Can We-Media Help Reduce the Consumption of Sugary Fruit Juice Drinks With the Presence of the Halo Effect?—Insights From A Discrete Choice Experiment in Beijing,China

Abstract

While fruit juices are widely considered healthy, their high sugar contents are often masked by a “halo effect” created by the positive image of fruits. This study, taking orange juice as a focal product, explores the role online self-media plays in reducing consumers’ sugary drink consumption with the presence of a halo effect. A discrete choice experiment (with attributes including fruit juice concentration, sugar content, taste, and price) was conducted in Beijing, China, to measure changes in consumers’ choices induced by randomly provided sugar content information. Four treatments containing the same basic information on the sugar content of sugary drinks, with different presentation forms (three text messages and a short video) and specificity levels, were designed for the experiment. Estimated mixed-logit models indicate that all four treatments significantly reduced participants’ likelihood of choosing high-sugar orange juices, with the short-video treatment being the most effective. It is likely that the video helped consumers convert a vague concept (sugar content) into a concrete one (sugar cubes). It is also possible that, with videos, consumers can passively assimilate information with less effort than reading text instructions.

Keywords

fruit juice halo effect sugar information online self-media China

Introduction

Sugary foods and beverages, such as cookies, ice cream, carbonated beverages, and fruit juices, are popular among consumers of almost all ages. Yet excessive sugar intake may be health-threatening, acting as a risk factor for health problems such as cardiovascular disease (Chen et al., 2010; Jayalath et al., 2015; Malik et al., 2010), diabetes (F. B. Hu, 2013), and fatty liver (Assy et al., 2008), to name a few. Due to improvements in nutrition and health awareness among the general public, the proportion of carbonated and flavored drinks in one’s diet has declined in recent years, accompanied by increased consumption of compound fruit juices (Q. R. Zhang et al., 2021). However, as healthy as fruit juices can be, their high sugar content is often overlooked (Duffy et al., 2021). In particular, the healthy image of fruits may form a positive impression of fruit juices in consumer perception, masking their high sugar content through a “halo” effect (Shefferly et al., 2016), a cognitive bias that often arises when a favorable claim on the packaging influences consumers’ judgment of other product attributes not mentioned on the packaging (Centurión et al., 2019; Hall et al., 2020; Roe et al., 1999). Even products labeled with “fruit sugar” are deemed healthier than those labeled with just “sugar” (Sutterlin & Siegrist, 2015). Positively sounding nutrition claims printed on product packages, such as “natural” and “vitamins,” may further aggravate the halo effect (André et al., 2019).

Partly due to the halo effect, many consumers are reluctant to classify fruit juices, especially “100%” ones, as sugary beverages (Bolling et al., 2009). Many sugar-controlling initiatives also precluded fruit juices as their target product. In Mexico and England, for instance, most tax policies regulating the consumption of high-sugar foods do not apply to fruit juices (Sah et al., 2021). As such, the halo effect not only poses potential health risks to juice consumers but also complicates the formulation of related policies. With the presence of the halo effect, how to correct the general public’s misunderstanding of fruit juices’ nutritional (sugar) content is a question of great policy relevance.

The present study attempts to provide an answer, at least a partial one, by investigating how exogenously provided information on sugar content may affect consumers’ choice of orange juice, the most common juice product on the market. A discrete choice experiment (DCE) involving four attributes (fruit-juice concentration, sugar content, taste, and price) and four information treatments (in the form of text messages and short videos) was conducted in Beijing, the Capital City of China. China currently ranks No. 3 worldwide in sugar consumption, and per capita consumption has increased year by year due to rapid income growth (Liu et al., 2022). Although its per capita sugar intake is still low compared with most Western countries, China’s rapid increase in sugar consumption triggers concern about its potential contribution to the country’s overweight epidemic—from 1991 to 2011, overweight and obesity prevalence among children (6–17 years) increased from 4.3% and 2.4% to 11.5% and 10.5%, respectively; the corresponding figures for adults increased from 16.8% and 3.8% to 31% and 11.3%, respectively (Pan et al., 2021). Since juice consumers in China mainly come from large cities with vibrant food markets like Beijing, insights from a study on Beijing consumers’ behavior can inform future consumption trends in China and other countries with large and active sugary beverage markets.

Inspired by the widespread popularity of smartphones and mobile networks in China (China Internet Network Information Center, 2022), our experiment emphasizes the role of internet-based self-media (also called “We Media”), one of today’s essential channels for information dissemination, in steering consumer fruit juice choices. Specifically, we designed a short video treatment selected from a well- known We Media channel (Bilibili), along with three text-message treatments containing information on sugary drinks’ sugar content, and randomly sent one of these four treatments to each participant (N = 142–146 in each treatment group; another 147 participants in a randomly assigned control group received no such information). The video clip consists of a 5-min lecture delivered by a lab technician, which expresses the sugar content as the number of sugar-cube equivalents; the three text messages contain the same basic information but with different levels of specificity and detail.

Our analysis of data from the DCE and its companion survey yields three informative findings. First, the survey reveals a nontrivial halo effect. For example, even though over 92% of the participants were aware of the existence of sugar in fruit juices, more than half (56%) mistakenly believed that “drinking 100% fruit juices has the same effect as eating fresh fruits, which is completely good for one’s health.” Second, our mixed-logit models suggest that compared with the control group, a “treated” consumer’s likelihood of choosing “100% orange juice” is significantly lower in all treatment groups, suggesting an essential role of (exogenously-provided) sugar-content information in reducing sugary drinks consumption, even with the presence of a halo effect. Finally, while all four information treatments reduced consumers’ willingness to consume 100% orange juices, the video-clip treatment was the most effective, highlighting the potential role of online self-media in promoting nutrition literacy among beverage consumers.

Literature Review

Existing Approaches to Controlling Sugary Drink Consumption

As a major channel for many consumers’ sugar intake and consumption worldwide, sugary drinks have attracted considerable attention from researchers and policymakers (Zenone et al., 2021). One particular concern is that a typical sugary drink contains more sugar than the recommended daily intake level (WHO, 2015). As the general public’s nutrition and health awareness have improved in recent decades, attention has increased to strategies that may help reduce sugary beverage consumption (Popkin & Hawkes, 2016). Two approaches have been widely examined in the literature. The first is sugar taxes. Recent studies from the U.K. and the U.S. suggest that taxes on sugary drinks effectively reduce consumers' likelihood of choosing these drinks (Cawley et al., 2020; Cornelsen et al., 2020). The second approach concerns nutrition-content labels, which often include warnings, health stars, and nutrition levels. Compared with other types of labels, “high in X” labels (where X = sugar, fat, etc.) were found to be more effective at reducing the consumption of targeted nutrients (Acton & Hammond, 2018; Acton et al., 2019).

The labels used for sugary beverages mainly include traffic-light and sugar-teaspoon labels. The former uses three colors (red, yellow, and green) to signal different levels (high, medium, and low) of nutrients per 100 ml (100 g) of a drink. Recent experimental evidence suggests that consumers exposed to color-coded labels are more likely to avoid unhealthy food choices and seek healthier options (Defago et al., 2020; Thorndike et al., 2012). Compared with traffic-light labels, sugar-teaspoon labels provide more vivid and detailed visual representations of sugar content, i.e., the number of sugar cubes or sugar teaspoons per serving of a drink. It has been found that when presented with specific information about sugar content (compared with abstract or no information), consumers felt more negatively toward sugary drinks and were less willing to consume them (Adams et al., 2014; Miller et al., 2024).

The Role of Video Persuasion

Effectiveness of Video Persuasion Compared With Text Messages

In addition to the aforementioned labels, video persuasion and intervention have been widely employed to enhance health by inducing changes in health knowledge and perceptions, attitudes, intentions, and behaviors (Abu Abed et al., 2014; Amaro et al., 2017; Ito-Jaeger et al., 2023; Juhong et al., 2022; Zhu et al., 2023), offering a new perspective for reducing the consumption of sugary drinks. Regarding information content and dissemination format, health-related videos can be broadly divided into three types: didactic, practice, and narrative presentations (Tuong et al., 2014). Of course, in the Digital Age, especially with the popularity of short videos, video content is not limited to these types.

Regarding the effectiveness of video persuasion, the “dual coding” theory suggests that individuals process information through two systems: one for visual information and one for verbal information (Paivio, 2013). Both systems are functional when watching videos, but the visual information processing system is suppressed when reading text messages, which renders the information in videos easier to digest. However, there has been no consensus in the existing literature on the effectiveness of video persuasion compared to text messages. Although many studies have pointed out that videos are more effective than text messages (Cheung et al., 2017; Walthouwer et al., 2015), some recent studies have provided new evidence showing no significant difference between videos and text messages in inducing behavioral changes (Aref et al., 2024; Vandelanotte et al., 2021). Thus, how videos may be used to effectively induce healthy behavior remains to be explored.

The Role of Video Content

One factor to consider is the content of videos. The explosive growth of short videos on online self-media platforms such as Bilibili and TikTok provides new carriers of nutrition and health information. Through eye-catching expressions, short videos offer audiences greater immersion, intuitiveness, and authenticity than text messages and have become increasingly popular on online self-media (Eghtesadi & Florea, 2020). Along with other content, such as sports and education, health-related information is also popular, and the rapid development of online self-media greatly facilitates its dissemination (Bora et al., 2018; Song et al., 2021). How to utilize such short video platforms to improve public health has become a hotly debated topic among health researchers and practitioners (Eghtesadi & Florea, 2020). For example, animated videos, a popular type of media, have been explored for their role in promoting improvements in health behavior (Supady et al., 2023; Vandormael et al., 2021). However, compared with animated videos, videos featuring people as the main subjects are more prevalent and generally more preferred by video viewers, which is also a factor to consider in the present study.

The Role of Platforms

Platforms also play a role. By providing ample opportunities for information sharing, online self-media has become a low-cost and efficient platform with great potential for further development (Mehmet et al., 2020). Traditionally, video interventions have limited effectiveness due to a lack of customization (Tuong et al., 2014). Today’s media platforms can circumvent this problem by using users’ search and viewing history to recommend content relevant to their interests and needs. Yet the quick and widespread dissemination of information facilitated by online self-media presents both opportunities and challenges for public health awareness. While it helps promote public health awareness by spreading knowledge about new medical developments and health-policy updates (Li et al., 2022; Vandormael et al., 2021), the constant spread of misinformation and the heavy mental burden of processing new information also create cognitive barriers that undermine public health awareness (Syed-Abdul et al., 2013). These opportunities and challenges suggest that for self-media to be an effective channel for popularizing nutrition and health information, ensuring the authenticity and reliability of the information disseminated is key (Madathil et al., 2015).

