Snack preferences among school-aged children: evidence from discrete choice experiments

Abstract

Background

Accelerated economic development and innovative food industry trends have contributed to worsening dietary imbalanced among Chinese school-aged population. To investigate snack preferences among school-aged children, determine the relative importance of different snack attributes, and explore how these preferences vary by social-demographic characteristics to inform targeted nutritional interventions.

Methods

A stratified cluster random sampling strategy was employed to recruit 854 school-aged children (grades 7 and 8) from Hubei and Anhui provinces. An evaluation framework comprising six attributes (taste, nutrient claims, purchase location, price, package size, and social influence) was constructed based on a Discrete Choice Experiment (DCE). A D-optimal design was implemented using SAS 9.4 software to generate 16 choice sets (divided into two versions), with two-stage questions incorporated to mitigate bias. A mixed logit model was applied to calculate preference coefficients (β), relative importance (RI), and willingness-to-pay (WTP) for different attributes.

Results

All six included attributes demonstrated statistically significant effects on local school-aged children’s snack choices (all P < 0.05). Specifically, school-aged children showed stronger preferences for snacks sold in supermarkets (β = 0.439), those rich in dietary fiber (β = 0.611), with sweet taste (β = 0.471), and commonly consumed by family members (β = 0.452). They were willing to pay an additional 46.49 CNY for snacks rich in dietary fiber. Large package size (β = -0.112) and price (β = -0.013) showed slight negative associations. Heterogeneity analysis revealed that boys preferred snacks rich in dietary fiber (β = 0.554) and were more price-sensitive (β = -0.026), while girls prioritized low-fat options (β = 0.738) and showed lower willingness to choose large packages (β = -0.259). School-aged children with less-educated parents placed greater emphasis on taste (RI = 25.33%), whereas those from highly-educated families valued package size more (RI = 10.43%). Overweight/obesity school-aged children preferred spicy snacks (β = 0.448) and were more price-sensitive (β = -0.115), while normal-weight school-aged children tended to choose sweet-tasting snacks (β = 0.510).

Conclusion

Snack preferences among school-aged children vary significantly by socio-demographic characteristics. Interventions should be tailored to the specific characteristics of them, while simultaneously engaging both the school-aged children and their parents to enhance participation and educational effectiveness, thereby reducing the purchase and consumption of unhealthy snacks.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12889-026-26883-7.

Introduction

In recent years, rapid socioeconomic development and dynamic transformations in the food industry have exacerbated the issue of imbalanced dietary patterns among school-aged children in China. Excessive consumption of snacks high in sugar, salt, and fat has not only accelerated the declining age trend of health problems such as obesity [1] and dental caries [2], but also emerged as a potential risk factor for chronic diseases including metabolic syndrome [3] and cardiovascular conditions [4]. Long-term intake of such snacks fosters a heightened preference for these products, reducing acceptance of healthy and balanced diets. Such dietary preferences may persist into adulthood, increasing the risk of chronic diseases in later life [5]. Although numerous studies have been conducted on children’s eating behaviors, existing research predominantly focuses on the isolated effects of single attributes (e.g., packaging [6]) on snack choices, with limited attention given to a systematic analysis of the formation mechanisms underlying school-aged children’s snack preferences from the perspective of multidimensional attribute interactions.

Against this background, this study employed a Discrete Choice Experiment (DCE) methodology combined with a stratified cluster random sampling strategy to recruit 854 school-aged children from Hubei and Anhui provinces for investigation. DCE is a quantitative research method grounded in the principles of utility maximization, attribute decomposition, and random utility theory, used to analyze individual or group preferences in different choice scenarios [7]. As a flexible preference analysis tool, this method enables dynamic simulation of multi-attribute interactions and has been widely applied in various fields such as health [8], marketing [9], and willingness-to-pay studies [10], thus compensating for the limitations of traditional questionnaires. However, the application of the DCE in the field of public health in China started relatively late, and there are notable limitations in directly extrapolating findings from international studies to explain the dietary preferences of the Chinese population [11–13]. Furthermore, existing research has predominantly focused on using DCE to analyze food choice preferences among adults [14, 15], which does not adequately inform effective intervention strategies for dietary behaviors among school-aged children in China. Based on the above, this study applies the DCE to analyze snack preferences among school-aged children in China. Heterogeneity analysis was conducted based on different demographic characteristics (gender, highest parental education level and BMI). The findings not only provide practical guidance for students’ parents, snack retailers, and policymakers but also offer a scientific basis for developing an integrated knowledge-attitude-practice intervention framework for student nutrition.

