, and contact information for the project manager.\u003c/p\u003e\u003ch2\u003e3.2 Pre-registration, data and analytic code\u003c/h2\u003e\u003cp\u003eThe following exclusion criteria, as well our hypotheses and the analysis of results, were pre-registered at the Open Source Framework’s open registries network (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://osf.io/nwv38/\u003c/span\u003e\u003cspan address=\"https://osf.io/nwv38/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e).\u003c/span\u003e Pre-registration guarantees that none of our results are affected by hindsight bias or subject to fishing for significance (Nosek et al., \u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e2018\u003c/span\u003e). The wording of all the stimuli, the data, analytic code, and further supporting materials are also available at \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://osf.io/nwv38/\u003c/span\u003e\u003cspan address=\"https://osf.io/nwv38/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/p\u003e\u003ch2\u003e3.3 Exclusion criteria\u003c/h2\u003e\u003cp\u003eIn total, \u003cb\u003e3835 participants\u003c/b\u003e completed the experiment. In line with the pre-registered procedures (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://osf.io/nwv38/\u003c/span\u003e\u003cspan address=\"https://osf.io/nwv38/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e)\u003c/span\u003e, we excluded 35 participants who failed a basic attention question and then the fastest 5% and the slowest 5%. This meant excluding participants who spent less than 7 minutes or more than 47 minutes on the experiment.\u003c/p\u003e\u003cp\u003eThe effective sample size was then \u003cb\u003e3420 participants\u003c/b\u003e. The average duration of the experiment for that sample was 19 minutes.\u003c/p\u003e\u003ch2\u003e3.4 Evaluation of the experiment by the participants\u003c/h2\u003e\u003cp\u003eIn general, participants considered the survey to be easy (mean score of 2.7 on a 1–10 difficulty scale), and interesting (mean score of 8.1 on a 1–10 interest scale).\u003c/p\u003e\u003ch2\u003e3.5 Socio-demographics\u003c/h2\u003e\u003cp\u003eThere were 1039 participants from Bulgaria, 1203 from Germany and 1178 from Spain. There was a nearly equal representation of different age groups from 18 to 65. The sample was also balanced in terms of gender. All education levels and location types were also well represented. Samples in each country differed slightly in terms of their composition. For example, compared to Bulgaria and Spain, German participants were slightly less educated and more likely to live in a village. Participants from Bulgaria were more likely to live in big cities.\u003c/p\u003e\u003cp\u003eIn the following analyses, we systematically test the robustness of our results to socio-demographic heterogeneity in our sample. In particular, we focus on vulnerable participants, defined as those who are older (above 55), less educated (primary and high school education), with lower income\u003csup\u003e5\u003c/sup\u003e or who report being in fairly difficult or very difficult financial situation, and who live in rural settings (in villages or the countryside).\u003c/p\u003e\u003ch2\u003e3.6 Experience with digital devices, shopping online and product labels\u003c/h2\u003e\u003cp\u003eAbout 97% of participants reported owning a smartphone, and 96% using it to access the Internet. 73% of participants had scanned a QR code on a food product; the latter number was higher in Spain (83%) than in Germany (69%) or Bulgaria (66%).\u003c/p\u003e\u003cp\u003e80% said they would be either very likely or quite likely to scan a QR code on a food product in the future. This percentage was highest in Spain and lowest in Germany. 82% of participants said they really liked or liked the idea of having QR code on food products. Again, this percentage was lowest in Germany.\u003c/p\u003e\u003cp\u003eWe also asked participants what they thought would be the main benefits and drawbacks of having QR codes on food products. In terms of benefits, 57% of participants mentioned being able to get more information on food products, and 51% mentioned easier access to that additional information. In terms of drawbacks, 34% of participants mentioned the time and effort to scan QR codes, and 33% mentioned the issue of having to go on the Internet to access that information.\u003c/p\u003e\u003cp\u003eSpanish participants were more likely than others to shop for groceries online at least sometimes (63%, compared to 43% in Bulgaria and 42% in Germany). Most participants claimed they “always” or “frequently” read food labels. They also expressed relatively high trust in those labels (mean of 7 on a 1–10 trustworthiness scale).\u003c/p\u003e"},{"header":"4 Main results","content":"\u003cp\u003eIn this section, we report the analyses of the choices, in particular related to the use of DFI.\u003c/p\u003e \u003cdiv id=\"Sec20\" class=\"Section2\"\u003e \u003ch2\u003e4.1 Preference for or against DFI\u003c/h2\u003e \u003cp\u003eIn this subsection, we test whether participants preferred products with DFI or products without DFI, when both types were available. For each participant, we computed the average frequency with which they chose the product on the left of the screen in the four information conditions, namely:\u003c/p\u003e \u003cp\u003e● Hybrid-Hybrid: both products have hybrid labels.\u003c/p\u003e \u003cp\u003e● Hybrid-Paper: only the label on the left is hybrid.\u003c/p\u003e \u003cp\u003e● Paper-Hybrid: only the label on the right is hybrid.\u003c/p\u003e \u003cp\u003e● Paper-Paper: both products have paper labels.\u003c/p\u003e \u003cp\u003eFigure \u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e shows the average and 95% confidence interval for this statistic depending on the information condition across all participants.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eThe graph shows average and 95% confidence intervals.\u003c/p\u003e \u003cp\u003eWe found that in the \u0026ldquo;Hybrid-Paper\u0026rdquo; condition, the Hybrid product was chosen 50.3% of the time, while in the \u0026ldquo;Paper-Hybrid\u0026rdquo; condition, the Paper product was chosen 49.3% of the time. The one percentage point difference was not statistically significant. Similarly, in the \u0026ldquo;Hybrid-Hybrid\u0026rdquo; condition the Hybrid product on the left was chosen 49.3% of the time, while it was chosen 50.3% of the time in the \u0026ldquo;Hybrid-Paper\u0026rdquo; condition. Again, this small difference was not significant. Other comparisons did not yield significant differences either. H1.1 is not verified.\u003c/p\u003e \u003cp\u003eOverall, the likelihood of choosing a product did not depend on whether some information was only available via a QR code. We further investigated whether at least some participants always avoided products with DFI. However, the number of participants who never chose products with DFI was not significantly higher than what one would expect to happen by chance if they were in fact indifferent (for an indifferent participant facing \u003cem\u003ek\u003c/em\u003e choices featuring one product with DFI each, the probability of not choosing any of them is \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{2}^{-k}\\)\u003c/span\u003e\u003c/span\u003e). The same was true of the number of participants who \u003cem\u003ealways\u003c/em\u003e chose products with DFI.