Gaps in Knowledge

Despite increasing interest among scholars and policymakers in the role of online self-media in promoting healthy food consumption, little attention has been paid to the sugar content of fruit juices and possible self-media interventions to increase consumers’ awareness. In addition to the general lack of understanding of sugary drinks’ sugar content, the healthy image of fruit creates a halo effect for fruit juice, leading consumers to believe that the latter is equally healthy; partly because of this, much of the world population has been loyal fruit juice consumers (Luckow & Delahunty, 2004). Yet, despite the many healthy nutrients in fruit juices that are absent in sugar-sweetened beverages (Mullen et al., 2008; Pereira-Caro et al., 2014), the high sugar content of fruit juices is often overlooked by the public. Even though “100% not-from-concentrate” juices contain no added sugar, they usually contain 80% to 130% of the fructose content of sugar-sweetened beverages, the excessive consumption of which is harmful to health (Walker et al., 2014). For example, research has shown that fruit juice intake is associated with an increased risk of diabetes (Bazzano et al., 2008). Indeed, the WHO (2015) recommended reducing the intake of fructose and other free sugars to below 10% of one’s daily energy intake. It is, therefore, imperative to inform consumers about the sugar content of fruit juices and the health risks associated with consuming them.

However, relevant studies rarely exist. Most existing studies on fruit-juice consumption examined how packaging (Pal et al., 2017), labeling (Schouteten et al., 2019), claims (Sabbe et al., 2009; Vidigal et al., 2011), and juice-processing technologies (Lee et al., 2016) affect consumers’ juice choice. It is encouraging, however, to see that studies aimed at reducing the consumption of other sugary foods using treatments such as “sugar content” labels have been increasing (Jindahra & Phumpradab, 2023; Jürkenbeck et al., 2022; Stoltze et al., 2021). To contribute to this emerging literature, we examine how different ways of providing information on the sugar content of fruit juices may affect consumers’ choices through a DCE. Our experimental design builds on insights from the literature discussed above, especially regarding the selection of intervention approaches and information carriers.

Experimental Design

Basic Design

We chose orange juice, the most common juice product in the market, as the focal product in the DCE. To avoid biases stemming from consumers’ preferences for specific brands, we named the product “Orange Desired,” a hypothetical new brand with no connection with existing brands. Participants were informed that the products to choose from were bottles of orange juice (330 ml/bottle) stored at room temperature, characterized by four attributes, each with 4 to 5 levels (Table 1):

Table 1.

Attributes and Levels of the Hypothetical Product (A Bottle of Orange Juice, 330 ml).

Attributes	Levels
Fruit juice concentration	100% juice (not-from-concentrate), 100% juice (from concentrate), Juice content >10%, Juice content >5%
Sugar content	0, 12, 24, 36 g
Taste	Sweet, slightly sweet, balanced, slightly sour, sour
	An example of a “star chart” for sweetness/sourness:
	sour	slight sour	balanced	slight sweet	sweet

Price	RMB 3, RMB 6, RMB 9, RMB 12

Fruit-Juice Concentration

By the level of juice concentration, fruit juices can be divided into “100%” and “non-100% juice” products. The former includes “not-from-concentrate (NFC)” 100% fruit juices, which are processed directly from fresh fruits, and “from-concentrate (FC)” 100% fruit juices made from concentrated juices with fresh water. “Non-100%” juices are all “FC” but with different concentrations. Based on the classification method provided by China’s National Standard (GB/T31121-2014), four levels were set for this attribute: “100% juice (NFC),”“100% juice (FC),”“juice >10% (FC),” and “juice >5% (FC).”

Sugar Content

Sugary drinks can also be classified by based on sugar content: sugar-free, low-sugar, and ordinary. “Sugar-free” means that the beverage contains no more than 0.5 g of sugar per 100 ml (or 100 g). “Low-sugar” means that the sugar content does not exceed 5 g/100 ml (or 100 g). “Ordinary” drinks contain more than 5 g/100 ml (or 100 g) of sugar. To ensure that our experiment covered all three categories, we set four levels for a bottle of orange juice: 0 g (sugar-free), 12 g (low-sugar), 24 g (ordinary), and 36 g (ordinary). Currently, there are no completely sugar-free fruit juices on the market, but as food development moves forward, it is helpful to explore consumers’ preferences and attitudes towards such products.

Note also that the sugar content of juices in this study refers to the total sugar content without distinguishing between fructose, sucrose, and glucose. There are two reasons for this design. First, the distinction between sucrose, glucose, and fructose is difficult for most consumers to understand. Introducing specific types of sugars could complicate consumer choices and render our findings difficult to interpret. Second, existing food products usually present sugar content in the standard nutritional composition tables and food labels. As such, designing the sugar content attribute in our DCE as total sugar content more closely mirrors the real-life context in which consumers make food choices, thus helping to elicit more accurate consumer responses.

Taste

Naturally, taste matters for consumers’ beverage choices. Five levels were set to capture its role: sweet, slightly sweet, balanced, slightly sour, and sour. Following Acton et al. (2019), we use a “star chart” to characterize the degree of sourness/sweetness (Table 1) and help consumers understand these levels more easily.

Price

Price is an essential component in any demand analysis. Given the current price range (3–10 RMB/bottle, roughly 0.5–1.5 US dollars) of orange juices in China, we chose four price levels for a bottle within the 3-to-12 RMB (about 0.5–1.8 US dollars) range, with equal increments of 3 RMB (about 0.5 US dollars).

Product-Profile Creation

The above design yields 320(= 4×4×5×4) potential orange-juice profiles. Such a large number imposes a heavy mental burden on consumers when making choices. An orthogonal fractional factorial design (Louviere et al., 2000) was thus used to reduce the number of product profiles to 100. The resultant profiles were divided into 10 blocks, each containing 10 single-choice questions, using blocking techniques. Participants were randomly assigned to one of the 10 blocks in the DCE. Figure 1 illustrates an actual choice task used in the DCE.

Figure 1.

An actual choice set used in the experiment.

Information Intervention

To better understand how consumers’ fruit juice choices respond to exogenously provided information on sugar content, we designed four treatments based on three lessons from the literature. The first concerns the choice of intervention approaches. Although China’s per capita sugar intake has been rising, it has not yet reached the level that triggered tax interventions in Western countries; thus, sugar taxes may not be a realistic option in the Chinese context. Our intervention, instead, is based on the idea of nutrition labels.

The second lesson concerns the choice of labels, particularly between traffic-light and sugar-teaspoon labels. The literature suggests that sugar-teaspoon labels tend to convey information more directly and clearly than traffic-light labels, thus being more effective at helping consumers make healthier choices (Mauri et al., 2021). Therefore, we tied our treatments to sugar-teaspoon labels: sugar content is presented as the number of sugar-cube equivalents in a sugary drink.

The third lesson regards the choice of information carriers. Existing studies seldom compare the effectiveness of different information carriers. To shed light on this issue, we designed four treatments in the form of text messages and short video clips (Table 2). The first, “basic information” treatment (T₁), briefly notes that the sugar content of fruit juice is high (measured in sugar cube equivalents) and that excessive sugar intake may cause disease. The second treatment (T₂), “basic information + authoritative source,” adds the WHO-recommended sugar intake level (WHO, 2015) to T₁. The third treatment (T₃), “information + authoritative source + specific figures,” further adds specific figures on the sugar contents of various drinks and their potential harm, again referencing WHO (2015). The last treatment (T₄) delivers the same content as T₃ but uses a short video as the information carrier.

Table 2.

Contents of Information Treatments.

Treatments	Level of specificity	Contents
Basic information (T₁)	Limited figures	The sugar content of beverages on the market shall not be ignored. A bottle of 500-ml sugary beverage contains more sugar than that in 10 sugar cubes. Sometimes sugar is called “sweet poison.” Chinese individuals’ sugar consumption has increased in the past few years. Surveys show that there are many patients diagnosed with dyslipidemia, diabetes, obesity, hypertension, and other chronic diseases in China. These diseases are directly or indirectly related to excessive sugar intake.
Basic information & authoritative source (T₂)	With the sugar intake level recommended by the WHO added	The World Health Organization recommends that adults' daily sugar intake of should not exceed 25 g. The sugar content of beverages on the market shall not be ignored. A 500-ml bottle of sugary beverage contains more sugar than ten sugar cubes. Sometimes sugar is called “sweet poison.” Chinese individuals’ sugar consumption has increased in the past few years. Surveys show that there are many patients diagnosed with dyslipidemia, diabetes, obesity, hypertension, and other chronic diseases in China. These diseases are directly or indirectly related to excessive sugar intake.
Information with specific figures & authoritative source (T₃)	With more specific figures to show the sugar content and its potential harm added	The World Health Organization recommends that adults' daily sugar intake should not exceed 25 g. The sugar content of beverages on the market shall not be ignored. With a 4.5-g sugar cube being a reference, the sugar content of a 450-ml orange juice drink can be converted into 11 sugar cubes, that of a bottle of 600-ml Coca-Cola into 14 sugar cubes, that of a 500-ml pear juice drink into 14 sugar cubes, that of a 600-ml apple-flavor carbonated beverage into 17 sugar cubes, and that of a bottle of 330-ml lactic acid bacteria beverage into 11 sugar cubes. Sometimes sugar is called “sweet poison.” Over the past 30 years, China’s per capita sugar consumption has increased year by year. According to the 2012 Consensus of Chronic Disease Prevention and Treatment Experts issued by the Ministry of Health, there are 33 million individuals diagnosed with dyslipidemia, 97 million with diabetes, 120 million with obesity, and more than 200 million with hypertension in China. These diseases are directly or indirectly related to excessive sugar intake.
Short video (T₄)	With the visualization of information on the sugar content and its potential harms	The content is taken from a video clip selected from the nationally popular online self-media platform Bilibili (https://v.qq.com/x/page/b3329hfh7po.html) and edited to fit the online questionnaire. This video visually presents the conversion of the sugar content of popular sugary drinks in the market into the equivalent number of sugar cubes.

Source. Author’s design.

Note. Figure 2 shows several screenshots of the video clip involved in T₄.

Since the main characters in existing health information videos are mostly hospitals (rather than human characters), which do not align with those in the present-day mainstream self-media, we chose a short video featuring a lab technician frequently spotted on online self-media as the basis for our video treatment. To find a video suitable for this study, the research team browsed hundreds of videos on popular social media apps on the topic of “sugar in beverages” in China. After discussing with other experts, we excluded videos that were either considered “boring” (e.g., those focusing on the principle that “sugar is bad for health”), very subjective, or those promoting a specific product to avoid offending participants. Several screenshots of the chosen video clip are presented in Figure 2.

Figure 2.

Screenshots of the video presented to DCE participants.

Quality Control

To urge the participants to make their choices carefully, we asked them to check the following statement before the choice tasks started: “I will provide honest answers throughout the experiment.” A note was also provided to remind them to read the instructions about the choice task carefully before making choices. We further included an example choice task in the questionnaire, with detailed explanations of the attributes of juice alternatives, to ensure participants understood how to make choices based on their preferences and budget constraints.