Methods

Data sources

The data of this study were derived from the “Research on Dietary Preferences and Intervention Strategies for School-Aged Children”. Using a stratified cluster random sampling method, one urban middle school and one rural middle school were selected from both Hubei and Anhui provinces as survey sites. Each school was stratified by grade (Grades 7 and 8), with at least 100 students recruited from each stratum to ensure a minimum sample size of ≥ 200 per school. Paper-based questionnaires (Supplementary file 1–3) were administered for on-site data collection. A total of 863 questionnaires were returned, among which 854 were valid, yielding an effective response rate of 98.96%. Informed consent was obtained from all subjects and their parents or legal guardians involved in the study. The study protocol was approved by the Ethics Committee of the National Institute for Nutrition and Health, Chinese Center for Disease Control and Prevention (Ethics Approval No.: 2024–007). This research was performed in compliance with the World Medical Association’s Declaration of Helsinki, which is a statement of ethical principles for medical research involving human participants.

Research methods

The study comprised two components: an individual questionnaire survey (Supplementary file 1) and a snack preference questionnaire survey (Supplementary file 2 and 3). The discrete choice experiment (DCE) method was employed to design and conduct the investigation of snack preferences among school-aged children.

The experimental design included the following components:

1. The definition of snack and identification of their attributes and levels: In this study, Snacks refer to non-staple food or beverage items consumed outside of main meals (i.e., breakfast, lunch, and dinner) [16], typically denoting commercially available, pre-packaged, ready-to-eat food products. Through literature analysis, expert consultation, and group discussions, six snack attributes were identified (taste, nutrient claims, purchase location, price, package size, and social influence) [13, 17–24]. Each attribute contained 3–4 levels [25] (See Supplementary file 2 and 3).

2. Discrete choice experiment design: The DCE, grounded in random utility theory and characteristic demand theory, is a key method in stated preference research [26]. By designing choice sets comprising different combinations of key attribute levels, respondents were able to make trade-offs based on their preferences. A fractional factorial design was used to optimize the choice sets and reduce the number of alternatives [27]. The %ChoicEff macro in SAS 9.4 was applied to implement a D-optimal design [28], generating 16 choice sets, each containing two alternatives. The 16 choice sets were divided into two versions (eight sets per version), each including two options per sets (See Supplementary file 2 and 3). Respondents were randomly assigned one version during the survey.

3. Questionnaire content: The survey also collected general demographic information (sex, age, highest parental education level, height, weight, etc.). Body mass index (BMI) was calculated using the formula BMI = weight (kg)/[height (m)]², with 18.5 ≤ BMI < 25.0 defined as normal weight and BMI ≥ 25 defined as overweight/obesity [29]. To avoid overestimating experimental attributes due to the lack of an opt-out option or insufficient preference information resulting from excessive opt-out selections [30], a two-stage question approach was adopted. Question 1 asked respondents to choose their preferred alternative, and Question 2 inquired whether they would actually choose that option in real life, thereby reducing biases caused by forced choices or opt-out preferences (Table 1).

4. Sample size: Based on the rule of thumb for sample size calculation, N ≥ 500c/t × a [31], where c is the largest number of levels in any attribute, t is the number of choice tasks, and a is the number of alternatives per task, the required sample size was determined as N ≥ 125(with c = 4, t = 8, and a = 2 in this study). To meet the precision requirements for subgroup analyses (e.g., sex, highest parental education level, obesity status), the sample size was increased to 800.

5. Quality Control: The questionnaire content (including snack attributes and levels) was developed based on field investigations and reviewed by an expert panel, with its structure and items further refined through two rounds of expert discussions to ensure scientific validity and operational feasibility. Prior to the survey, all investigators received standardized training to ensure adherence to uniform procedures. Data collection was conducted centrally on a class-by-class basis, with two investigators assigned to each class; two versions of the preference questionnaire were randomly distributed. A standardized briefing was provided before the survey to ensure consistent comprehension among respondents. Researchers were present throughout the process to address any questions that arose during completion and repeatedly emphasized that participants should respond based on their genuine preferences, without concern for a “correct answer.” Upon collection, investigators performed on-site verification to confirm the absence of omissions or logical errors; questionnaires were coded only after confirmation.

Table 1.

Example of a snack preference choice set (DCE)

Statistical analysis

This study utilized R 4.3.3 for data coding and preprocessing. Based on random utility theory U_ij = V_ij + Ɛ_ij, V_ij, (Where U_ij represents the total utility of individual i for alternative j, V_ij denotes the systematic fixed utility, and Ɛ_ij is the random utility, X_n represents observable attributes), the “price” attribute of snacks was treated as a continuous variable using linear encoding, while all other attributes were encoded as categorical variables using dummy coding. The normal distribution was assumed for all parameters, Stata 17.0 software was employed to establish a mixed logit model to calculate the preference coefficients (), relative importance (RI), and willingness-to-pay (WTP) for different snack attributes.