\u003c/p\u003e \u003cp\u003e \u003cstrong\u003eProduct differences\u003c/strong\u003e \u003cp\u003eWe further ran fixed-effects (within-individual) panel regressions, whereby we considered how labels affect choice while controlling for characteristics of the product on the left and on the right, for product category and for the type of information in the QR code.\u003c/p\u003e \u003c/p\u003e \u003cp\u003eColumn 1 in Table D1, Appendix D, confirms that choice is not affected by the type of label on the product on the right. Column 2 controls for the order in which products were presented, and for differences between products other than their labels. This includes not only the product characteristics but also whether the consumer knew or purchased either product before. We found that consumers were more likely to choose the Hybrid product in later choices (\u0026ldquo;Choice order\u0026rdquo;, +\u0026thinsp;0.001 per choice period, p\u0026thinsp;\u0026lt;\u0026thinsp;0.05). A product was more likely to be chosen if it had been known by the consumer while the other product had not (\u0026ldquo;diff_known\u0026rdquo;, +\u0026thinsp;0.181, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001), if it had a higher weight (\u0026ldquo;diff_weight\u0026rdquo;, +\u0026thinsp;0.0003 per gram, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001), or if it had a later expiry date (\u0026ldquo;diff_date\u0026rdquo;, +\u0026thinsp;0.001 per day, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001). Similarly, it was less likely to be chosen if it was more expensive (\u0026ldquo;diff_price\u0026rdquo;), -0.109, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001), or more calorific (\u0026ldquo;diff in kcal/100g\u0026rdquo;, -0.0002 per kcal, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001). Results in terms of preference for the product with a QR code are not affected by controlling for those variables. H1.2 is thus not verified.\u003c/p\u003e \u003cp\u003eWe furthermore test if the likelihood to choose Hybrid products depended on the category the product was in, and on the type of food information accessible through the QR code (column 3). We found no such dependence on the category and the food information.\u003c/p\u003e \u003cp\u003eFinally, we tested if differences in weight, calories and best before date mattered less if that information was shown via a QR code for the hybrid product (column 4). This would be the case if consumers did not access the QR codes and therefore could not judge the difference between the two products. We find that impact of differences in weight and calories are similar whether those differences are shown digitally or on label, but the impact of differences in the best before date is lower if it is shown digitally.\u003c/p\u003e \u003cp\u003e \u003cstrong\u003eIndividual differences\u003c/strong\u003e \u003cp\u003eWe also test whether there were differences in the likelihood to choose DFI products across different socio-demographic groups, focusing on vulnerable groups (column 5). We found that\u003c/p\u003e \u003c/p\u003e \u003cp\u003e● Germans were more likely to favour hybrid labels, though this was only marginally significant (+\u0026thinsp;0.014, p\u0026thinsp;\u0026lt;\u0026thinsp;0.1).\u003c/p\u003e \u003cp\u003e● Older individuals (\u0026gt;\u0026thinsp;55 years) were less likely to choose hybrid labels (-0.017, p\u0026thinsp;\u0026lt;\u0026thinsp;0.05).\u003c/p\u003e \u003cp\u003e● Less educated individuals were also less likely to choose hybrid labels, but again this was marginally significant (-0.012, p\u0026thinsp;\u0026lt;\u0026thinsp;0.1).\u003c/p\u003e \u003cp\u003e● Place of residence, income or financial situation do not affect the likelihood to choose a product with DFI.\u003c/p\u003e \u003cp\u003e \u003cstrong\u003eDiscussion\u003c/strong\u003e \u003cp\u003eWe expected that consumers would be less likely to choose a product with a Hybrid label. Indeed, as we will see in the following sections, QR code scanning frequency is relatively low, choice takes longer if there is a QR code, and consumers end up knowing less about products that have a Hybrid label. This should have resulted in a lower likelihood to choose a product with a Hybrid label. However, this is not the case in our experiment.\u003c/p\u003e \u003c/p\u003e \u003cp\u003eSome specificities of our design can explain this result. First, paper label and hybrid labels contained the same type of information, the only difference was how easy it was to access it Second, consumers could infer form the experimental set-up that the decision to adopt a Hybrid label was not an attempt by the firm to hide information about the product.\u003c/p\u003e \u003cp\u003eResults could be different, therefore, if information provided via QR codes was additional, not available for other products, or if firms were free to choose which information to show on label.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec21\" class=\"Section2\"\u003e \u003ch2\u003e4.2 Frequency of access to DFI\u003c/h2\u003e \u003cp\u003eOn average, participants scanned QR codes only 23% of the time. If both products in a pair had a QR code, then participants scanned both 20% of the time, and only one 5% of the time (Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e). H2.1 is verified.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab1\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eNumber of QR codes scanned, by information condition.\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"3\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/th\u003e \u003cth align=\"left\" colspan=\"2\" nameend=\"c3\" namest=\"c2\"\u003e \u003cp\u003eInformation condition\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eNumber of QR\u003c/p\u003e \u003cp\u003ecodes scanned\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eOne hybrid label\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eBoth hybrid labels\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e76.86%\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e75.26%\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e23.14%\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e4.83%\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eNA\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e19.90%\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eN\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e27360\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e13680\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003eParticipants were more likely to scan a QR code the first time they encountered one. Indeed, as many as 46% scanned the first QR code they encountered. The QR scanning frequency then dropped and stabilized at 20%. H2.2 is verified.\u003c/p\u003e \u003cp\u003e \u003cstrong\u003eProduct differences\u003c/strong\u003e \u003cp\u003eWe ran panel regressions to find determinants of the likelihood to open a QR code (Table D2, Appendix D). We found that, compared to hybrid-paper and paper-hybrid situations, participants were more likely to scan at least one QR code if both products had hybrid labels (variable \u0026ldquo;both hybrid\u0026rdquo;, +\u0026thinsp;1.6%, p\u0026thinsp;\u0026lt;\u0026thinsp;0.01) (column 1). As mentioned previously, we also found they were becoming less likely to scan a QR code over time (variable \u0026ldquo;order\u0026rdquo;, column 2). They were also less likely to open a QR code if there was a large difference in price or in weight between the two products in the pair, possibly because their choice could then be made primarily on that basis (variables \u0026ldquo;diff_weight\u0026rdquo;, \u0026ldquo;diff_price\u0026rdquo;, column 2).