To make the choice scenario as realistic as possible, a cheap-talk script was included to remind participants to imagine facing various juice products in the market when making their choices. The cheap-talk script reads: “Similar surveys have found that people tend to overestimate how much they are willing to pay for a given product. Please behave as you would if you had to pay for the product and take it home. Next, suppose you are going to buy a bottle of orange juice at a supermarket that offers a range of juice products. Please choose the product you prefer the most from the options, subject to your daily budget.”

Survey and Data

The study was conducted in accordance with the Declaration of Helsinki and was approved by the Human Research Ethics Committee of China Agricultural University. A pilot survey involving 68 participants was conducted in April 2022 to test the validity of the experimental design and the questionnaire’s content. The formal survey, involving 725 participants from Beijing, was implemented in September 2022 through Questionnaire Star, the largest online survey platform in China. This platform, powered by Changsha Ranxing IT Ltd., provides functions equivalent to those of Amazon Mechanical Turk. With the ability to reach nearly 300 million users per month, the platform can filter out specific target groups and control data quality using techniques such as IP restrictions, trap questions, and similarity detection. Of particular importance to our study is that the platform allows us to incorporate videos in the questionnaire through a web link.

During the online survey, informed consent was obtained from all the participants (by choosing the “consent” option). They were informed about the purpose of the study and the minimal harm their participation may impose on them at the beginning of the survey. The participants were also informed that they could stop at any point during the study for any reason. The survey lasted 3 to 5 minutes per participant, depending on the information provided. Each participant received an allowance of 11 RMB (≈1.7 US dollars) for their participation; the payment was nearly twice the average wage that employed workers in Beijing would earn in 5 minutes in 2022—the average monthly wage in 2022 was 13,369 RMB (≈2,057 US dollars) in Beijing (https://www.beijing.gov.cn/ywdt/gzdt/202211/t20221121_2863122.html accessed: August 4, 2025). Thus, the participants were fully compensated for their time spent participating in our study. The health information they received during the study may lead to additional health benefits for them in the future.

We opted to conduct the DCE online for three reasons. First, online survey platforms provide a more convenient way to deliver video messages than in-person surveys. Second, an online intervention naturally takes advantage of consumers’ access to online self-media; by using cell phones or computers to complete the choice tasks, they can pause the video at any time desired. Moreover, due to the COVID-19 pandemic, online participants may have become much more representative of the food-/juice-shopper population than in the pre-pandemic period. In fact, given Beijing’s frequent lockdowns in the past few years and the rapid development of online shopping systems (e.g., Hema, Meituan, and Elema) during the pandemic, many food shoppers have chosen online shopping as the primary food-purchasing method (Beijing Municipal Bureau of Statistics, 2021; Ministry of Commerce, 2020). This shift in primary food-purchasing methods may continue in the years to come. Currently, consumers who mainly purchase foods at supermarkets and grocery stores tend to be older, less educated, and unfamiliar with digital technology; they may not represent the juice-consumer population of interest to our study.

Questionnaires with different choice tasks were randomly distributed to the participants through Questionnaire Star. The platform used two methods to ensure that each participant completed the questionnaire only once. First, it linked accounts in the user pool to users’ unique ID card numbers, WeChat verification, and facial recognition, allowing each account to have at most one completed questionnaire submission. Second, the link sent to the participants was specially programmed so that once the completed questionnaire had been submitted, the link became invalid and could not be reopened or forwarded to another participant. To assess the quality of randomization, a Chi-square test of the similarity in socio-demographic characteristics across subgroups (four treatment groups and one control group), collected after the DCE choice tasks were completed, was performed. The test detected no statistically significant between-group differences (Table 3).

Table 3.

Balance in Socio-Demographic Characteristics of Participants Across Groups.

Characteristics	Description	Control group (%)	Treatment groups				Chi-test (p-value)
Characteristics	Description	Control group (%)	T₁: basic Information (%)	T₂: Basic information & authoritative source (%)	T₃: Information with specific figures & authoritative source (%)	T₄: 5-min video clip (%)	Chi-test (p-value)
Gender	Male	40.82	39.04	40.85	36.30	36.81	.9997
Gender	Female	59.18	60.96	59.15	63.70	63.19	.9999
Age	20 or under	7.48	4.79	1.41	2.74	2.08	.9883
	21–25	14.29	14.38	12.68	9.59	10.42	.9988
	26–30	34.01	32.19	33.80	32.88	29.17	.9998
	31–35	25.17	30.82	29.58	28.08	29.86	.9985
	36–40	7.48	9.59	11.27	17.12	15.28	.9725
	41–45	6.12	4.11	7.04	5.48	7.64	.9996
	46–50	2.72	2.05	2.11	2.74	5.56	.9985
	51–55	0.68	0.68	1.41	0.68	0.00	.9995
	56–60	2.04	0.00	0.00	0.68	0.00	.9952
	61 or older	0.00	1.37	0.70	0.00	0.00	/
Education	Primary school	0.34	0.33	0.00	0.00	0.52	.9998
	Junior high school	1.36	0.33	0.00	0.00	0.69	.9980
	(Vocational) high school	3.07	1.00	3.59	1.00	2.07	.9985
	College/university	79.05	76.67	75.90	82.39	77.80	.9999
	Master’s degree or higher	16.18	21.67	20.51	16.61	18.93	.9983
Monthly income (RMB)	≤4,999	2.25	2.01	1.91	0.84	1.81	.9997
	5,000–9,999	17.57	14.77	19.57	10.02	16.33	.9979
	10,000–19,999	31.76	39.60	35.80	38.83	35.37	.9975
	20,000–49,999	38.74	29.53	36.28	35.07	35.37	.9986
	50,000–99,999	5.63	10.07	3.58	11.48	5.67	.9914
	≥100,000	4.05	4.03	2.86	3.76	5.44	.9998
N		147	146	142	146	144	/

Source. Author’s survey.

Admittedly, collecting data through a third-party platform has several drawbacks. In particular, respondents participating through online survey panels may answer questions differently from the general population. One may expect the participants to be more interested in juice consumption than non-participants. Nonetheless, many recent studies (Bartneck et al., 2015; Casler et al., 2013; Lindhjem and Navrud, 2011; Windle and Rolfe, 2011) have demonstrated the reliability of stated-preference studies conducted with online participants. Recent studies using data collected through Questionnaire Star have also yielded important findings (Ning et al., 2020; H. S. Wang et al., 2023; Zheng and Zheng, 2014). Even if the participants are more interested in juice consumption than non-participants, the former’s preferences for orange juices are likely tainted with a stronger halo effect. As such, they still serve as a relevant sample for studying potential approaches to reducing sugary drink consumption with the presence of a halo effect.

Analytical Framework

To assess how information provided on online self-media may help reduce the likelihood of choosing sugary fruit juices, we developed an analytical framework based on McFadden’s (1974) random utility theory. In this framework, the impacts of information treatments can be modeled as following certain distributions (rather than being fixed parameters) whose variations are potentially affected by those in the attributes discussed in “Experimental Design” section, such as sugar content and taste (Train, 2009). Specifically, assume that the utility ( $U_{ijt}$ ) a consumer i derives from choosing orange juice product $j (= 1, \dots, J)$ in choice scenario $t (= 1, \dots, 10)$ has the following form:

U_{ijt} = β' x_{ijt} + ε_{ijt}

(1)

where $x_{ijt}$ is the vector of observed attribute-level combinations of alternative $j (= 1, \dots, J)$ defined in the “Experimental Design” section and $ε_{ijt}$ represents $U_{ijt}' s$ unobserved components. With these attribute-level combinations, Equation (1) can be further specified as:

\begin{array}{l} U_{i j t} = β_{0} A S C_{i j t} + β_{1} 100 %_{NFC}_{i j t} + β_{2} 100 %_F C_{i j t} \\ + β_{3} f r u i t_j u i c e > 10 %_{i j t} + β_{4} s u g a r_0 g_{i j t} \\ + β_{5} s u g a r_12 g_{i j t} + β_{6} s u g a r_24 g_{i j t} + β_{7} s o u r_{i j t} \\ + β_{8} s l i g h t l y_s o u r_{i j t} + β_{9} b a l a n c e d_{i j t} \\ + β_{10} s l i g h t l y_s w e e t_{i j t} + β_{11} p r i c e_{i j t} + ε_{i j t}, \end{array}

(2)

where $β_{0}$ is an alternative-specific constant (ASC) representing the choice of “None of these.” All attribute-level combinations except prices enter the model as dummy variables (=1 if present in alternative j; =0 otherwise). The reference levels of “fruit-juice concentration,”“sugar content,” and “taste” are, respectively, fruit_juice >5%, sugar_36 g, and sweet.

Following the standard practice in the recent DCE literature, we estimated the parameters in Equation (2) using mixed-logit models (Train, 2009). Compared with conventional models, such as multinomial-logit and conditional-logit, mixed-logit is more flexible. By relaxing the “independence of irrelevant alternatives” restriction, mixed-logit allows for unobserved heterogeneity captured by random coefficients on the attribute-level combinations, which can be affected by our information treatments. Following Ortega et al. (2016), Revelt and Train (1998), and Tonsor et al. (2009), among others, we treated the coefficients of prices ( $β_{11}$ ) and ASC ( $β_{0}$ ) as fixed parameters and other parameters ( $β_{1}$ $~ β_{10}$ ) as normally distributed random parameters. Given estimated parameters, the willingness-to-pay (WTP) for a non-price attribute/level can be recovered as $WT P_{k} = - \frac{β_{k}}{β_{11}}$ , with $β_{k}$ being the coefficient of the kth attribute/level in Equation (2) (Train, 2009).

To further assess the effects of our information treatments, we further included interaction terms between the treatment-group dummies, $T_{l} (l = 1, \dots, 4)$ , and the related attributes in Equation (2): $X_{ijt} \times T_{l}$ , where $X_{ijt}$ is the full set of attribute-level combinations included in Equation (2), except “sugar content,” which is treated as a continuous variable here. Given the repeated-choice structure of the data for each consumer, the level terms of the treatment-group dummies and any factors that do not vary within the chosen product necessarily drop out of the model. The impact of a particular treatment, $T_{l}$ , is then captured by the coefficient of the corresponding interaction term, $x_{ijt} \times T_{l}$ , with the interpretation as “the impact of treatment T_l, which changes the likelihood of consumer i choosing product j through $x_{ijt}$ (e.g., a sugar content or taste level), relative to the control group.”