Specifically, , Where max is the maximum utility of an attribute and min is the minimum utility; , where is the attribute coefficient and is the price attribute coefficient [32].

The magnitude and direction of the regression coefficient β in the model results reflect the degree and direction of the influence of each attribute on school-aged children’s snack preferences. A positive coefficient indicates a positive preference for the attribute, while a negative coefficient indicates a negative preference. Based on the mixed logit model, the nlcom command in Stata 17.0 was used to estimate the marginal WTP and its 95% confidence interval for different snack attributes among school-aged children, with P < 0.05 indicating statistical significance.

Participants characteristics

The study included a total of 854 participants in Anhui and Hubei provinces, all of whom were school-aged children with a mean age of 13.3 years. Demographic characteristics showed a balanced gender distribution: boys accounted for 55.04% (n = 470), and girls comprised 44.96% (n = 384). Age distribution was relatively homogeneous: Grade 7 students represented 50.23% (n = 429), while Grade 8 students accounted for 49.77% (n = 425). The proportion of school-aged children with normal weight was relatively large (63.93%, n = 546), while overweight/obese school-aged children constituted a smaller proportion (36.07%, n = 308). In terms of socioeconomic characteristics, the highest parental education level was predominantly junior high school or below (37.59%, n = 321), with a small proportion holding postgraduate or higher degrees (5.97%, n = 51); data were missing for 9.37% (n = 80) of participants. Regarding daily living arrangements, the primary caregivers were mostly parents (92.15%, n = 787), and students boarding at school constituted a minority (19.56%, n = 167). However, collective school dining was widely adopted (97.54%, n = 833). Economic support and nutritional awareness indicators revealed that most students had disposable pocket money (78.34%, n = 669) and were aware of nutrition labels on food packages (78.57%, n = 671) Table 2.

Table 2.

Baseline characteristics of the study participants (n = 854)

Variable	Number(n)	Proportion(%)
Gender
Males	470	55.04
Females	384	44.96
Age(Years)
Grade 7	429	50.23
Grade 8	425	49.77
School Location
Rural	409	47.89
Urban	445	52.11
BMI
Normal	546	63.93
Overweight/Obesity	308	36.07
Highest Parental Education Level
Junior School or Below	321	37.59
High School/Vocational School	220	25.76
University/College	182	21.31
Postgraduate or Above	51	5.97
Unknown	80	9.37
Primary Caregiver
Parent(s)	787	92.15
Grandparent(s)	57	6.67
Nanny/Other	8	0.94
Unknown	2	0.23
Boarding at School
Yes	167	19.56
No	687	80.44
Eating Collective Meals at School
Yes	833	97.54
No	21	2.46
Having Discretionary Pocket Money
Yes	669	78.34
No	185	21.66
Awareness of Nutrition Labels
Yes	671	78.57
No	183	21.43

Results

Snack preferences among school-aged children

The analysis included 854 participants, yielding 13,664 observed data points. The model demonstrated good overall fit (log-likelihood ratio = −4377.84; Likelihood Ratio χ² = 244.37; P < 0.001), indicating its effectiveness in explaining school-aged children’s snack choice behaviors in Anhui and Hubei provinces.

Overall, all six attributes demonstrated statistically significant effects on school-aged children’s snack choices in Anhui and Hubei provinces (P < 0.05). The relative importance of attributes in descending order was: nutrient claims (34.13%), taste (22.89%), social influence (19.72%), purchase location (16.17%), package size (3.70%), and price (3.39%).

Specifically, school-aged children in Anhui and Hubei provinces demonstrated a preference for snacks that are commonly consumed by family members (β = 0.452, P < 0.001), purchased in supermarkets (β = 0.439, P < 0.001), with a sweet taste (β = 0.471, P < 0.001), and rich in dietary fiber (β = 0.611, P < 0.001) or vitamins/minerals (β = 0.463, P < 0.001). In contrast, package size and price had relatively smaller influences on preferences, with large package size (β = −0.112, P = 0.019) and price (β = −0.013, P = 0.007) showing slight negative associations. Additionally, the effects of salty taste and medium package size were not statistically significant (P > 0.05) Table 3.

Table 3.