\u003c/p\u003e \u003c/p\u003e \u003cp\u003eWe also found that the likelihood to scan a QR code did not depend on the food category (column 3). H7 is thus not verified in terms of DFI access. However, the likelihood depended on the type of information in the QR code. Participants were most likely to scan a QR code giving access to nutrition information (32%), followed by quantity information (25%), and best before date information (22%). Those differences are statistically significant (column 3). H6 is thus verified.\u003c/p\u003e \u003cp\u003e \u003cstrong\u003eIndividual differences\u003c/strong\u003e \u003cp\u003eWhen considering differences in behaviour across participants, we found that 37% of the participants never scanned any QR code across all the 16 choices they had to make. An additional 17% scanned a QR code only once. Only 4% of the total scanned QR codes across all choices they made.\u003c/p\u003e \u003c/p\u003e \u003cp\u003eFinally, we considered the impact of socio-demographic characteristics of participants (column 4, Table D2, Appendix D). We found that Spanish participants were more likely to scan QR codes (+\u0026thinsp;3.6%, p\u0026thinsp;\u0026lt;\u0026thinsp;0.01) compared to the base category of Bulgarians. Vulnerable participants were not statistically different from the rest of the sample, regardless of whether vulnerability was defined in terms of age, education, income, or place of residence. H5 is thus not verified in terms of DFI access frequency.\u003c/p\u003e \u003cp\u003e \u003cb\u003eDiscussion\u003c/b\u003e: Rates of QR code scanning in our experiment were quite high compared to rates observed in previous studies we discussed before. This is what we aimed for: we made QR code scanning as easy as possible to be able to identify the effect of scanning QR code on behaviour.\u003c/p\u003e \u003cp\u003eOur setting was indeed a \u003cem\u003ebest-case scenario\u003c/em\u003e for the use of QR code, whereby QR code scanning meant only clicking on the screen, and participants only had 16 choices to make, of which 12 where at least one of the products had a QR code.\u003c/p\u003e \u003cp\u003eYet, participants made that effort only 24% of the time. This shows that delivering food information by that channel cannot possibly ensure participants are exposed to it.\u003c/p\u003e \u003cp\u003eOur experiment allowed us to identify interesting differences between participants. The 37% who never scanned any QR codes were also those who made the fastest choices; in other words, not scanning QR codes saved time. This was at the expense of knowing less about products, as we will see.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec22\" class=\"Section2\"\u003e \u003ch2\u003e4.3 Speed of choice\u003c/h2\u003e \u003cp\u003eWe measured how long participants took to make choices for each of the 16 choice situations they faced. We also measured how long they took to access information via the QR code.\u003c/p\u003e \u003cp\u003eWe report statistics after eliminating outliers, namely the 1% of participants who were the slowest (taking more than 2 minutes and 46 seconds to make choices). This makes our comparisons of statistics across situations more robust, as such outliers may not be equally distributed in our sample and may thus unduly influence our statistics. We also disregarded the last three choices, as participants also had to report information about their chosen product for those choices, which was included in their choice time.\u003c/p\u003e \u003cp\u003eWe found that participants took 16.9 seconds on average to make a choice and spent 3.6 seconds accessing DFI if they clicked on the QR code. Choice became faster as participants progressed through choice: the first choice took 26.8 seconds on average, while the 13th choice took only 14.8 seconds on average.\u003c/p\u003e \u003cp\u003eWe found that mean decision times were longer if there were some QR codes to be opened. Choice took 15.9 seconds on average when there were no QR codes (\u0026ldquo;Both Paper\u0026rdquo;), 16.9 seconds if one product in the pair had a QR code (\u0026ldquo;One Hybrid\u0026rdquo;), and 17.5 seconds if both products in the pair had QR codes (\u0026ldquo;Both Hybrid\u0026rdquo;) (Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e). Those differences are statistically significant (Table D3, Appendix D). H3.2 is thus verified.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab2\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 2\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eAverage choice speed, in seconds, depending on the information condition.\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"4\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e\u0026nbsp;\u003c/th\u003e \u003cth align=\"left\" colspan=\"3\" nameend=\"c4\" namest=\"c2\"\u003e \u003cp\u003eInformation condition\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eBoth Paper\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eOne Hybrid\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eBoth Hybrid\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003emean\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e15.9\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e16.9\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e17.5\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003esd\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e12.3\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e11.8\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e13.6\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eN\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e3420\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e3420\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e3420\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003eThese differences in total time spent making a choice are consistent with the likelihood to access DFIs and time spent accessing them. Indeed, participants opened about 25% of the QR codes and spent about 4 seconds reading the information in them, so this would explain why time spent making a choice is about 25% times 4 seconds\u0026thinsp;=\u0026thinsp;1 seconds more when there is a QR code.\u003c/p\u003e \u003cp\u003eWe now consider how time spent making a choice varied depending on whether a QR code was opened or not. Choices took 14.6 seconds if one product had a QR code and it was not opened, compared to 26.4 seconds if it was opened. Similarly, choices took 14.3 seconds if two products had a QR code and they were not opened, vs. 21.6 seconds if one QR code was opened, and 31.1 seconds if both were opened. Regressions (not shown) confirm those differences are significant. H3.1 is thus verified.\u003c/p\u003e \u003cp\u003eThe increase in time spent making a choice when there were QR codes, and they were opened was greater than could be explained simply by the time spent accessing the DFIs (we saw this was about 4 seconds for each QR code). There is therefore likely to be a selection effect, whereby people not opening QR codes were also those who made fast choices. Later in this section we show regressions intended to isolate the selection effects from the loss in speed that is due to QR codes.