Results

Participant Profiles, Use of Online Media, and Cognition of Fruit Juices

Table 4 summarizes the participants’ socio-demographic characteristics. 61% are female. The largest group was 26 to 30 years old at the time of the survey, with a university degree and a monthly household income of 10,000-19,999 RMB (roughly 1,480–2,960 US dollars). While the participants were younger and better educated than the Beijing population, they shared a similar socio-demographic profile with beverage/juice consumers in China (Ye et al., 2005), the target population of our study.

Table 4.

Socio-Demographic Characteristics of Respondents.

Characteristics	Description	Percentage
Gender	Male	38.76
Gender	Female	61.24
Age (years)	20 or under	3.72
	21–25	12.28
	26–30	32.41
	31–35	28.69
	36–40	12.14
	41–45	6.07
	46–50	3.03
	51–55	0.69
	56–60	0.55
	61 or older	0.41
Education	Primary school	0.97
	Junior high school	0.97
	(Vocational) high school	2.90
	College/university	79.86
	Master’s degree or higher	15.31
Monthly income (RMB)	≤4,999	5.38
	5,000–9,999	23.86
	10,000–19,999	37.24
	20,000–49,999	26.90
	50,000–99,999	4.55
	≥100,000	2.07
Residential district	Xicheng District	12.14
	Haidian District	7.86
	Dongcheng District	26.07
	Shijingshan District	11.17
	Chaoyang District	4.00
	Fengtai District	13.66
	Other (8 outer districts)	20.96

Source. Author’s survey.

Table 5 reports participants’ use of online media, revealing the popularity of online media and the large amount of time participants spend watching short videos online. Eighty-one percent of participants often watch short videos online, far exceeding the proportion who often read text information (48.69%). Notably, the proportion of participants who watch short videos for over 2 hr/day is almost twice that of those who read text information for over 2 hr/day. Given the high frequency and long time spent watching short videos, the information they provide will likely impact consumers’ behavior in a notable way.

Table 5.

Respondents’ Use of Short Videos and Text Information.

Frequency/time of use	Description	Percentage
Frequency of watching short-form videos	Never	0.41
	Hardly	1.1
	Occasionally	17.52
	Often	80.97
Frequency of reading text information	Never	0.97
	Hardly	3.45
	Occasionally	46.9
	Often	48.69
Average daily time spent watching short-form videos	<0.5 hr	6.65
	0.5–1 hr	29.22
	1–2 hr	31.58
	2–3 hr	17.87
	≥3 hr	14.68
Average daily time spent reading text information	<0.5 hr	24.09
	0.5–1 hr	36.77
	1–2 hr	25.35
	2–3 hr	7.24
	≥3 hr	6.55

Source. Author’s survey.

Table 6 shows participants’ cognition of fruit juices and their sugar content, assessed by seven questions, revealing a potentially nontrivial halo effect. For example, 56% of the participants mistakenly believed that “Drinking 100% fruit juices has the same effect as eating fresh fruits, which is completely good for one’s health” (Table 6, question 1), even though the vast majority (>92%) were aware of the existence of sugar in fruit juices (Table 6, questions 3 and 6). The existence of a halo effect was also evident when comparing answers from residents of different districts in Beijing. For example, Xicheng and Haidian, the two most affluent urban districts of Beijing, have the highest income and education levels among all Beijing districts. Still, their residents’ correct-answer rates are the lowest among all six inner-city districts in Beijing. In particular, even though Xicheng residents had the highest correct-answer rates for questions 5 (“The sugar content of 100% fruit juices is almost negligible”, 91.23%) and 7 (“Reducing sugar intake will benefit one’s health, e.g., by reducing the risk of some diseases,” 84.21%), nearly two-thirds Xicheng participants mistakenly answered question 1 (“Drinking 100% fruit juices has the same effect as eating fresh fruits, which is completely good for one’s health”).

Table 6.

Respondents of Cognition of Fruit Juice.

Questions	Correct answer	Percentage choosing the correct answer (%)
Questions	Correct answer	Full sample	Xicheng District	Haidian District
(1) Drinking 100% fruit juices has the same effect as eating fresh fruits, which is completely good for one’s health.	False	43.59	36.84	41.41
(2) 100% fruit juice is nutritious and healthy and can replace water as a daily drink.	False	76.69	75.44	80.81
(3) Fruit juices contain no sugar as long as no sugar is added to it.	False	92.41	85.96	92.93
(4) The sugar content in fruit juices cannot be ignored, which can even be higher than that in carbonated drinks.	True	71.86	75.44	68.69
(5) The sugar content in 100% fruit juices is almost negligible.	False	82.48	91.23	80.81
(6) Sugar widely exists in our daily diet and can provide us with the energy needed for daily activities.	True	92.83	94.74	93.94
(7) Reducing sugar intake will benefit one’s health, for example, by reducing the risk of some diseases.	True	78.62	84.21	77.78

Source. Author’s survey.

To assess whether participants have a cognitive bias (in particular, underestimation) in estimating the sugar content of orange juice, we informed them that the total sugar content of a bottle of orange juice (330 ml) is equivalent to that of 8 sugar cubes (4.5 g) at the end of the survey. The results show that nearly three-quarters underestimated the sugar content of orange juice, reporting that 8 sugar cubes/bottle was “slightly” (38%) or “much” (35%) higher than they expected. Such underestimations provide further evidence of the halo effect of fruit juice drinks among Chinese consumers.

Estimation Results of Mixed-Logit Models

Roles of Product Attributes

Table 7 reports the results of estimating two mixed-logit models with 500 Halton draws using NLOGIT 5.0. Model 1 includes only the estimated random coefficients of juice attributes/levels. Model 2 further includes interactions between these attributes/levels and information-treatment dummies, allowing consumers’ preferences to differ across treatment groups. We opt to report findings of a large model with interaction terms rather than estimating the model separately for the five groups (four treatment groups and one control group) for two reasons. First, the former approach allows us to include more observations (from all five groups) in the model, thereby improving the precision of the estimates. Second, it allows us to test directly whether the between-group difference in the coefficients on the same explanatory variable (e.g., “100% NFC”), captured by the coefficient on the interaction term (e.g., “100% NFC × video (T₄)”), is statistically significant, which is not feasible when the model is estimated separately for each group. Note that in a “full” interaction model in which all explanatory variables—even if their impacts are statistically insignificant—are fully interacted with group dummies (T₁–T₄), these two sets of estimates will be mathematically equivalent—one can recover coefficients in each group from between-group differences, and vice versa. For example, the coefficient on “100% NFC × video (T₄)”, say, $δ_{1}$ , captures the difference in the coefficient on “100% NFC” in the video-treatment group (T₄) and that in the control group, $β_{1}$ ; thus, given an estimate of the coefficient on “100% NFC” in the control group, that in the T₄ group can be easily recovered as $(β_{1} + δ_{1})$ . Because Model 2 converts information on sugar content (four dummies in Model 1) into a continuous variable to reduce the number of interaction terms, we expect slight differences between the (recovered) estimates in Model 2 and the estimates for each group (based on Model 1’s specification). To check for robustness, Table A1 reports the results of estimating the mixed-logit model separately for the five groups, which are comparable to those of Model 2. To put the estimated impacts into perspective, Table 8 reports the corresponding mean WTPs for different treatments.

Table 7.

Mixed-Logit Estimates.

Attributes/Levels	Model 1		Model 2
Attributes/Levels	Mean of parameter	SD of parameter	Mean of parameter	SD of parameter
100% NFC	1.052*** (0.072)	1.196*** (0.071)	1.490*** (0.152)	1.229*** (0.071)
100% FC	0.946*** (0.068)	1.046*** (0.072)	1.273*** (0.159)	1.030*** (0.072)
Juice content >10% (reference level: juice content >5%)	−0.029 (0.077)	1.088*** (0.092)	−0.040 (0.175)	1.141*** (0.086)
Sugar 0 g	1.452*** (0.085)	1.682*** (0.084)	1.324*** (0.181)	1.686*** (0.085)
Sugar 12 g	0.874*** (0.067)	0.914*** (0.076)	0.966*** (0.142)	0.917*** (0.074)
Sugar 24 g (reference level: sugar 36 g)	0.320*** (0.067)	0.717*** (0.091)	0.548*** (0.149)	0.781*** (0.084)
Sour	−0.517*** (0.098)	1.719*** (0.101)	−0.677*** (0.200)	1.685*** (0.097)
Slightly-sour	0.035 (0.072)	0.894*** (0.091)	−0.328** (0.164)	0.898*** (0.087)
Balanced	0.299*** (0.070)	0.872*** (0.083)	0.174 (0.169)	0.895*** (0.087)
Slightly-sweet (reference level: sweet)	0.263*** (0.068)	0.610*** (0.106)	0.180 (0.157)	0.632*** (0.096)
100% NFC × basic information (T₁)			−0.307 (0.204)
100% NFC × basic information & authoritative source (T₂)			−0.631*** (0.218)
100% NFC × information with specific figures & authoritative source (T₃)			−0.442** (0.216)
100% NFC × video (T₄)			−1.083*** (0.227)
100% FC × basic information (T₁)			−0.468** (0.202)
100% FC × basic information & authoritative source (T₂)			−0.689*** (0.206)
100% FC × information with specific figures & authoritative source (T₃)			−0.578*** (0.206)
100% FC × video (T₄)			−0.843*** (0.220)
Juice content >10% × basic information (T₁)			0.115 (0.228)
Juice content >10% × basic information & authoritative source (T₂)			0.090 (0.232)
Juice content >10% × information with specific figures & authoritative source (T₃)			0.030 (0.225)
Juice content >10% × video (T₄)			−0.356 (0.241)
Sugar × basic information (T₁)			0.006 (0.006)
Sugar × basic information & authoritative source (T₂)			−0.012* (0.006)
Sugar × information with specific figures & authoritative source (T₃)			0.001 (0.006)
Sugar × video (T₄)			−0.016*** (0.006)
Sour × basic information (T₁)			0.440 (0.268)
Sour × basic information & authoritative source (T₂)			0.469 (0.286)
Sour × information with specific figures & authoritative source (T₃)			0.384 (0.281)
Sour × video (T₄)			0.867*** (0.285)
Slightly-sour × basic information (T₁)			0.295 (0.213)
Slightly-sour × basic information & authoritative source (T₂)			0.106 (0.219)
Slightly-sour × information with specific figures & authoritative source (T₃)			0.054 (0.215)
Slightly-sour × video (T₄)			0.304 (0.243)
Balanced × basic information (T₁)			−0.009 (0.222)
Balanced × basic information & authoritative source (T₂)			0.232 (0.215)
Balanced × information with specific figures & authoritative source (T₃)			0.040 (0.210)
Balanced × video (T₄)			0.349 (0.232)
Slightly-sweet × basic information (T₁)			0.015 (0.209)
Slightly-sweet × basic information & authoritative source (T₂)			0.010 (0.202)
Slightly-sweet × information with specific figures & authoritative source (T₃)			−0.258 (0.209)
Slightly-sweet × video (T₄)			0.159 (0.216)
Price	−0.111*** (0.008)		−0.117*** (0.008)
ASC	−0.183* (0.093)		−0.196** (0.098)
Log-likelihood	−8,082.606		−8,024.603
AIC	16,209.2		16,157.2
Mcfadden R²	0.196		0.202
N	725		725

Note. Mixed-logit estimates for each of the treatment groups are reported in Table A1. Standard errors in parentheses.