Analysis of snack preferences based on mixed logit models

Attributes and levels	β	SD	P	RI(%)
Taste				22.89
Original Flavor	Ref
Sweet Flavor	0.471	0.067	< 0.001
Savory Flavor	0.015	0.053	0.771
Spicy Flavor	0.399	0.072	< 0.001
Nutrient Claims				34.13
None	Ref
Low-fat	0.399	0.058	< 0.001
Rich Minerals/Vitamins	0.463	0.055	< 0.001
High Dietary Fiber	0.611	0.055	< 0.001
Purchase Location				16.17
Near School	Ref
Convenience Store	0.098	0.050	0.047
Supermarket/Hypermarket	0.439	0.048	< 0.001
Social Influence				19.72
Advertising/Influencer Recommendations	Ref
Peer Recommendations	0.253	0.041	< 0.001
Family Consumption Habits	0.452	0.048	< 0.001
Package Sizea				3.7
Small	Ref
Medium	0.010	0.048	0.830
Large	−0.112	0.048	0.019
Price	−0.013	0.005	0.007	3.39
Sample Size	854
Number of Observations	13,664
Log Likelihood	−4377.84
LRchi2(12)	244.37
Prob > chi2	0.00

^aSmall:15 ~ 100 g; medium:101 ~ 300 g; large: > 300g

Heterogeneity analysis of snack preferences among school-aged children with different characteristics

Analysis of snack preferences by gender

Among school-aged children in Anhui and Hubei provinces, the relative importance ranking of snack attributes was consistent across genders. However, compared to girls, boys showed stronger influences from social influence (RI = 21.01%) and price (RI = 1.04%), demonstrating a marked preference for snacks rich in dietary fiber (β = 0.554, P < 0.01). Package size did not significantly affect their preferences, while they exhibited a distinct negative preference for higher prices (β = −0.026, P < 0.01). In contrast, girls placed greater emphasis on package size (RI = 8.02%) and preferred low-fat snacks (β = 0.738, P < 0.01). They showed a significant negative preference for large package sizes (β = −0.259, P < 0.01), while price had no statistically significant effect on their snack choices. See Table 4.

Table 4.

Analysis of heterogeneity in snack preferences by gender among school-aged children (n = 854)

Attributes and levels	Boys		Girls
	β(SD)	RI(%)	β(SD)	RI(%)
Taste		24.30		21.41
Original Flavor	Ref		Ref
Sweet Flavor	0.423**(0.095)		0.540**(0.100)
Savory Flavor	0.037(0.072)		−0.030(0.082)
Spicy Flavor	0.334**(0.095)		0.521**(0.111)
Nutrient Claims		34.96		36.34
None	Ref		Ref
Low-fat	0.162*(0.078)		0.738**(0.097)
Rich Minerals/Vitamins	0.481**(0.075)		0.472**(0.085)
High Dietary Fiber	0.554**(0.076)		0.718**(0.087)
Purchase Location		15.97		16.37
Near School	Ref		Ref
Convenience Store	0.030(0.068)		0.187*(0.075)
Supermarket/Hypermarket	0.385**(0.065)		0.507**(0.078)
Social Influence		21.01		17.83
Advertising/Influencer Recommendations	Ref		Ref
Peer Recommendations	0.180**(0.055)		0.366**(0.065)
Family Consumption Habits	0.436**(0.065)		0.471**(0.072)
Package Sizea		2.72		8.02
Small	Ref		Ref
Medium	0.064(0.066)		−0.070(0.076)
Large	0.008(0.066)		−0.259**(0.074)
Price	−0.026**(0.007)	1.04	0.001(0.007)	0.03
Sample Size	470		384
Number of Observations	7520		6144
Log Likelihood	−2414.94		−1931.38
LRchi2(12)	142.75**		106.33**

^aSmall:15 ~ 100 g; medium:101 ~ 300 g; large: > 300g

**p < 0.01,*p < 0.05

Analysis of snack preferences by parental education level

Based on parental education levels, among school-aged children in Anhui and Hubei provinces, whose parents had a junior high school education or below placed greater emphasis on taste (RI = 25.33%) and nutrient claims (RI = 37.50%). For those with parents holding a high school or vocational secondary education, social influence (RI = 25.46%) was considered more important than taste (RI = 21.55%). In contrast, those parents with college or higher education levels showed greater concern for package size (RI = 10.43%).

In Anhui and Hubei provinces, school-aged children whose parents had a junior high school education or below demonstrated a preference for sweet-tasting snacks (β = 0.485, P < 0.01) and those rich in dietary fiber (β = 0.596, P < 0.01), while showing a slight negative preference for price (β = −0.017, P < 0.05). In contrast, school-aged children with parents holding a high school or vocational secondary education preferred spicy-tasting snacks (β = 0.434, P < 0.01). Meanwhile, school-aged children of parents with college or higher education exhibited stronger preferences for sweet taste (β = 0.529, P < 0.01) and snacks rich in vitamins/minerals (β = 0.667, P < 0.01), along with a significant negative preference for large package sizes (β = −0.325, P < 0.01). Price did not significantly influence their preferences. See Table 5.