\u003c/p\u003e \u003cp\u003eWe also tested for a selection effect by relating average individual time spent making a choice in menus without QR codes and the individual likelihood to open QR codes when some are available. We found such a relation, whereby for example participants who never opened QR codes also were the fastest in making choices when there were no QR codes, taking 12.6 seconds, while those who always opened QR codes took an average of 24.5 seconds in the same situation. This points to the existence of different types of participants: those who try to make fast choices at the expense of taking less information into account, and those who try to make more considered choices at the expense of spending more time gathering information (Kahneman, \u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e2011\u003c/span\u003e).\u003c/p\u003e \u003cp\u003e \u003cb\u003eIndividual differences\u003c/b\u003e Given how differences in choice speed depending on information conditions may also be linked to individual differences, we run fixed effect (\u0026ldquo;within\u0026rdquo;) panel regressions to take out variability in choice speed across individuals and focus on differences in choice speed across choice situation for the same individual. Indeed, in fixed effects regressions, we computed the mean choice speed of each individual, and then considered deviations from that speed depending on the information condition (Table D3, Appendix D).2F\u003csup\u003e6\u003c/sup\u003e\u003c/p\u003e \u003cp\u003eWe found that hybrid labels did slow choice down even when controlling for individual differences. Compared to the paper-paper condition, choices were 0.882 seconds slower on average when there was one hybrid label and 1.625 seconds slower when there were two (column 1, Table D3, Appendix D). This confirms H3.2.\u003c/p\u003e \u003cp\u003eWe further controlled for choice order (column 2) and found that individuals made faster choice over time. Choice was also faster if only one product was known (\u0026ldquo;abs_diff_known\u0026rdquo;), when there was a larger price difference (\u0026ldquo;abs_diff_price\u0026rdquo;), or when there was a large difference in best before dates (\u0026ldquo;abs_diff_date\u0026rdquo;). This maybe be due to easier application of simple heuristics in those cases, such as going for the known product, the cheaper one or the one that will last longer. Indeed, if differences on one dimension are large, then the buyer can more easily spare time considering other aspects since they are not likely to change the decision.\u003c/p\u003e \u003cp\u003eIn terms of differences in product categories and type of information in QR codes (column 3), we found that choice for vegetable products was fastest, and that choice was slowest when the QR code contained nutritional information. This may result from consumers\u0026rsquo; previously mentioned greater propensity to access the DFI in the choices in which it contained this type of information.\u003c/p\u003e \u003cp\u003eFinally, we consider the impact of individual difference on choice speed (column 4), and how the presence of a QR code on one or both products impacted this speed differently for vulnerable participants (column 5). We found that German consumers were the fastest in making choices, followed by the Spanish, whereas Bulgarians were the slowest. Older participants (those above 55) took 5.6 seconds more to make a choice, while those with low income took 1.3 seconds less (column 4). We also see that hybrid labels impacted older participants more than the rest, whereby they slowed their choice by an additional 0.953 second (\u0026ldquo;DFI \u0026amp; Age\u0026thinsp;\u0026gt;\u0026thinsp;55\u0026rdquo;, column 5).\u003c/p\u003e \u003cp\u003e \u003cstrong\u003eDiscussion\u003c/strong\u003e \u003cp\u003eThis part shows that products with a QR code slowed choice down by about one second. We also saw that participants who were slow in making choices were also more likely to open QR codes, which slowed them even more. Slow decision makers were thus most affected by QR codes. This may be because they were those who tried to make considered choices, which involves getting all information even if behind a QR code. Unlike what could have been expected, slow decision makers did not refrain from opening QR codes (to avoid being slowed down even more). Rather, they proceed to open QR codes even at the cost of further slowdown.\u003c/p\u003e \u003c/p\u003e \u003cp\u003eEven considering differences in choice speed across participants, we found that participants were slower in making choices when there was a QR code on one or both of the products. Their choice was 0.886 second slower when one product had a Hybrid label than when both products had a paper label, and 1.625 seconds slower when both products have a hybrid label. Consumers did not compensate for time spent scanning QR codes and accessing DFIs by spending less time considering other information on food labels. Rather, the time spent scanning QR codes resulted directly in an increase in time spent making choices.\u003c/p\u003e \u003cp\u003eThis finding helps predict whether choices would be slower if Hybrid labels were widely introduced in the field. We would expect QR code scanning to be much less prevalent than in our experiment, where scanning QR codes was much easier and faster than it would be in the outside world. However, time spent scanning the QR code and accessing information would also be much higher than in our experiment, for the same reasons. It is hard to put a number for the net effect, but it would likely also result in slower choice overall. The introduction of QR codes would particularly affect slow decision makers, who as we saw before, were also the most likely to want to access DFIs. They would thus lose the most time.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec23\" class=\"Section2\"\u003e \u003ch2\u003e4.4 Product knowledge\u003c/h2\u003e \u003cp\u003eWe asked participants to report their best estimate of information about products they had chosen from: the weight, the number of kilocalories per 100g, and the \u0026ldquo;best before\u0026rdquo; dates. They answered these questions for the last three choices they had to make, right after each of those three choices, for both products in the pair they had to choose from. We varied whether the information they were asked about was shown on the product label or accessible only via a QR code.\u003c/p\u003e \u003cp\u003eWe measured errors in their responses in two ways: For the weight and the number of calories, we defined the error as abs(ln(stated value / true value)). In words, we took the absolute value of the logarithm of the ratio between the value reported by the participant and the true value. The error is 0 if the participant reported the true value. This is a good way to normalise the error because for example, the error is the same, equal to ln(2)\u0026thinsp;=\u0026thinsp;0.69 whether the participant reported twice the true value, or half of the true value. Those are indeed errors of analogous magnitude.\u003c/p\u003e \u003cp\u003eWe compute the error in a slightly different way when considering best before dates, by considering the number of days left before the best before date, as estimated by the participant vs. as shown to them with the best before date.\u003csup\u003e7\u003c/sup\u003e The error is then abs(ln(stated days left / true days left)).