***

p < .01, **p < .05, *p < .1.

Table 8.

WTP Estimates for Different Groups (Unit: RMB).

Attributes/levels	Full sample	Control group	Basic information (T₁)	Basic information & authoritative sources (T₂)	Information with specific figures & authoritative sources (T₃)	Video (T₄)
100% NFC	9.38 [7.23, 8.01]	10.93 [9.63, 12.23]	9.95 [8.48, 11.42]	8.90 [7.68, 10.12]	9.43 [8.17, 10.69]	7.62 [6.74, 8.49]
100% FC	8.91 [7.96, 8.83]	10.32 [9.3, 11.34]	9.05 [8.14, 9.96]	8.55 [7.53, 9.56]	8.21 [7.15, 9.27]	8.39 [7.42, 9.36]
Juice content >10%(reference group: juice content >5%)	−0.36 [−0.53, 0.42]	−1.39 [−2.34, −0.45]	−0.16 [−1.06, 0.75]	0.21 [−0.75, 1.17]	−0.40 [−1.52, 0.73]	−0.05 [−1.12, 1.01]
Sugar 0 g	13.26 [15.95, 18.05]	11.51 [9.93, 13.25]	11.34 [9.56, 13.13]	14.58 [12.62, 16.54]	11.91 [9.83, 13.98]	17.00 [14.65, 19.36]
Sugar 12 g	8.08 [7.61, 8.24]	8.24 [7.55, 9.21]	7.69 [6.84, 8.54]	8.51 [7.68, 9.33]	7.90 [7.03, 8.76]	7.92 [7.21, 8.63]
Sugar 24 g(reference group: sugar 36 g)	2.93 [2.23, 2.67]	3.21 [2.89, 3.91]	3.02 [2.44, 3.66]	3.14 [2.59, 3.69]	2.63 [2.06, 3.20]	2.45 [1.96, 2.94]
Sour	−4.33 [−3.93, −2.37]	−5.65 [−7.61, −3.69]	−3.99 [−0.5.75, −2.23]	−4.84 [−6.73, −2.95]	−4.02 [−6.03, −2.02]	−3.15 [−4.89, −1.40]
Slightly-sour	−0.11 [0.1, 0.68]	−0.80 [−1.46, −0.15]	0.19 [−0.55, 0.94]	−0.38 [−1.14, 0.37]	0.04 [−0.71, 0.79]	0.39 [−0.26, 1.05]
Balanced	2.71 [2.84, 3.43]	2.44 [1.78, 3.10]	2.60 [1.90, 3.30]	2.95 [2.22, 3.67]	2.47 [1.71, 3.22]	3.13 [2.47, 3.79]
Slightly-sweet(reference group: sweet)	2.37 [1.97, 2.34]	2.82 [2.41, 3.24]	2.34 [1.90, 2.77]	2.36 [1.98, 2.74]	2.14 [1.73, 2.56]	2.16 [1.75, 2.57]
N	725	147	146	142	146	144

Model 1 reveals three informative findings. First, in general, consumers prefer orange juices with higher juice concentrations to those with lower concentrations. Specifically, consumers prefer 100% orange juices (both FC and NFC), but not those with juice content >10%, to those with juice content >5% (the reference group). Corresponding to the positive coefficients of “100% FC” and “100% NFC” juices, consumers’ WTPs for these two attribute levels are 9.38 and 8.91 RMB, respectively (Table 8, column 1). Other things equal, the higher WTPs for 100% orange juices than for non-100% juices suggest a halo effect (Table 8).

Second, consumers prefer orange juices with relatively low sugar content to those with high sugar content. With reference to orange juice with 36 g sugar per bottle, consumers prefer sugar-free juice the most ( ${\hat{β}}_{sugar_0 g} = 1.45$ ), followed by low-sugar ( ${\hat{β}}_{sugar_12 g} = 0.87$ ) and high-sugar ( ${\hat{β}}_{sugar_24 g} = 0.32$ ) juices. Consistent with this order, estimated WTPs decrease as the sugar content increases (Table 8, column 1).

Finally, taste serves as an essential factor affecting consumers’ orange juice choices. Compared with orange juice with a sweet taste (the reference group), slightly sweet ( ${\hat{β}}_{slightly_sweet} = 0.26$ ) and balanced-flavor ( ${\hat{β}}_{balanced} = 0.30$ ) juices are more popular among the participants. Echoing such preferences, consumers’ WTPs for slightly sweet and balanced tastes are 2.37 and 2.71 RMB, respectively (Table 8, column 1). In contrast, consumers dislike sour-tasting orange juice ( ${\hat{β}}_{sour} = - 0.52$ ); they are willing to receive a discount of 4.33 RMB for juice with a sour taste than that with a sweet taste (Table 8, column 1).

Effects of Sugar-Content Information

Model 2 further reveals the role of the exogenously provided sugar content information. Three findings are notable. First, all treatments reduced consumers’ likelihood of selecting NFC and FC juices. The WTPs for “100% orange juice” in all treatment groups are lower than those in the control group (Table 8, first and second rows). A possible explanation is that after receiving the treatments, consumers realized that fruit juices also pose potential health risks due to their high sugar content, thereby reducing their willingness to consume “100% orange juices.” It bears noting that the short-video treatment led to the largest reduction in the WTP for “100% NFC” juices, by 30.28% [=(10.93-7.62)/10.93], among all four treatments (Table 8, first row).

Second, the coefficients of “sugar × T_l” suggest that only the participants receiving the “basic information + authoritative source (T₂)” and “video (T₄)” treatments increased their likelihood of choosing lower-sugar orange juices significantly. Compared with the mean WTP for sugar-free juice in the control group (11.51 RMB), that in the video-treatment group (17 RMB), indicates a maximum increase of 47.70%. Presumably, the video helped consumers convert a vague concept (sugar content) into a concrete one (the number of sugar cubes) when making choices. Our survey after the DCE choice tasks showed that only 45% of the “control” participants converted sugar content into sugar cubes when making choices, compared to 67% in the video-treatment group.

Another possible reason is that, with videos, one can passively assimilate information with less effort than reading text instructions, which may require effort to engage (—we thank an anonymous reviewer for pointing this out). According to dual coding theory, individuals process information through two systems: one for visual information and one for verbal information (Paivio, 2013). Both systems are functional when watching videos, but the visual information processing system is suppressed when reading text messages, which renders the information in videos easier to understand. Additionally, processing visual stimuli involves a more automatic process, requiring minimal effort; in contrast, reading text often demands more active thought and engagement (Green & Brock, 2002). The “effort” effect is also supported by the insignificant impact of treatment T₃, the “information + authoritative sources + specific figures.”T₃ obviously contains more detailed information, but the many figures involved in the message may create information overload, inducing consumer anxiety and decision-making errors (H. F. Hu & Krishen, 2019).

Finally, only the video intervention has affected the choice of taste. The statistically positive coefficients of the “sour × video” interactions suggest that the video treatment significantly reduces consumers’ rejection of sour juices: they are willing to sacrifice taste for health.

Discussion and Conclusion

As public health awareness improves, consumers are paying closer attention to the nutritional content of foods and beverages. However, when it comes to fruit juices, their high sugar content is often overlooked, possibly masked by a halo effect generated by the positive image of fruits. Motivated by documented consumer misunderstanding of nutrition labels, this study explored the role of online self-media as an alternative means of conveying health and nutrition information to consumers. Based on a discrete choice experiment conducted in Beijing, China, we found that exogenously provided sugar content information reduced consumers’ likelihood of choosing high-sugar fruit juice. Among the four treatments, a short video treatment consisting of a 1.5-min lecture delivered by a lab technician expressing the sugar content as the number of sugar-cube equivalents had the most pronounced impact. It is likely that the video helped consumers convert a vague concept (sugar content) into a concrete one (the number of sugar cubes) when making choices. It is also possible that, with videos, consumers can passively assimilate information with less effort than reading text instructions. Echoing this “effort effect,” a text treatment with more details and figures had a less significant effect than a more concise text treatment containing the same basic information from authoritative sources. These findings suggest that while the government should encourage the dissemination of nutrition and health knowledge from authoritative sources through online self-media to induce healthy food choices among consumers, the specific content should be concise yet authoritative.

With these analyses and findings, our study makes two contributions to the literature. First, to our knowledge, this study is among the few that focus on the sugar content and halo effect of fruit (orange) juices, laying the groundwork for correcting misperceptions about the nutrient content of juice products. Second, regarding consumers’ juice choices, this study is among the first to assess the effectiveness of online self-media—short educational videos in particular—as an information carrier, helping inform policies on the use of social media to disseminate nutrition knowledge and correct public misperceptions.

From a policy perspective, several measures can be taken to facilitate the use of nutrition-education videos as a vehicle for health interventions. First, videos on the sugar content of popular fruit juices (e.g., apple, orange, and lemonade) on the market may be created, especially under the supervision of public health authorities. Second, user profiling may be performed to identify target fruit-juice consumer groups with specific demographic characteristics (e.g., age, gender, and education level) or health risks (e.g., young children, older adults, and patients with chronic diseases) (Romano et al., 2025; Harshbarger et al., 2021). Third, algorithms that help match specific consumer groups and suitable nutrition-education videos may be developed. Finally, effective influencers (e.g., public health institutions, celebrities, and renowned nutrition experts) may be hired to disseminate nutrition-education videos to their followers and, if appropriate, to a wider audience, thereby raising public health awareness and stimulating active information-seeking for health-related content.

Yet, while promoting nutrition-education videos through self-media is desirable as a means of health intervention, challenges remain. First, high production costs pose a challenge in tailoring video content to fit all major products and individual consumers’ specific needs. Due to the wide variety of fruit-juice products and their varying sugar content, it would be highly costly to create videos tailored to each product. A potential solution is to develop AI-customizable videos to achieve cost-effective personalized dissemination. While this investment requires additional cost, we believe it is worthwhile because self-media interventions may help reduce healthcare costs and generate health benefits at the societal level (Gentili et al., 2022; Stanczyk et al., 2014).

Second, platform dependency may constrain the dissemination and promotion of intervention videos, as they rely on specific platforms. Given the significant heterogeneity in user-group characteristics across different platforms, this dependence may create barriers for scaling up video-based interventions. A possible solution is to adopt a de-platformization strategy (e.g., using WeChat Mini Programs rather than apps) to eliminate platform dependency and cross-platform barriers in the dissemination of video-based health information.

