"Warning: ultra-processed": an online experiment examining the impact of ultra-processed warning labels on consumers' product perceptions and behavioral intentions · D'Angelo Campos A, Ng SW, Duran AC et al.

Results

Participants’ mean age was 37.1 years (SD = 12.2) and 54% identified as men. The sample had a high level of educational attainment compared to the Brazilian population: around 62% of participants had a bachelor’s degree or more, compared to 19% of the Brazilian population [52]. Around 53% of participants identified white, 33% as pardo (i.e., “brown” or mixed race), and 12% as Black – compared to 44%, 45%, and 10%, respectively, of the Brazilian population [53]. In terms of regional distribution, the sample resembled the Brazilian population: 46% of participants reported living in the Southeast, 30% in the Northeast, and 12% in the South – compared to 42%, 27%, and 15%, respectively, of the population [53]. More than half of participants reported that they were confident in their ability to identify ultra-processed foods prior to the experiment (Table 3).

Table 1 details the descriptors that participants used in their open-ended responses related to meaning of the term “ultra-processed.” The most used descriptor, employed by 56% of participants, was a reference to food processing, either in a tautological manner (e.g., “very processed”) or using other terms to describe food transformation processes. The second most used descriptor, employed by 14% of participants, was a reference to additives, either directly or using terms that allude to additives, such as “artificial ingredients,” “lab-made ingredients,” and “ingredients not used in home cooking.” Overall, 58% of participants used at least one descriptor that referred to lack of whole foods as a UPF characteristic, and 18% used at least one descriptor that referred to synthetic substances. Other descriptors used by smaller percentages of participants included nutrient content (6%), large number of ingredients (5%), and convenience (6%). Around 7% of participants used unhealthfulness as a UPF descriptor.

Participants also used descriptors that did not refer to any UPF characteristics. Around 4% of participants mentioned descriptors that are food-related but not specifically UPF-related, such as “genetically modified foods,” “non-organic foods,” or “non-fresh foods,” among others. Around 5% provided answers that defined UPFs as “outdated” or “surpassed” – possibly a result of misreading the Portuguese word for ultra-processed (ultraprocessado) as a similar word with this meaning (ultrapassado). Around 4% provided answers that exhibited a positive interpretation of UPFs as “good-quality” or “sophisticated” products, which contradicts the Brazilian Dietary Guidelines’ description of UPFs. Lastly, some participants provided non-food related descriptors (6%), stated that they did not know what the term “ultra-processed” means, or did not provide a response (5%). Based on their use of these descriptors, we classified 8% of the sample as having a high understanding, 69% as having a moderate understanding, and 23% as having a low understanding of the concept of UPF (Table 2).

The experiment demonstrated that a higher share of participants who saw the nutrients plus ultra-processed warning labels correctly identified all products shown as being UPFs, compared to participants who saw nutrients-only warning labels (58% vs. 51% respectively, d = 0.15, p = 0.021, Fig. 3). However, the nutrients plus ultra-processed warning labels did not differ from nutrients-only warning labels on participants’ intentions to purchase the products, perceived product healthfulness, and perceived label effectiveness (all p > 0.05, Fig. 4). Similarly, there were no differences by experimental arm on purchase intentions, perceived product healthfulness, and perceived label effectiveness when examining effects for individual products (Table S1).

Lastly, moderation analyses did not reveal significantly different effects of experimental arm on identification of products as ultra-processed or purchase intentions by age, gender, education, self-reported health status, self-reported confidence in understanding of the term “ultra-processed,” or low understanding of the term “ultra-processed” (all p > 0.05, Tables S2 and S3).