\u003c/p\u003e \u003cp\u003e \u003cb\u003eAverage errors\u003c/b\u003e \u003c/p\u003e \u003cp\u003eWe found that the average error for weight was 0.39 (Table\u0026nbsp;\u003cspan refid=\"Tab3\" class=\"InternalRef\"\u003e3\u003c/span\u003e, rightmost column), which is equivalent to overestimating weight by 48% or underestimating it by 32%. Average error for nutrition information (number of kilocalories per 100g) was 0.67, which is equivalent to overestimating kilocalories by 95% or underestimating it by 49%. Participants were thus significantly worse at reporting caloric content of the food than in reporting weight. This is probably because most consumers are less experienced estimating caloric content per 100g. Average error for days left was still higher, at 0.89, which is equivalent to overestimating the number of days left by 143% or underestimating it by 59%. We can speculate on the reason for this low precision; it may be that participants expected in any case that the products would not be out of date and did not care for length of conservation beyond this.\u003c/p\u003e \u003cp\u003e \u003cb\u003eImpact of the means of food information delivery\u003c/b\u003e \u003c/p\u003e \u003cp\u003eTable\u0026nbsp;\u003cspan refid=\"Tab3\" class=\"InternalRef\"\u003e3\u003c/span\u003e also shows how label types affected error rates.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab3\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 3\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eErrors for weight, calories and best before date, depending on information condition\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"7\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c7\" colnum=\"7\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003ePaper-Paper IC\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003ePaper label in \u003c/p\u003e \u003cp\u003ePaper-Hybrid IC\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003eHybrid label in \u003c/p\u003e \u003cp\u003ePaper-Hybrid IC\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c6\"\u003e \u003cp\u003eHybrid-Hybrid IC\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c7\"\u003e \u003cp\u003eTotal\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eWeight\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eMean (sd)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.38 (0.43)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.39 (0.40)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.39 (0.40)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.42 (0.43)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.39 (0.42)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eN\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e4205\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e856\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e852\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e855\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e6768\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCalories\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eMean (sd)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.68 (0.63)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.65 (0.61)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.67 (0.62)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.68 (0.61)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.67 (0.63)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eN\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e5050\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e429\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e430\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e862\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e6771\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBest before date\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eMean (sd)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.84 (0.70)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.92 (0.72)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.94 (0.72)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e1.05 (0.80)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.89 (0.72\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eN\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e4170\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e839\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e841\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e809\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e6659\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003eWe show average error depending on if none, one or both products in the pair had a hybrid label. In case only one had a paper label, we show error for the product with a paper label and error for the product with a hybrid label.\u003c/p\u003e \u003cp\u003eFor weight, the mean error was higher if information was available via the QR code, increasing from 0.38 if weight was shown on paper for both products, to 0.42 if weight was shown as DFI for both products (Table\u0026nbsp;6). For calories, the error was the same if calories were both shown on paper or both shown as DFI. For the best before date, the error increased from 0.84 if \u0026ldquo;best before\u0026rdquo; dates were shown on the paper label for both products to 1.05 if \u0026ldquo;best before\u0026rdquo; dates were accessible only as DFI for both products. Regressions presented in Table D4, Appendix D, confirm H4.2.\u003c/p\u003e \u003cp\u003e \u003cb\u003eImpact of clicking the QR code\u003c/b\u003e \u003c/p\u003e \u003cp\u003eDisaggregating further the statistic for the case of weight information being accessible via the QR code, we found that the error for weight was 0.44 if the person did not access the weight information in the DFI, vs. 0.26 if the person did access the DFI. This is to be compared with the error of 0.38 if the weight information was on the label.\u003c/p\u003e \u003cp\u003eWe thereby see a self-selection effect at play, whereby those who are motivated to access the DFI do make lower error than the average of those who get this information on the label. However, conversely, those who do not access the QR code make higher errors. Overall, too few consumers access the DFI, so the overall effect of putting information behind the QR code is to increase the error.\u003c/p\u003e \u003cp\u003eFor calorie information, we found that the error was 0.69 if the person did not access the DFI with the calorie information, vs. 0.62 if she did. This is to be compared with the error of 0.67 when information was on the label. The pattern is thus the same as for weights, but weaker and since more people accessed digital nutrition information than other types of information, the total effect is null.\u003c/p\u003e \u003cp\u003eFinally, we found that the error for best before date was 1.08 if the person did not access the DFI with that information, vs. 0.85 if the person did access the DFI. This is to be compared with the error of 0.67 if the information is on the label. Regressions (not shown) confirm those differences are significant. H4.1 and H4.3 are thus verified.\u003c/p\u003e \u003cp\u003eOverall, therefore, we found that consumers did gain in precision by accessing the QR code with the information, but also lose in precision when not accessing the QR code. The net effect of QR codes is thereby negative because too few of them click on QR codes.\u003c/p\u003e \u003cp\u003e \u003cstrong\u003eIndividual differences\u003c/strong\u003e \u003cp\u003eWe run fixed effect (\u0026ldquo;within\u0026rdquo;) panel regressions to control for differences across individuals. We normalize errors in terms of dates, weight and calories with the mean and standard deviation of the error in each of those categories.