The third challenge is the difficulty public health institutions face in ensuring consistent quality across educational videos produced by different content creators. Providing the public with health-knowledge training, setting qualification standards for content creators, and building a multi-partner collaboration system may help address this challenge. The WHO, for example, recently announced a partnership with TikTok to promote public health awareness (Daramola et al., 2025).

Finally, “algorithmic recommendation,” originally designed to provide personalized information sets tailored to individual interests, has evolved into “information cocoons,” a phenomenon characterized by narrowed user information intake (X. Zhang et al., 2023). This challenge might be addressed from the perspectives of content and technology, which is also worth exploring in future research. Regarding content, sensitive vocabulary should be avoided, and expressions should be used in line with spoken language and popular topics. It is also necessary to update the content and keywords of video interventions in a timely and regular manner (Kilb et al., 2022). Regarding technology, building a multi-channel dissemination system (e.g., via social networks and email) could help break “information cocoons.” For example, a characteristic-driven distribution algorithm based on multidimensional health data (e.g., body weight and exercise habits) associated with demographic data could be combined with interest-dependent recommendation algorithms.

Before closing, a note on the limitations of this study is in order. First of all, the reliance on a third-party online platform to conduct our experiment has inherent drawbacks, as discussed in the “Survey and Data” section. Even though a series of steps was taken to ensure survey quality, concerns remain. For example, as an anonymous reviewer pointed out, some participants who received different treatments may have discussed their answers before completing their choice tasks, especially when they come from the same household or office. By checking the responses’ IP addresses, we found that 126 participants share IP addresses with other participants, suggesting they may be employees working in the same office building. As a robustness check, we re-estimated our mixed-logit models, excluding these observations. The results (Table A2) are quite comparable to our estimates with the full sample (Table 7). Although these overlapping IP addresses did not pose a serious threat to our key findings, they raise concerns about the quality of the data collection (e.g., random sampling) conducted by the third-party platform.

Second, there are currently no completely sugar-free juice products on the market. Sugar-free drinks mainly use sweeteners to mimic the flavor of sugar. However, given its focus on the role of online self-media, our DCE did not consider additives, sweeteners, and other attributes. While our randomized treatment design ensures that omitting these attributes will not bias the estimated impacts of sugar content and information treatments, future research examining their impact will be fruitful.

Third, although video and text messages are used to design treatments, the information provided is still somewhat different from that encountered in real-life self-media. Future research that designs interventions closer to real-life self-media use shall yield more insights into the role online self-media can play. For example, consumers are often repeatedly exposed to messages with the same content (e.g., commercials) on online self-media, which may lead to further changes in consumer perceptions of sugary drinks compared to our one-shot interventions. To better mirror such real-life social media consumption patterns, future research could employ repeated exposure designs in the choice experiment. For example, the online survey platform can use IP tracking technology to follow consumers who have already participated in the one-shot intervention, resend them the educational video and survey questionnaires with choice tasks (perhaps with varying frequencies across individuals), and record their responses in the choice tasks. That way, the impact of one-shot interventions and that of repeated exposure can be distinguished.

Finally, while our experiment has shed some light on how to reduce consumers’ likelihood of choosing high-sugar fruit juices with the presence of a halo effect using information treatments, it lacks the power to assess the extent to which our treatments actually reduced the halo effect. Thus, future research should develop appropriate methods and indicators to measure the halo effect and quantify changes in the effect induced by well-designed treatments.

The literature has provided some useful guidance. For example, perceptions of nutrient content (Fernan et al., 2018; Stoltze et al., 2021; Z. Wang & Begho, 2024), increased perceived benefits, and reduced perceived risks (Choi & Reid, 2015; Choi et al., 2013; Lim et al., 2021) have been used to measure the halo effect of food products. In consumers’ minds, fruit juices are often associated with fresh fruits, which are considered rich in vitamins that can help improve health, vitality, and immune function, thereby contributing to their popularity. Given this cognitive association, future studies may be designed to capture the halo effect by asking consumers about the vitamin content in fruit juices, the health benefits it provides, and its ability to help avoid illness. Additionally, measuring and comparing respondents’ evaluations of fruit juices and fresh fruits regarding health benefits could also be a way to assess the halo effect. Finally, nutrition claims may also help. In the case of fruit juices, the positive image of vitamins may induce consumers to choose juices labeled “high in vitamins” and ignore their high sugar content. Thus, by designing juice products with nutrition labels such as “high in vitamins,”“high in vitamins and sugar,” and “high sugar,” the differences in consumer perceptions of the sugar content with these labels may serve as a measure of the halo effect.

Footnotes

Appendix

Table A2.

Mixed-Logit Estimates.

Attributes/Levels	Model 1		Model 2
Attributes/Levels	Mean of parameter	SD of parameter	Mean of parameter	SD of parameter
100% NFC	1.045*** (0.079)	1.162*** (0.076)	1.452*** (0.167)	1.134*** (0.077)
100% FC	0.944*** (0.076)	1.058*** (0.079)	1.394*** (0.176)	1.067*** (0.803)
Juice content >10% (reference level: juice content >5%)	−0.069 (0.085)	1.114*** (0.095)	−0.192 (0.187)	1.130*** (0.096)
Sugar 0 g	1.307*** (0.093)	1.634*** (0.090)	1.105*** (0.171)	1.629*** (0.091)
Sugar 12 g	0.854*** (0.072)	0.848*** (0.079)	0.705*** (0.122)	0.847*** (0.080)
Sugar 24 g (reference level: sugar 36 g)	0.359*** (0.071)	0.712*** (0.104)	0.254*** (0.090)	0.743*** (0.101)
Sour	−0.619*** (0.113)	1.774*** (0.112)	−0.831*** (0.218)	1.809*** (0.113)
Slightly-sour	−0.048 (0.081)	0.954*** (0.097)	−0.134 (0.175)	0.985*** (0.099)
Balanced	0.366*** (0.075)	0.832*** (0.093)	0.237 (0.188)	0.842*** (0.093)
Slightly-sweet (reference level: sweet)	0.271*** (0.074)	0.579*** (0.107)	0.388** (0.170)	0.587*** (0.104)
100% NFC × basic information (T₁)			−0.312 (0.221)
100% NFC × basic information & authoritative source (T₂)			−0.465** (0.234)
100% NFC × information with specific figures & authoritative source (T₃)			−0.493** (0.240)
100% NFC × video (T₄)			−1.144*** (0.238)
100% FC × basic information (T₁)			−0.559** (0.228)
100% FC × basic information & authoritative source (T₂)			−0.512** (0.233)
100% FC × information with specific figures & authoritative source (T₃)			−0.544** (0.235)
100% FC × video (T₄)			−0.868*** (0.239)
Juice content >10% × basic information (T₁)			0.174 (0.246)
Juice content >10% × basic information & authoritative source (T₂)			0.314 (0.256)
Juice content >10% × information with specific figures & authoritative source (T₃)			0.072 (0.244)
Juice content >10% × video (T₄)			−0.115 (0.251)
Sugar × basic information (T₁)			0.006 (0.006)
Sugar × basic information & authoritative source (T₂)			−0.015** (0.007)
Sugar × information with specific figures & authoritative source (T₃)			0.005 (0.007)
Sugar × video (T₄)			−0.016** (0.007)
Sour × basic information (T₁)			0.215 (0.305)
Sour × basic information & authoritative source (T₂)			0.010 (0.325)
Sour × information with specific figures & authoritative source (T₃)			0.216 (0.331)
Sour × video (T₄)			0.797** (0.312)
Slightly-sour × basic information (T₁)			0.269 (0.236)
Slightly-sour × basic information & authoritative source (T₂)			−0.291 (0.243)
Slightly-sour × information with specific figures & authoritative source (T₃)			0.130 (0.239)
Slightly-sour × video (T₄)			0.506* (0.265)
Balanced × basic information (T₁)			−0.102 (0.247)
Balanced × basic information & authoritative source (T₂)			0.113 (0.239)
Balanced × information with specific figures & authoritative source (T₃)			0.009 (0.233)
Balanced × video (T₄)			0.373 (0.250)
Slightly-sweet × basic information (T₁)			−0.061 (0.229)
Slightly-sweet × basic information & authoritative source (T₂)			−0.164 (0.222)
Slightly-sweet × information with specific figures & authoritative source (T₃)			−0.418 (0.229)
Slightly-sweet × video (T₄)			0.125 (0.228)
Price	−0.115*** (0.008)		−0.117*** (0.009)
ASC	−0.292* (0.102)		−0.335*** (0.106)
Log-likelihood	−8,303.903		−6,668.148
AIC	13,456.8		13,444.3
Mcfadden R²	0.192		0.197
N	599		599

Note. Standard errors in parentheses.

***

p < .01, **p < .05, *p < .1.

Acknowledgements

The authors thank Junfei Bai, Juerong Huang, Xiaoou Liu, Yanjun Ren, Tian Xu, Qiran Zhao, and Chen Zhu, as well as participants of the 2022 annual meeting of Beijing Agricultural Economists Association and the 2024 Chinese Conference on Agricultural and Foresty Economics and Management, for their helpful suggestions on several early versions of this paper.

ORCID iD

Qihui Chen

Ethical Considerations

The study was conducted in accordance with the Declaration of Helsinki, and approved by the Human Research Ethics Committee of China Agricultural University (protocol code CAUHR-2020-05).

Consent to Participate

Informed consent was obtained from all subjects involved in the study.

Author Contributions

Jiale Bao and Qihui Chen—designed the study. Jiale Bao and Hongsha Wang—acquired the data, performed the research, and performed data analysis. Hongsha Wang and Qihui Chen—provided supervision and advice on data analysis and interpretation of findings. Jiale Bao, Qihui Chen, and Hongsha Wang—wrote and revised the manuscript. All authors have read and approved the final manuscript.

Funding

The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by the 2115 Talent Development Program of China Agricultural University. This research was funded by the National Natural Science Foundation, grant number 71973134. Support from the Beijing Food Safety Policy and Strategy Research Base is also gratefully acknowledged.

Declaration of Conflicting Interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Data Availability Statement

Data analyzed in this study can be requested from the corresponding author.

References

Abu Abed

Himmel

Vormfelde

Koschack

(2014). Video-assisted patient education to modify behavior: A systematic review. Patient Education and Counseling, 97(1), 16–22.

Acton

R. B.

Hammond

(2018). The impact of price and nutrition labelling on sugary drink purchases: Results from an experimental marketplace study. Appetite, 121, 129–137.

Acton

R. B.

Jones

A. C.

Kirkpatrick

S. I.

Roberto

C. A.

Hammond

(2019). Taxes and front-of-package labels improve the healthiness of beverage and snack purchases: A randomized experimental marketplace. International Journal of Behavioral Nutrition and Physical Activity, 16(1), 46.