Table 5.

Analysis of heterogeneity in snack preferences by parental education level among school-aged children (n = 774)

Attributes and levels	Junior School or Below		High School/Vocational School		University/College or Above
	β(SD)	RI(%)	β(SD)	RI(%)	β(SD)	RI(%)
Taste		25.33		21.55		20.84
Original Flavor	Ref		Ref		Ref
Sweet Flavor	0.485**(0.110)		0.429**(0.150)		0.529**(0.136)
Savory Flavor	−0.047(0.088)		0.026(0.112)		−0.003(0.108)
Spicy Flavor	0.407**(0.116)		0.434**(0.144)		0.477**(0.158)
Nutrient Claims		37.50		33.58		33.62
None	Ref		Ref		Ref
Low-fat	0.354**(0.098)		0.319*(0.124)		0.494**(0.122)
Rich Minerals/Vitamins	0.423**(0.091)		0.441**(0.118)		0.667**(0.123)
High Dietary Fiber	0.596**(0.093)		0.651**(0.117)		0.644**(0.113)
Purchase Location		16.55		17.39		16.58
Near School	Ref		Ref		Ref
Convenience Store	0.041(0.088)		0.138(0.101)		0.101(0.100)
Supermarket/Hypermarket	0.414**(0.078)		0.465**(0.113)		0.559**(0.111)
Social Influence		17.66		25.46		17.91
Advertising/Influencer Recommendations	Ref		Ref		Ref
Peer Recommendations	0.171*(0.070)		0.359**(0.087)		0.339**(0.084)
Family Consumption Habits	0.378**(0.076)		0.602**(0.104)		0.489**(0.097)
Package Sizea		2.30		1.48		10.43
Small	Ref		Ref		Ref
Medium	0.066(0.082)		0.050(0.104)		−0.117(0.101)
Large	−0.055(0.079)		−0.009(0.103)		−0.325**(0.100)
Price	−0.017*(0.008)	0.65	−0.020*(0.010)	0.65	0.023(0.045)	0.63
Sample Size	321		220		233
Number of Observations	5136		3520		3728
Log Likelihood	−1649.82		−1120.30		−1174.74
LRchi2(12)	98.21**		66.25**		88.66**

^aSmall:15 ~ 100 g; medium:101 ~ 300 g; large: > 300g

**p < 0.01,*p < 0.05

Analysis of snack preferences by BMI

In Anhui and Hubei provinces, compared to overweight/obese school-aged children, those with normal weight placed greater emphasis on purchase location (RI = 19.01%) and preferred snacks with sweet taste (β = 0.510, P < 0.01) that were rich in dietary fiber (β = 0.612, P < 0.01). Package size and price did not significantly influence their preferences. In contrast, overweight/obese school-aged children were more influenced by nutrient claims (RI = 39.47%) and price (RI = 3.84%), selecting snacks that were similarly rich in dietary fiber (β = 0.684, P < 0.01) but with a spicy taste (β = 0.448, P < 0.01). They also exhibited a negative preference for price (β = −0.115, P < 0.01). See Table 6.

Table 6.

Analysis of heterogeneity in snack preferences by BMI among school-aged children (n = 854)

Attributes and levels	Normal BMI		Overweight/Obesity
	β(SD)	RI(%)	β(SD)	RI(%)
Taste		24.00		21.60
Original Flavor	Ref		Ref
Sweet Flavor	0.510**(0.087)		0.434**(0.117)
Savory Flavor	0.032(0.064)		−0.035(0.099)
Spicy Flavor	0.396**(0.093)		0.448**(0.119)
Nutrient Claims		33.94		39.47
None	Ref		Ref
Low-fat	0.381**(0.075)		0.441**(0.104)
Rich Minerals/Vitamins	0.443**(0.072)		0.557**(0.100)
High Dietary Fiber	0.612**(0.072)		0.684**(0.101)
Purchase Location		19.01		9.58
Near School	Ref		Ref
Convenience Store	0.165**(0.062)		−0.092(0.091)
Supermarket/Hypermarket	0.491**(0.062)		0.333**(0.085)
Social Influence		19.59		19.70
Advertising/Influencer Recommendations	Ref		Ref
Peer Recommendations	0.284**(0.052)		0.205**(0.074)
Family Consumption Habits	0.449**(0.060)		0.488**(0.084)
Package Sizea		2.70		5.81
Small	Ref		Ref
Medium	0.039(0.061)		−0.032(0.084)
Large	−0.101(0.06)		−0.158(0.088)
Price	−0.023(0.027)	0.76	−0.115**(0.039)	3.84
Sample Size	546		308
Number of Observations	8736		4928
Log Likelihood	−2789.44		−1575.38
LRchi2(12)	157.86		100.81