\u003csup\u003e8\u003c/sup\u003e Those regressions allow us to check how differences in label presentations affect the error while controlling for a possible selection effect whereby for example those who do not click on QR codes are those who already know the relevant food information. Results are shown in Table D4, Appendix D. We find that providing the information via a QR code consistently increases the error (columns 1\u0026ndash;2). German and Spanish participants appear to make higher errors than Bulgarians. Older participants make lower errors, but those with low income make higher errors (column 3). H5 is thus verified in terms of the error. Providing information via a QR code increases the error a lot more in Bulgaria (baseline, +\u0026thinsp;0.21) than for participants in Germany or Spain, although the effect is still significant in each of those countries. The impact does not appear to differ consistently for vulnerable participants, except for those who report being financially constrained (+\u0026thinsp;0.11) (column 4).\u003c/p\u003e \u003c/p\u003e \u003cp\u003e \u003cstrong\u003eDiscussion\u003c/strong\u003e \u003cp\u003eThis part on errors confirms that a consequence of participants not scanning QR codes is that they know less about the product they choose. Hybrid labels result in them knowing less about product characteristics than if all information is on a paper label.\u003c/p\u003e \u003c/p\u003e \u003c/div\u003e"},{"header":"5 Conclusions","content":"\u003cp\u003eWe found that most consumers did not show a preference for or against products that displayed some information through a QR code compared to products that displayed all information on printed paper labels. However, this was not the case for older and less educated consumers, who were less likely to choose products with QR codes.\u003c/p\u003e \u003cp\u003eConsumers showed little interest in accessing information through QR codes, even though our experiment was designed to make it comparatively easy to access digital food information. As many as 37% of the participants in the experiment never scanned any QR code over all the choices they made in our experiment. Only 4% scanned all of them. Overall, participants scanned QR codes in only 24% of the cases when one was shown on product labels.\u003c/p\u003e \u003cp\u003eProviding food information via QR codes had a negative impact on consumers in two respects. First, scanning QR codes slowed choice down. Participants took longer making choices when a product had a QR code. This was especially the case among older participants. This slowdown was proportional to the time spent accessing the digital food information and to the likelihood of accessing it. This means that consumers did not compensate for the time they spent scanning QR codes by accelerating other aspects of their choice process. Second, participants were less precise in their knowledge of product information if it was shown via a QR code than if it was shown printed on the labels. Indeed, consumers made more mistakes when reporting information about products with digital food information. They were not able to correctly report product information when they did not scan the QR codes. This means that consumers who did not scan QR codes did not do so because they already knew the information therein. Rather, they simply chose to forgo knowledge about the product.\u003c/p\u003e \u003cp\u003eIn summary, we found that consumers were unlikely to access information provided through QR codes. Giving access to information via QR codes slowed choice down. Providing food information digitally also resulted in consumers knowing less about the food products they had chosen from. However, consumers did not respond to this by avoiding products with QR codes.\u003c/p\u003e \u003cp\u003eThis study illustrates that printed paper labels are more effective than QR codes in giving access to food information. Labels that require digital means to access food information are not a good substitute for printed paper labels that give direct access to food information. This is due to the time spent scanning QR codes, the low likelihood to scan them, and the resulting lack of knowledge about the characteristics of the products. These negative effects of the use of QR codes emerged even though our setup made it particularly easy to access digital information.\u003c/p\u003e \u003cp\u003eWe note however that our study did not aim to investigate possible benefits of QR codes when used as a complement, rather than as a substitute for information printed on labels. For example, QR codes could provide more in-depth information than could fit when printed on a label. Furthermore, changes in technology and in consumers\u0026rsquo; habits could affect the potential of digital means of access to food information.\u003c/p\u003e \u003cp\u003eFunding Declaration: This work was funded by the European Commission.\u003c/p\u003e \u003cp\u003eClinical trial number: not applicable.\u003c/p\u003e \u003cp\u003eHuman Ethics and Consent to Participate declarations: see attached ethics committee approval\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003eFunding Declaration: This work was funded by the European Commission.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eClinical trial number: not applicable.\u003c/p\u003e\n\u003cp\u003eHuman Ethics and Consent to Participate declarations: see attached ethics committee approval\u003c/p\u003e\u003ch2\u003eAuthor Contribution\u003c/h2\u003e\u003cp\u003eAG took part in all the stages of this project.MK consulted the design and data analysis, supervised development of stimuli, co-edited the text.\u003c/p\u003e\u003ch2\u003eData Availability\u003c/h2\u003e\u003cp\u003eThe wording of all the stimuli, the data, analytic code, and further supporting materials are also available at https://osf.io/nwv38/.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eBashir, H. (2022). Leveraging technology to communicate sustainability-related product information: Evidence from the field. \u003cem\u003eJournal of Cleaner Production\u003c/em\u003e, \u003cem\u003e362\u003c/em\u003e, 132508.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBettman, J. R., Johnson, E. J., \u0026amp; Payne, J. W. (1990). A componential analysis of cognitive effort in choice. \u003cem\u003eOrganizational behavior and human decision processes\u003c/em\u003e, \u003cem\u003e45\u003c/em\u003e(1), 111\u0026ndash;139.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBradford, H., McKernan, C., Elliott, C., \u0026amp; Dean, M. (2022). Consumer purchase intention towards a quick response (QR) code for antibiotic information: an exploratory study. \u003cem\u003enpj Science of Food\u003c/em\u003e, \u003cem\u003e6\u003c/em\u003e(1), 23.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBray, J., Hartwell, H., Price, S., Viglia, G., Kapuściński, G., Appleton, K., \u0026amp; Mavridis, I. (2019). Food information presentation: consumer preferences when eating out. \u003cem\u003eBritish Food Journal\u003c/em\u003e, \u003cem\u003e121\u003c/em\u003e(8), 1744\u0026ndash;1762.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eClinton, V. (2019). Reading from paper compared to screens: A systematic review and meta-analysis. \u003cem\u003eJournal of research in reading\u003c/em\u003e, \u003cem\u003e42\u003c/em\u003e(2), 288\u0026ndash;325.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDickinson, D. L., \u0026amp; Kakoschke, N. (2021). Seeking confirmation? Biased information search and deliberation in the food domain. \u003cem\u003eFood quality and preference\u003c/em\u003e, \u003cem\u003e91\u003c/em\u003e, 104189.