Adams

J. M.

Hart

Gilmer

D. O.

Lloyd-Richardson

E. E.

Burton

K. A.

(2014). Concrete images of the sugar content in sugar-sweetened beverages reduce attraction to and selection of these beverages. Appetite, 83, 10–18.

Amaro

Cortes

D. E.

Garcia

Duan

Black

D. S.

(2017). Video-based grocery shopping intervention effect on purchasing behaviors among Latina shoppers. American Journal of Public Health, 107(5), 800–806.

André

Chandon

Haws

(2019). Healthy through presence or absence, nature or science? A framework for understanding front-of-package food claims. Journal of Public Policy & Marketing, 38(2), 172–191.

Aref

H. A. T.

Westrick

Chou

C. H.

Worthington

Garza

(2024). How to inform college students about meningitis B vaccine? Comparative effectiveness of an online theory-based text and video intervention. Journal of American College Health, 72(3), 970–979.

Assy

Nasser

Kamayse

Nseir

Beniashvili

Djibre

Grosovski

(2008). Soft drink consumption linked with fatty liver in the absence of traditional risk factors. Canadian Journal of Gastroenterology, 22(10), 811–816.

Bartneck

Duenser

Moltchanova

Zawieska

(2015). Comparing the similarity of responses received from studies in Amazon’s Mechanical Turk to studies conducted online and with direct recruitment. Plos One, 10(4), e0121595.

10.

Bazzano

L. A.

T. Y.

Joshipura

K. J.

F. B.

(2008). Intake of fruit, vegetables, and fruit juices and risk of diabetes in women. Diabetes Care, 31(7), 1311–1317.

11.

Beijing Municipal Bureau of Statistics. (2021). Special research report on the upgrade of Beijing residents’ online shopping consumption in 2021. https://tjj.beijing.gov.cn/tjsj_31433/sjjd_31444/202106/t20210622_2418248.html

12.

Bolling

Crosby

Boles

Stark

(2009). How pediatricians can improve diet and activity for overweight preschoolers: a qualitative study of parental attitudes. Academic Pediatrics, 9(3), 172–178.

13.

Bora

Das

Barman

Borah

(2018). Are Internet videos useful sources of information during global public health emergencies? A case study of YouTube videos during the 2015–16 Zika virus pandemic. Pathogens and Global Health, 112(6), 320–328.

14.

Casler

Bickel

Hackett

(2013). Separate but equal? A comparison of participants and data gathered via Amazon’s MTurk, social media, and face-to-face behavioral testing. Computers in Human Behavior, 29(6), 2156–2160.

15.

Cawley

Frisvold

Jones

(2020). The impact of sugar-sweetened beverage taxes on purchases: Evidence from four city-level taxes in the United States. Health Economics, 29(10), 1289–1306.

16.

Centurión

Machín

Ares

(2019). Relative impact of nutritional warnings and other label features on cereal bar healthfulness evaluations. Journal of Nutrition Education and Behavior, 51(7), 850–856.

17.

Chen

Caballero

Mitchell

D. C.

Loria

Lin

P.-H.

Champagne

C. M.

Elmer

P. J.

Ard

J. D.

Batch

B. C.

Anderson

C. A. M.

Appel

L. J.

(2010). Reducing consumption of sugar-sweetened beverages is associated with reduced blood pressure. Circulation, 121(22), 2398–2406.

18.

Cheung

K. L.

Schwabe

Walthouwer

M. J. L.

Oenema

Lechner

de Vries

(2017). Effectiveness of a video-versus text-based computer-tailored intervention for obesity prevention after one year: A randomized controlled trial. International Journal of Environmental Research and Public Health, 14(10), 1275.

19.

China Internet Network Information Center. (2022). Statistical report on the development of China’s internet. http://www.gov.cn/xinwen/2022-09/01/content_5707695.htm

20.

Choi

Reid

L. N.

(2015). Differential impact of message appeals, food healthiness, and poverty status on evaluative responses to nutrient-content claimed food advertisements. Journal of Health Communication, 20(11), 1355–1365.

21.

Choi

Yoo

Baek

T. H.

Reid

L. N.

Macias

(2013). Presence and effects of health and nutrition-related (HNR) claims with benefit-seeking and risk-avoidance appeals in female-orientated magazine food advertisements. International Journal of Advertising, 32(4), 587–616.

22.

Cornelsen

Quaife

Lagarde

Smith

R. D.

(2020). Framing and signalling effects of taxes on sugary drinks: A discrete choice experiment among households in Great Britain. Health Economics, 29(10), 1132–1147.

23.

Daramola

Ezeagu

C. N.

Aderinto

Olatunji

Kokori

(2025). Short videos, big impact: WHO and TikTok join forces. Discover Public Health, 22(1), 372.

24.

Defago

Geng

J. F.

Molina

Santa María

(2020). Can traffic light nutritional labels induce healthier consumer choices? Experimental evidence from a developing country. International Journal of Consumer Studies, 44(2), 151–161.

25.

Duffy

E. W.

Hall

M. G.

Dillman

C. F.

Musicus

A. A.

Meyer

M. L.

Rimm

Smith

T. L.

(2021). Nutrition claims on fruit drinks are inconsistent indicators of nutritional profile: a content analysis of fruit drinks purchased by households with young children. Journal of the Academy of Nutrition and Dietetics, 121(1), 36–46.e34.

26.

Eghtesadi

Florea

(2020). Facebook, Instagram, Reddit and TikTok: a proposal for health authorities to integrate popular social media platforms in contingency planning amid a global pandemic outbreak. Canadian Journal of Public Health-Revue Canadienne De Sante Publique, 111(3), 389–391.

27.

Fernan

Schuldt

J. P.

Niederdeppe

(2018). Health halo effects from product titles and nutrient content claims in the context of “protein” bars. Health Communication, 33(12), 1425–1433.

28.

Gentili

Failla

Melnyk

Puleo

Tanna

G. L. D.

Ricciardi

Cascini

(2022). The cost-effectiveness of digital health interventions: A systematic review of the literature. Frontiers in Public Health, 10, 787135.

29.

Green

M. C.

Brock

T. C.

(2002). Persuasion: Psychological insights and perspectives. Sage Publications.

30.

Hall

M. G.

Lazard

A. J.

Grummon

A. H.

Mendel

J. R.

Taillie

L. S.

(2020). The impact of front-of-package claims, fruit images, and health warnings on consumers’ perceptions of sugar-sweetened fruit drinks: Three randomized experiments. Preventive Medicine, 132, 105998.

31.

Harshbarger

Burrus

Rangarajan

Bollenbacher

Zulkiewicz

Verma

Galindo

C. A.

Lewis

M. A.

(2021). Challenges of and solutions for developing tailored video interventions that integrate multiple digital assets to promote engagement and improve health outcomes: Tutorial. JMIR Mhealth and Uhealth, 9(3), e21128.

32.

F. B.

(2013). Resolved: there is sufficient scientific evidence that decreasing sugar-sweetened beverage consumption will reduce the prevalence of obesity and obesity-related diseases. Obesity Reviews, 14(8), 606–619.

33.

H. F.

Krishen

A. S.

(2019). When is enough, enough? Investigating product reviews and information overload from a consumer empowerment perspective. Journal of Business Research, 100, 27–37.

34.

Ito-Jaeger

Vallejos

E. P.

Curran

Spors

Long

Y. F.

Liguori

Warwick

Wilson

Crawford

(2023). Digital video interventions and mental health literacy among young people: a scoping review. Journal of Mental Health, 31(6), 873–883.

35.

Jayalath

V. H.

de Souza

R. J.

Mirrahimi

Blanco-Mejia

Di Buono

Jenkins

A. L.

Leiter

L. A.

Wolever

T. M.

Beyene

Kendall

C. W.

Jenkins

D. J.

Sievenpiper

J. L.

(2015). Sugar-sweetened beverage consumption and incident hypertension: A systematic review and meta-analysis of prospective cohorts. American Journal of Clinical Nutrition, 102(4), 914–921.

36.

Jindahra

Phumpradab

(2023). Label copresence for healthier choices: How sugar content per daily limit and sugar warning labels balance out the health halos of nutrient-content claim. Food Quality and Preference, 107, 104853.

37.

Juhong

Mordmuang

Jewboonchu

Udomwech

(2022). Effectiveness of an online educational video intervention to improve the knowledge and behavior of contact lens care during the COVID-19 pandemic: A pre-test/post-test design. Heliyon, 8(10), e11009.

38.

Jürkenbeck

Mehlhose

Zühlsdorf

(2022). The influence of the Nutri-Score on the perceived healthiness of foods labelled with a nutrition claim of sugar. Plos One, 17(8), e0272220.

39.

Kilb

Dickhauser

Mata

(2022). A theory-based video intervention to enhance communication and engagement in online health communities: two experiments. Health Psychology and Behavioral Medicine, 10(1), 199–228.

40.

Lee

P. Y.

Lusk

Mirosa

Oey

(2016). Effect of information on Chinese consumers’ acceptance of thermal and non-thermal treated apple juices: A study of young Chinese immigrants in New Zealand. Food Quality and Preference, 48, 118–129.

41.

P. Y.

Chen

Deveaux

Luo

Y. M.

Tao

W. J.

W. M.

Wen

Zheng

(2022). Cross-verification of COVID-19 information obtained from unofficial social media accounts and associated changes in health behaviors: Web-based questionnaire study among Chinese netizens. JMIR Public Health and Surveillance, 8(5), e33577.

42.

Lim

K. H.

W. Y.

Nayga

R. M.

(2021). Consumer preference for grass-fed beef: A case of food safety halo effect. Journal of Agricultural and Resource Economics, 46(3), 447–463.

43.

Lindhjem

Navrud

(2011) Using internet in stated preference surveys: a review and comparison of survey modes. International Review of Environmental and Resource Economics, 5, 309–351.

44.

Liu

Cheng

Wan

L. J.

Chen

(2022). Total and added sugar intakes are increasing among children and adolescents in China: Findings from CHNS 1997–2011. Nutrients, 14(16), 3340.

45.

Louviere

J. J.

Hensher

D. A.

Swait

J. D.

(2000). Stated choice methods: Analysis and applications. Cambridge University Press.

46.

Luckow

Delahunty

(2004). Which juice is “healthier”? A consumer study of probiotic non-dairy juice drinks. Food Quality and Preference, 15(7–8), 751–759.

47.

Madathil

K. C.

Rivera-Rodriguez

A. J.

Greenstein

J. S.

Gramopadhye

A. K.

(2015). Healthcare information on YouTube: A systematic review. Health Informatics Journal, 21(3), 173–194.

48.

Malik

V. S.

Popkin

B. M.

Bray

G. A.

Despres

J. P.

F. B.

(2010). Sugar-sweetened beverages, obesity, type 2 diabetes mellitus, and cardiovascular disease risk. Circulation, 121(11), 1356–1364.