^aSmall:15 ~ 100 g; medium:101 ~ 300 g; large: > 300g

**p < 0.01,*p < 0.05

Willingness-to-pay analysis

Substantial heterogeneity was observed in school-aged children’s willingness-to-pay for different snack attributes in Anhui and Hubei Provinces. Specifically, snacks rich in dietary fiber commanded the highest premium (WTP = 46.49 CNY). Regarding taste, school-aged children were willing to pay 35.83 CNY more for sweet-tasting snacks compared to original alternatives. For social influence, snacks commonly consumed by family members yielded a 34.40 CNY higher WTP than those promoted by advertisements or internet celebrities. In terms of purchase location, school-aged children demonstrated a 33.40 CNY greater WTP for snacks available in supermarkets compared to those sold near schools. It should be noted that package size did not exert a statistically significant influence on WTP (P > 0.05). See Fig. 1.

Fig. 1.

Willingness-to-pay analysis of snack preferences (n = 854)

Sensitivity analysis

This study’s Model I included all 13,664 observations, while Model II excluded 3,276 observations from respondents who opted out (selecting “would not choose”) in Question 2. Mixed logit regression analyses were conducted for both models, revealing statistically significant goodness-of-fit for each (P < 0.001). Furthermore, the regression coefficients for snack attribute levels in both models were consistent in direction, with negligible differences in magnitude and statistical significance. This indicates that the exclusion of opt-out observations did not substantially alter the overall regression results. Collectively, these findings demonstrate the robustness and validity of all analytical outcomes. See Table 7.

Table 7.

Results of sensitivity analysis for snack preferences based on mixed logit models

Attributes and levels	Model I		Model II
	β	SD	β	SD
Taste
Original Flavor	Ref		Ref
Sweet Flavor	0.471**	0.067	0.626**	0.078
Savory Flavor	0.015	0.053	0.060	0.064
Spicy Flavor	0.399**	0.072	0.528**	0.083
Nutrient Claims
None	Ref		Ref
Low-fat	0.399**	0.058	0.482**	0.071
Rich Minerals/Vitamins	0.463**	0.055	0.450**	0.066
High Dietary Fiber	0.611**	0.055	0.599**	0.067
Purchase Location
Near School	Ref		Ref
Convenience Store	0.098*	0.050	0.112	0.058
Supermarket/Hypermarket	0.439**	0.048	0.444**	0.057
Social Influence
Advertising/Influencer Recommendations	Ref		Ref
Peer Recommendations	0.253**	0.041	0.338**	0.050
Family Consumption Habits	0.452**	0.048	0.577**	0.057
Package Sizea
Small	Ref		Ref
Medium	0.010	0.048	0.015	0.056
Large	−0.112*	0.048	−0.069	0.057
Price	−0.013**	0.005	−0.016**	0.006
Sample Size	854		854
Number of Observations	13,664		10,388
Loglikelihood	−4377.84		−3289.73
LRchi2(12)	244.37		172,79
Attributes and levels	0.00		0.00

Model I comprised observations from the full sample, while Model II included only those observations that selected “Yes” in Question 2

^aSmall:15 ~ 100 g; medium:101 ~ 300 g; large: > 300g

**p < 0.01,*p < 0.05

Discussion

This study systematically assessed the key attribute levels and heterogeneity in snack preferences among school-age children in Anhui and Hubei provinces through a DCE. It identifies nutrient claims, taste, and social influence as the core factors driving snack choices among school-age children. The findings are consistent with those of existing literature [33, 34], while also revealing characteristics specific to the Chinese context, thereby providing new insights for designing targeted interventions.