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFenko, A., Nicolaas, I., \u0026amp; Galetzka, M. (2018). Does attention to health labels predict a healthy food choice? An eye-tracking study. \u003cem\u003eFood quality and preference\u003c/em\u003e, \u003cem\u003e69\u003c/em\u003e, 57\u0026ndash;65.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFrancis, J., Ball, C., Kadylak, T., \u0026amp; Cotten, S. R. (2019). Aging in the digital age: Conceptualizing technology adoption and digital inequalities. In Ageing and digital technology: Designing and evaluating emerging technologies for older adults, 35\u0026ndash;49.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFoodDrinkEurope (2024, January 17). \u003cem\u003eMaking consumer information accessible to all\u003c/em\u003e. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.linkedin.com/pulse/making-consumer-information-accessible-all-fooddrinkeurope-nmwde/\u003c/span\u003e\u003cspan address=\"https://www.linkedin.com/pulse/making-consumer-information-accessible-all-fooddrinkeurope-nmwde/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGarbarino, E. C., \u0026amp; Edell, J. A. (1997). Cognitive Effort, Affect, and Choice. \u003cem\u003eJournal of Consumer Research\u003c/em\u003e, \u003cem\u003e24\u003c/em\u003e(2), 147\u0026ndash;158.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGl\u0026ouml;ckner, A., \u0026amp; Betsch, T. (2012). Decisions beyond boundaries: When more information is processed faster than less. \u003cem\u003eActa psychologica\u003c/em\u003e, \u003cem\u003e139\u003c/em\u003e(3), 532\u0026ndash;542.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGolman, R., Hagmann, D., \u0026amp; Loewenstein, G. (2017). Information avoidance. \u003cem\u003eJournal of economic literature\u003c/em\u003e, \u003cem\u003e55\u003c/em\u003e(1), 96\u0026ndash;135.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJha, S., Balaji, M. S., Stafford, M. B. R., \u0026amp; Spears, N. (2020). Haptic information processing and need for touch in product evaluation. \u003cem\u003eJournal of Consumer Marketing\u003c/em\u003e, \u003cem\u003e37\u003c/em\u003e(1), 55\u0026ndash;64.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKahneman, D. (2011). \u003cem\u003eThinking, Fast and Slow\u003c/em\u003e. Macmillan.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKang, M. J., Hsu, M., Krajbich, I. M., Loewenstein, G., McClure, S. M., Wang, J. T. Y., \u0026amp; Camerer, C. F. (2009). The wick in the candle of learning: Epistemic curiosity activates reward circuitry and enhances memory. \u003cem\u003ePsychological science\u003c/em\u003e, \u003cem\u003e20\u003c/em\u003e(8), 963\u0026ndash;973.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKim, Y. G., \u0026amp; Woo, E. (2016). Consumer acceptance of a quick response (QR) code for the food traceability system: Application of an extended technology acceptance model (TAM). \u003cem\u003eFood Research International\u003c/em\u003e, \u003cem\u003e85\u003c/em\u003e, 266\u0026ndash;272.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLacroix, A., Muller, L., \u0026amp; Ruffieux, B. (2019). Labeling for Sustainable Food: The Consumer's Point of View. In V. Clavier, \u0026amp; De J. P. Oliveira (Eds.), \u003cem\u003eFood and Health\u003c/em\u003e. Actor Strategies in Information and Communication, Wiley.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLi, T., \u0026amp; Messer, K. D. (2019). To Scan or Not to Scan: The Question of Consumer Behavior and QR Codes on Food Packages. \u003cem\u003eJournal of Agricultural and Resource Economics\u003c/em\u003e, \u003cem\u003e44\u003c/em\u003e(2), 311\u0026ndash;327.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLombardi, A., Carfora, V., Cicia, G., Del Giudice, T., Lombardi, P., \u0026amp; Panico, T. (2017). Exploring willingness to pay for QR code labeled extra-virgin olive oil: An application of the theory of planned behavior. \u003cem\u003eInternational Journal on Food System Dynamics\u003c/em\u003e, \u003cem\u003e8\u003c/em\u003e(1), 14\u0026ndash;31.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMa, G., \u0026amp; Zhuang, X. (2021). Nutrition label processing in the past 10 years: Contributions from eye tracking approach. \u003cem\u003eAppetite\u003c/em\u003e, \u003cem\u003e156\u003c/em\u003e, 104859.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMcFarlane, T., \u0026amp; Pliner, P. (1997). Increasing willingness to taste novel foods: effects of nutrition and taste information. \u003cem\u003eAppetite\u003c/em\u003e, \u003cem\u003e28\u003c/em\u003e(3), 227\u0026ndash;238.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMendelson, J., \u0026amp; Romano Bergstrom, J. C. (2013). Age differences in the knowledge and usage of QR codes. In \u003cem\u003eUniversal Access in Human-Computer Interaction. User and Context Diversity: 7th International Conference, UAHCI 2013, Held as Part of HCI International 2013, Las Vegas, NV, USA, July 21\u0026ndash;26, 2013, Proceedings, Part II 7\u003c/em\u003e (pp. 156\u0026ndash;161). Springer Berlin Heidelberg.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eNeijens, P. C., \u0026amp; Voorveld, H. A. (2018). Digital replica editions versus printed newspapers: Different reading styles? Different recall? \u003cem\u003eNew media \u0026amp; society\u003c/em\u003e, \u003cem\u003e20\u003c/em\u003e(2), 760\u0026ndash;776.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eNosek, B. A., Ebersole, C. R., DeHaven, A. C., \u0026amp; Mellor, D. T. (2018). The preregistration revolution. Proceedings of the National Academy of Sciences, 115(11), 2600\u0026ndash;2606.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eOonk, L. (2013). QR codes, quick response or quick rejection: A study about the contribution of the phenomenon QR codes on food products, on the intention to seek information and the purchase intention. Master\u0026rsquo;s thesis, University of Twente.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eOrquin, J. L., Bagger, M. P., Lahm, E. S., Grunert, K. G., \u0026amp; Scholderer, J. (2020). The visual ecology of product packaging and its effects on consumer attention. \u003cem\u003eJournal of Business Research\u003c/em\u003e, \u003cem\u003e111\u003c/em\u003e, 187\u0026ndash;195.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003e\u0026Oslash;vrum, A., Alfnes, F., Almli, V. L., \u0026amp; Rickertsen, K. (2012). Health information and diet choices: Results from a cheese experiment. \u003cem\u003eFood Policy\u003c/em\u003e, \u003cem\u003e37\u003c/em\u003e(5), 520\u0026ndash;529.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eOzkaya, E., Ozkaya, H. E., Roxas, J., Bryant, F., \u0026amp; Whitson, D. (2015). Factors affecting consumer usage of QR codes. \u003cem\u003eJournal of Direct Data and Digital Marketing Practice\u003c/em\u003e, \u003cem\u003e16\u003c/em\u003e, 209\u0026ndash;224.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eQualityChain (2024, January 16). \u003cem\u003eWe just can\u0026rsquo;t put everything on the label.\u003c/em\u003e \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.linkedin.com/pulse/we-just-cant-put-everything-label-qualitychain-qhysf\u003c/span\u003e\u003cspan address=\"https://www.linkedin.com/pulse/we-just-cant-put-everything-label-qualitychain-qhysf\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRiddle, D. R. (Ed.). (2007). \u003cem\u003eBrain Aging Models, Methods, and Mechanisms\u003c/em\u003e. Routledge.