49.

Mauri

Grazzini

Ulqinaku

Poletti

(2021). The effect of front-of-package nutrition labels on the choice of low sugar products. Psychology & Marketing, 38(8), 1323–1339.

50.

Mcfadden

(1974). Conditional logit analysis of qualitative choice behavior. In Zarembka

(Ed.), Frontiers in Econometrics (pp. 105–142). Academic Press.

51.

Mehmet

Roberts

Nayeem

(2020). Using digital and social media for health promotion: A social marketing approach for addressing co-morbid physical and mental health. Australian Journal of Rural Health, 28(2), 149–158.

52.

Miller

Ettridge

Pettigrew

Wittert

Coveney

Eld

M. W. F.

Roder

Durkin

Martin

Kay

Dono

(2024). Warning labels for sugar-sweetened beverages and fruit juice: evaluation of 27 different labels on health effects, sugar content, energy and exercise equivalency. Public Health, 230, 138–148.

53.

Ministry of Commerce. (2020). The new trend of online shopping under the epidemic. http://www.mofcom.gov.cn/article/yqfkscbg/dfgz/202002/20200202934990.shtml

54.

Mullen

Archeveque

M. A.

Edwards

C. A.

Matsumoto

Crozier

(2008). Bioavailability and metabolism of orange juice flavanones in humans: Impact of a full-fat yogurt. Journal of Agricultural and Food Chemistry, 56(23), 11157–11164.

55.

Ning

L. W.

Niu

J. Y.

X. J.

Yang

Liu

Q. H.

Ning

Liang

L. B.

Liu

A. S.

Hao

Y. H.

Gao

L. J.

Liu

C. J.

(2020). The impacts of knowledge, risk perception, emotion and information on citizens’ protective behaviors during the outbreak of COVID-19: A cross-sectional study in China. BMC Public Health, 20(1), 1751.

56.

Ortega

D. L.

Hong

S. J.

Wang

H. H.

L. P.

(2016). Emerging markets for imported beef in China: Results from a consumer choice experiment in Beijing. Meat Science, 121, 317–323.

57.

Paivio

(2013). Imagery and verbal processes. Psychology Press.

58.

Pal

Holkar

Yevalkar

Bhattacharjee

(2017). Juice packaging design: effects of transparency on consumers’ perception leading toward purchase preference for packaged juice. Advances in Ergonomics in Design: (Ahfe 2017), 588, 177–184.

59.

Pan

X. F.

Wang

L. M.

Pan

(2021). Epidemiology and determinants of obesity in China. Lancet Diabetes & Endocrinology, 9(6), 373–392.

60.

Pereira-Caro

Borges

van der Hooft

J. D.

Clifford

M. N.

Del Rio

Lean

M. E. J.

Roberts

S. A.

Kellerhals

M. B.

Crozier

(2014). Orange juice (poly)phenols are highly bioavailable in humans. American Journal of Clinical Nutrition, 100(5), 1378–1384.

61.

Popkin

B. M.

Hawkes

(2016). Sweetening of the global diet, particularly beverages: Patterns, trends, and policy responses. Lancet Diabetes & Endocrinology, 4(2), 174–186.

62.

Revelt

Train

(1998). Mixed logit with repeated choices: Households’ choices of appliance efficiency level. Review of Economics and Statistics, 80(4), 647–657.

63.

Roe

Levy

A. S.

Derby

B. M.

(1999). The impact of health claims on consumer search and product evaluation outcomes: Results from FDA experimental data. Journal of Public Policy & Marketing, 18(1), 89–105.

64.

Romano

Abarca

Baehman

(2025). A low-cost, social media-supported intervention for caregivers to enhance toddlers’ language learning: Mixed methods feasibility and acceptability study. JMIR Pediatrics and Parenting, 8, e66175.

65.

Sabbe

Verbeke

Deliza

Matta

Van Damme

(2009). Effect of a health claim and personal characteristics on consumer acceptance of fruit juices with different concentrations of acai (Euterpe oleracea Mart). Appetite, 53(1), 84–92.

66.

Sah

Hillenbrand

Vogt

(2021). Visible sugar: Salient sugar information impacts health perception of fruit juices but only when motivated to be responsible and not when motivated to enjoy. Appetite, 164, 105262.

67.

Schouteten

J. J.

Gellynck

Slabbinck

(2019). Influence of organic labels on consumer’s flavor perception and emotional profiling: Comparison between a central location test and home-use test. Food Research International, 116, 1000–1009.

68.

Shefferly

Scharf

R. J.

DeBoer

M. D.

(2016). Longitudinal evaluation of 100% fruit juice consumption on BMI status in 2–5-year-old children. Pediatric Obesity, 11(3), 221–227.

69.

Song

S. J.

Zhao

Y. X .C.

Yao

X. L.

Z. C.

Zhu

Q. H.

(2021). Short video apps as a health information source: An investigation of affordances, user experience and users’ intention to continue the use of TikTok. Internet Research, 31(6), 2120–2142.

70.

Stanczyk

N. E.

Smit

E. S.

Schulz

D. N.

de Vries

Bolman

Muris

J. W. M.

Evers

(2014). An economic evaluation of a video- and text-based computer-tailored intervention for smoking cessation: a cost-effectiveness and cost-utility analysis of a randomized controlled trial. Plos One, 9(10), e110117.

71.

Stoltze

F. M.

Busey

Taillie

L. S.

Carpentier

F. R. D.

(2021). Impact of warning labels on reducing health halo effects of nutrient content claims on breakfast cereal packages: A mixed-measures experiment. Appetite, 163, 105229.

72.

Supady

Nguyen

Abd El Aziz

Ueberreiter

Baernighausen

Adam

(2023). A short, animated storytelling video about sodium intake as a major cardiovascular risk factor and recommendations for a healthy diet: An online, randomized, controlled trial. Trials, 24(1), 390.

73.

Sutterlin

Siegrist

(2015). Simply adding the word “fruit” makes sugar healthier: The misleading effect of symbolic information on the perceived healthiness of food. Appetite, 95, 252–261.

74.

Syed-Abdul

Fernandez-Luque

Jian

W. S.

Y. C.

Crain

Hsu

M. H.

Wang

Y. C.

Khandregzen

Chuluunbaatar

Nguyen

P. A.

Liou

D. M.

(2013). Misleading health-related information promoted through video-based social media: Anorexia on YouTube. Journal of Medical Internet Research, 15(2), e30.

75.

Thorndike

A. N.

Sonnenberg

Riis

Barraclough

Levy

D. E.

(2012). A 2-phase labeling and choice architecture intervention to improve healthy food and beverage choices. American Journal of Public Health, 102(3), 527–533.

76.

Tonsor

G. T.

Olynk

Wolf

(2009). Consumer preferences for animal welfare attributes: The case of gestation crates. Journal of Agricultural and Applied Economics, 41(3), 713–730.

77.

Train

K. E.

(2009). Discrete choice methods with simulation. Cambridge University Press.

78.

Tuong

Larsen

E. R.

Armstrong

A. W.

(2014). Videos to influence: A systematic review of effectiveness of video-based education in modifying health behaviors. Journal of Behavioral Medicine, 37(2), 218–233.

79.

Vandelanotte

Short

C. E.

Plotnikoff

R. C.

Rebar

Alley

Schoeppe

Canoy

D. F.

Hooker

Power

Oldmeadow

Leigh

Mummery

W. K.

Duncan

M. J.

(2021). Are web-based personally tailored physical activity videos more effective than personally tailored text-based interventions? Results from the three-arm randomised controlled TaylorActive trial. British Journal of Sports Medicine, 55(6), 336–343.

80.

Vandormael

Adam

Greuel

Gates

Favaretti

Hachaturyan

Bärnighausen

(2021). The effect of a wordless, animated, social media video intervention on COVID-19 prevention: online randomized controlled trial. JMIR Public Health and Surveillance, 7(7), e29060.

81.

Vidigal

Minim

V. P. R.

Carvalho

N. B.

Milagres

M. P.

Gonçalves

A. C. A.

(2011). Effect of a health claim on consumer acceptance of exotic Brazilian fruit juices: Acai (Euterpe oleracea Mart.), Camu-camu (Myrciaria dubia), Caja (Spondias lutea L.) and Umbu (Spondias tuberosa Arruda). Food Research International, 44(7), 1988–1996.

82.

Walker

R. W.

Dumke

K. A.

Goran

M. I.

(2014). Fructose content in popular beverages made with and without high-fructose corn syrup. Nutrition, 30(7–8), 928–935.

83.

Walthouwer

M. J. L.

Oenema

Lechner

de Vries

(2015). Comparing a video and text version of a web-based computer-tailored intervention for obesity prevention: A randomized controlled trial. Journal of Medical Internet Research, 17(10), e236.

84.

Wang

H. S.

Chen

Q. H.

Katare

(2023). Nudging Chinese consumers to embrace sustainable milk consumption: How should information be provided? Agribusiness, 39, 1512–1534.

85.

Wang

Begho

(2024). Investigating cognitive biases: Does halo effect from nutrition or health claims drive negative calorie illusion in food combinations? International Journal of Food Science and Technology, 59(1), 208–218.

86.

WHO (World Health Organization). (2015). Guideline: Sugars Intake for Adults and Children. https://apps.who.int/iris/bitstream/handle/10665/155735/WHO_NMH_NHD_15.2_chi.pdf;sequence=4

87.

Windle

Rolfe

(2011). Comparing responses from internet and paper-based collection methods in more complex stated preference environmental valuation surveys. Economic Analysis and Policy, 41(1), 83–97.

88.

D. H.

W. S.

Shi

Z. T.

You

Y. Q.

(2005). A study on consumer behavior and brand characteristics of juice in the urban areas of China. Journal of Qingdao Agricultural University, 4, 318–321. (In Chinese)

89.

Zenone

Silva

Smith

Lee

(2021). How does the British soft drink association respond to media research reporting on the health consequences of sugary drinks? Globalization and Health, 17(1), 72.

90.

Zhang

Q. R.

Liu

X. X.

X. Y.

Cao

Shi

(2021). Development and prospect analysis of compound fruit and vegetable juice beverages. Food and Fermentation Industries, 47(14), 94–299. (In Chinese)

91.

Zhang

Cai

Y. T.

Zhao

M. Q.

Zhou

(2023). Generation mechanism of “Information Cocoons” of network users: An evolutionary game approach. Systems, 11(8), 414.

92.

Zheng

(2014). Online sexual activity in Mainland China: Relationship to sexual sensation seeking and sociosexuality. Computers in Human Behavior, 36, 323–329.

93.

Zhu

Z. C.

Liu

S. Y.

Zhang

R. W.

(2023). Examining the persuasive effects of health communication in short videos: Systematic review. Journal of Medical Internet Research, 25, e48508.