Overall, school-age children in Anhui and Hubei provinces prefer to purchase sweet snacks, high dietary fiber, family consumption habits, and in smaller packages, mainly in supermarkets, while showing a negative preference for price. Batista [35] noted that positive nutrition labeling is a key driver of healthy food choices. In recent years, as nutrition labeling has become more widespread, school-age children have developed a certain level of dietary knowledge (in this study, approximately 80% were aware of nutrition labels) and have higher nutritional expectations for snacks, with a growing number opting for healthier options. According to The Snack Guidelines for Chinese Children and Adolescents (2018) [36] and related data [37], the most commonly consumed snacks among Chinese children include fruit (56.3%), dairy products (25.7%), sugar-sweetened beverages (12.2%), instant foods (4.7%), and grain-based snacks (1.7%), which also indicates a trend toward healthier snack choices among school-age children. The WTP analysis in this study revealed that local school-age children are willing to pay an additional 46.49 CNY for snacks rich in dietary fiber. This suggests that manufacturers and retailers should prioritize supply-side reform, emphasizing “nutrition” as a core competitive advantage by producing or selling nutrient-dense snacks and clearly highlighting key nutritional components on labels. Given that children in these regions prefer purchasing snacks in supermarkets—likely due to the wider variety available [38]—the government could encourage supermarkets to establish dedicated sections for nutritious snacks, categorized by nutritional claims, to attract both children and their parents. However, it is important to note that this study simulated a real shopping scenario. On one hand, participants did not actually pay for the snacks; on the other, as minors, their snack purchasing decisions and behaviors are influenced by guardians [39] (who are the final payers in real settings). This may have reduced children’s sensitivity to price attributes, leading to an overestimation of WTP results. Therefore, the WTP findings do not reflect actual market conditions but rather indicate the strength of children’s preferences for different snack attributes in an ideal context. In a study conducted in a U.S. middle school, Rusmevichientong et al. [13] found that students were more likely to choose snacks commonly consumed by their families, even if the snacks were low in nutritional value. This aligns with Gibson [40], who posits that children’s snacking behaviors are significantly influenced by their parents, and helps explain why school-age children in this study preferred snacks frequently purchased by their families. This underscores that the family environment remains a critical leverage point for fostering healthy eating habits among school-age children [22, 41]. Enhancing parents’nutritional literacy is essential for shaping positive dietary behaviors in children. As primary caregivers, parents should model healthy snacking behaviors and avoid providing unhealthy snacks. Finally, a sensitivity analysis was conducted by excluding children who chose the opt-out option, confirming the robustness of the study findings.

In Anhui and Hubei provinces, subgroup analysis revealed differences in snack preferences among school-age children with different characteristics. Boys showed a preference for snacks high in dietary fiber, while girls paid more attention to low-fat options and smaller packages. Studies indicate that people generally hold both implicit and explicit associations linking males with large portions and females with small portions [42], such gender stereotypes may originate from sociocultural expectations regarding eating behaviors for men and women. Such gender stereotypes may stem from sociocultural expectations regarding eating behaviors for men and women. In this study, girls in adolescence often face pressure to maintain a slim physique, making them particularly sensitive to weight management and dietary choices to avoid feelings of guilt from overeatin; School-age children with less-educated parents tended to prefer snacks rich in dietary fiber, whereas those from highly educated families considered dietary fiber and mineral/vitamin content equally important. Moreover, children from highly educated families showed greater concern for package size and a slightly positive preference for price, possibly because parental education influences food choices indirectly by shaping health awareness [43]. Due to the intergenerational transmission of health consciousness, children in highly educated households consider multiple attributes when selecting snacks. This further underscores the importance of parents in fostering healthy eating habits among school-age children; Local overweight/obesity school-age children displayed nearly equal preferences for spicy and sweet snacks, which may be related to sensory-seeking behavior [44]. The pronounced preference for spicy taste could also be attributed to the widespread consumption of spicy snacks such as spicy strips and puffed foods in Chinese local dietary culture [45, 46]. Furthermore, a review [47] noted that frequent or high consumption of spicy foods is positively associated with overweight/obesity compared to infrequent or no consumption. One potential explanation is that spicy food intake may increase cravings for sweets, leading to significant weight gain [48], which is consistent with the findings of this study. However, the above explanations remain speculative, and the precise underlying mechanisms linking spicy food consumption to overweight/obesity are still unclear. Further research is needed to explore these mechanisms.