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRoberto, C. A., \u0026amp; Khandpur, N. (2014). Improving the design of nutrition labels to promote healthy food choices and reasonable portion sizes. \u003cem\u003eInternational Journal of Obesity\u003c/em\u003e, \u003cem\u003e38\u003c/em\u003e(1), S25\u0026ndash;S33.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTemple, N. J., \u0026amp; Fraser, J. (2014). Food labels: a critical assessment. \u003cem\u003eNutrition (Burbank, Los Angeles County, Calif.)\u003c/em\u003e, \u003cem\u003e30\u003c/em\u003e(3), 257\u0026ndash;260.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003evan der Laan, L. N., \u0026amp; Orcholska, O. (2022). Effects of digital Just-In-Time nudges on healthy food choice \u0026ndash; A field experiment. \u003cem\u003eFood Quality and Preference\u003c/em\u003e, \u003cem\u003e98\u003c/em\u003e, 104535. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.foodqual.2022.104535\u003c/span\u003e\u003cspan address=\"10.1016/j.foodqual.2022.104535\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWansink, B., \u0026amp; Sobel, J. (2007). Hidden persuaders and 200 daily decisions. \u003cem\u003eEnviron Behav\u003c/em\u003e, \u003cem\u003e39\u003c/em\u003e(1), 106\u0026ndash;123.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZentall, T. R., \u0026amp; Stagner, J. P. (2012). Do pigeons prefer information in the absence of differential reinforcement? \u003cem\u003eLearning \u0026amp; behavior\u003c/em\u003e, \u003cem\u003e40\u003c/em\u003e, 465\u0026ndash;475.\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"},{"header":"Footnotes","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003e An alternative to our choice to vary presentation of MFI would have been to consider digital delivery of voluntary food information, such as measures of a product\u0026rsquo;s Environmental, Social and Governance dimensions, or nutritional aspects of the food that are not yet generally provided (such as their Nutriscore and/or alternatives and variations thereof). We decided not to do so because that type of information is not available for all products, and consumers are not yet knowledgeable about their meaning and importance.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003e For practical and budgetary reasons, this was implemented for 25% of randomly suggested participants only (and they were aware of this).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003e We asked this only for the last three choices to make the first 14 choices as close to normal choice as possible. In the last two choices, participants may anticipate they will be asked to recall information they consulted, which may affect their choice and information search.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003e Postal addresses were collected at the end of the experiment from those participants who won the lottery. However, we made sure at the beginning of the experiment that all participants were ready to give their address in case of a win. This guarantees that all participants participated with the knowledge that one of their choices would be implemented.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003e Monthly household income less than 1400 LEV in Bulgaria, less than 2400\u0026euro; in Germany, less than 1500\u0026euro; in Spain. Those thresholds correspond to the bottom 40% in terms of household income in our sample in each country.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003e We need to do so because as we saw, slower individuals are also more likely to open QR codes, which would explain why choice is slower when the QR code is opened.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003e Unlike weight and calories, which have a lowest value of 0, dates do not have such a clear and meaningful lowest value. We therefore use the date at which a participant made their choice as the reference point.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003e For example, the normalized error for weight is (error - mean error for weight)/std dev. of the error for weight.\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"journal-of-consumer-policy","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"copo","sideBox":"Learn more about [Journal of Consumer Policy](http://link.springer.com/journal/10603)","snPcode":"10603","submissionUrl":"https://submission.springernature.com/new-submission/10603/3","title":"Journal of Consumer Policy","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"Springer Hybrid","inReviewEnabled":true,"inReviewRevisionsEnabled":false},"keywords":"consumer choice, digital labelling, food information, discrete choice experiment, food labels, QR codes","lastPublishedDoi":"10.21203/rs.3.rs-8549197/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-8549197/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eWe conducted an online experiment with 3420 participants in Bulgaria, Germany and Spain to evaluate the consequences of digital access to food information via QR codes. Participants made incentivized choices between food products with labels that were either \u0026ldquo;paper\u0026rdquo; or \u0026ldquo;hybrid\u0026rdquo;. Only the latter type featured a QR code; scanning it was necessary to access some of the information. Presence of QR codes did not affect probability of choosing given product. As many as 37% of participants never scanned QR codes, and only 4% scanned all of them. Hybrid labels led to longer decision times and less accurate knowledge, indicating a negative impact on consumers.\u003c/p\u003e","manuscriptTitle":"Using QR codes to access food information: A behavioural study with European consumers","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-02-03 15:45:09","doi":"10.21203/rs.3.rs-8549197/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"decision","content":"Revision requested","date":"2026-03-03T12:33:26+00:00","index":"","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2026-03-02T17:13:27+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2026-03-02T14:51:16+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"54029121483189642195222845855489331656","date":"2026-02-03T05:36:17+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"41768427998178875713015761028502484261","date":"2026-02-02T09:56:46+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2026-01-29T10:50:31+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2026-01-27T15:59:37+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2026-01-15T23:27:35+00:00","index":"","fulltext":""},{"type":"submitted","content":"Journal of Consumer Policy","date":"2026-01-08T08:42:18+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"journal-of-consumer-policy","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"copo","sideBox":"Learn more about [Journal of Consumer Policy](http://link.springer.com/journal/10603)","snPcode":"10603","submissionUrl":"https://submission.springernature.com/new-submission/10603/3","title":"Journal of Consumer Policy","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"Springer Hybrid","inReviewEnabled":true,"inReviewRevisionsEnabled":false}}],"origin":"","ownerIdentity":"ae83b01c-bbd4-4289-8c36-c90769b04298","owner":[],"postedDate":"February 3rd, 2026","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"in-revision","subjectAreas":[],"tags":[],"updatedAt":"2026-03-03T12:41:13+00:00","versionOfRecord":[],"versionCreatedAt":"2026-02-03 15:45:09","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-8549197","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-8549197","identity":"rs-8549197","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}