The Snack Guidelines for Chinese Children and Adolescents (2018) [36] provide specific recommendations for school-aged children: snacks should supplement main meals and be consumed in small quantities (not exceeding 10% of daily energy intake); snacks during breaks should prioritize fruits, dairy, and nuts (avoiding high-sugar beverages and fried foods); high-salt, high-sugar, and high-fat snacks should be limited, and alcohol- or caffeine-containing beverages should be avoided; oral hygiene should be maintained (rinsing or brushing after snacking), and no snacks should be consumed within one hour before bedtime. Based on the findings of this study, we also propose the following recommendations: (1) Strengthen health education for parents of school-age children in Anhui and Hubei provinces. Schools should integrate food safety and nutrition education into daily teaching activities. Through channels such as parent-teacher meetings and public health courses, a “home-school collaboration” model can be implemented to provide parents with nutrition education opportunities. This approach will enable both students and parents to make informed snack choices using nutrition labels or claims and cultivate healthy dietary habits; (2) Leverage the exemplary role of families. Capitalizing on the strong influence of “family eating habits,” families should jointly establish a healthy snack list. As primary caregivers, parents should consciously increase their consumption of healthy snacks such as fruits, yogurt, and nuts; (3) Optimize the school snack environment. Given children’s preference for purchasing snacks in supermarkets and for smaller packages, schools are encouraged to collaborate with reputable supermarkets to install vending machines on campus offering snacks in smaller portions. Snacks with nutritional claims could be labeled as “Nutrition Stars.” The sale of snacks high in sugar, salt, or fat on campus should be strictly prohibited; (4) Improve the social marketing environment. As noted by Qian et al. [49], the lack of access to supermarkets, which typically provide healthier food options, has been identified as a risk factor for childhood obesity in several studies. Since our findings indicate that school-age children prefer buying snacks in supermarkets, supermarkets should offer more healthy snack options to help mitigate the local trend of childhood obesity; (5) Reduce added sugars (artificial sweeteners). While this study reveals a widespread preference for sweet snacks among local children, the harms of added sugars are well-documented. For health reasons, snack manufacturers should consider innovating and reformulating products by replacing artificial sweeteners with natural alternatives. Additionally, parents can prepare naturally sweet snacks at home for their children; (6) Provide targeted snack interventions and guidance tailored to different subgroups of school-age children. In this study, 97.54% of children had meals collectively at school. For those already overweight or obese, interventions can leverage their dietary characteristics—such as a preference for spicy flavors and price sensitivity. School cafeterias could offer small portions of healthy snacks made with natural spices at discounted prices, with appropriate purchase limits. Clear guidance on healthy eating should be provided, and parents should be advised to reasonably control the proportion of pocket money allocated for snacks. These measures can help slow weight gain and support weight management.

In summary, this study employed DCE to examine snack preferences among school-age children in Anhui and Hubei provinces. The findings provide an empirical basis for designing targeted snack intervention strategies. The recommendations derived from this study can also be extended to other regions with similar contexts, thereby contributing to the dietary health of school-age children.

Limitations

The survey was limited to Anhui and Hubei provinces, resulting in limited national representativeness, the findings are influenced by regional dietary habits and should be extrapolated with caution; Attributes such as food additives, packaging style, and brand were not included, potentially underestimating school-aged children’s attention to technological food properties; Considering statistical efficiency, subgroup analyses were not performed for every attribute; Due to the inherent hypothetical bias of DCE, the actual selection probabilities of certain snack attributes may be overestimated [50]. For example, the WTP values estimated in this study should be interpreted with caution as quantitative indicators reflecting consumers’ relative valuation of the attributes, rather than precise, market-applicable absolute price predictions; The choice sets in this study were designed to maximally reflect currently available market products rather than pre-defining an “ideal” snack based on nutritional guidelines. While this approach limits the normative guidance of the findings, it enables a better understanding of the trade-offs made by school-aged children in real-world market settings.

Future research

Future studies could expand nationwide to establish a nationally representative sample and explore the impact of various emerging attributes on snack choices among school-aged children; Simultaneously, Subsequent studies could combine multiple methods such as eye-tracking experiments [51] and neuroimaging techniques [52] to elucidate the neural mechanisms underlying school-aged children’s dietary decision-making, providing biological evidence to support intervention frameworks; Besides, Future research also could build a normative benchmark (i.e., an “ideal” snack) to quantify the gap between subjects’ preferences and public health objectives.

Abbreviations

DCE

Discrete Choice Experiment

WTP

Willingness-to-Pay

RI

Relative Importance

β

Beta coefficient

P

P-Value

SD

Standard Deviation

BMI

Body Mass Index

Ref

Reference

Authors’ contributions

HL drafted and revised the manuscript. WZ, XW and XZ collected and analyzed the data. FG, XJ, QG, JZ, and SL provided guidance and technical support. WD conceived the study, obtained funding, and critically reviewed the manuscript. HW and AL provided support and guidance. All authors read and approved the final manuscript.

Funding

China Student Nutrition and Health Promotion Association—Mead Johnson Academic Excellence Nutrition Research Fund Project (CASNHP-MJN2023-18).

Data availability

The datasets generated and/or analysed during the current study are not publicly available due [REASON WHY DATA ARE NOT PUBLIC] but are available from the corresponding author on reasonable request.

Declarations

Ethics approval and consent to participate

Informed consent was obtained from all subjects and their parents or legal guardians involved in the study.The study protocol was approved by the Ethics Committee of the National Institute for Nutrition and Health, Chinese Center for Disease Control and Prevention (Ethics Approval No. 2024–007). This research was performed in compliance with the World Medical Association’s Declaration of Helsinki, which is a statement of ethical principles for medical research involving human participants.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.