Current research on rural residents’ nutrition primarily focuses on measurement methods and influencing factors.

First, regarding nutritional status measurement, diverse methodologies exist. Among these, the Dietary Diversity Score (DDS) is widely adopted, as it better reflects changes in nutritional composition. Common approaches include nutrient intake levels, the Food Consumption Score (FCS) used by the United Nations World Food Program, and the Dietary Diversity Score (DDS) employed by the Food and Agriculture Organization (FAO). The core of the nutrient intake level method lies in measuring caloric intake, protein, and fat consumption. However, these indicators only reflect changes in nutrient quantities and cannot capture shifts in nutritional composition. The DDS and FCS share similar measurement mechanisms but differ slightly in food grouping, the food consumption recall period, and food group weighting. The choice of method depends on data collection timing, resources, and evaluator requirements.

Second, regarding the factors influencing rural residents’ nutrition, these encompass both macro and micro factors. Among these, income is the primary factor, while other factors primarily exert their influence through income effects. At the macro level, rural nutrition is primarily affected by urbanization, market access, and agricultural production diversity. Urbanization continuously optimizes rural dietary consumption patterns by increasing non-agricultural income, reducing carbohydrate and fat intake, and boosting protein consumption (Ren et al., 2021). Market access lowers transaction costs, increases farmers’ sales revenue, and reduces expenses for agricultural inputs and food purchases, thereby enhancing household welfare and nutrition (Usman and Haile, 2022). Specialization reduces agricultural production diversity. In highly commercialized rural areas, production specialization can improve dietary diversity by increasing agricultural income, diminishing the importance of production diversification for dietary diversity. However, in less commercialized rural areas, production diversification remains crucial for dietary diversity (Ickowitz et al., 2019). At the micro level, the nutrition of rural residents is influenced by factors such as income level, education attainment, and nutritional knowledge. Zheng et al. (2023) found that nutrition education programs contribute to increased dietary diversity and total calorie intake among rural residents.

Against the backdrop of digital transformation, the internet—as a carrier and channel for information and digital technologies—has permeated rural production and living spaces with more advanced production methods, lifestyles, and concepts, profoundly impacting the “rural development-agricultural efficiency-farmer income” dynamic. Extensive research exists on the effects of internet use on the development of agriculture, rural areas, and farmers.

Some studies have examined the impact of internet development on residents’ nutritional levels and quality, with differing perspectives emerging. On one hand, there are findings regarding the positive effects of internet usage on nutritional quality. Parlasca et al. (2020) examined the impact of mobile phone use on nutritional status in pastoralist communities in northern Kenya, finding that mobile phone use significantly enhanced dietary diversity. Twumasi et al. (2021) investigated the effects of internet use on food and nutrition security among rural households in Ghana, concluding that internet access improves household food availability (HFIAS) and dietary diversity, with more pronounced effects observed in households engaged in non-agricultural work and large-scale land management. Ma and Jin (2022) examined the impact of internet use on dietary quality among rural residents using CHNS (2006–2011) data. They found internet use increased China Food Pyramid scores by 10.4%. Leng et al. (2024) investigated the relationship between digital technologies, labor allocation, and nutritional intake using CFPS data. They discovered digital skills can boost rural residents’ protein consumption by facilitating non-agricultural employment. On the other hand, regarding the negative impact of internet usage on nutritional quality, Zou and Lin (2024) examined the effect of takeout food consumption on overweight and obesity risks among urban residents. Their study found a direct correlation between frequent consumption of O2O takeout and increased risks of overweight and obesity. Liu et al. (2025) examined the impact of internet usage on adolescent nutrition. Their findings revealed that internet use increases protein intake while simultaneously leading to higher consumption of fast food and beverages, thereby elevating the risk of adolescent obesity.

This review identifies several areas warranting further exploration. Regarding research content, the mechanisms through which internet use affects rural residents’ nutrition remain understudied. Domestic literature has not examined this topic, while only a few international studies touch upon it without fully investigating the underlying mechanisms linking internet use to rural nutrition. Regarding research indicators, existing studies on dietary diversity among rural residents primarily employ household-level dietary diversity scores (HDDS) as a proxy for nutrition. However, HDDS is not a highly precise indicator of dietary quality, as it overlooks intra-household food distribution issues. Individual dietary diversity scores (DDS) offer a more accurate measure of dietary quality (Parlasca et al., 2020). Furthermore, existing studies have not sufficiently discussed the heterogeneous effects of internet use on dietary diversity among rural residents. Based on this, this paper further expands the research on the aforementioned issues.

Improvements in the nutritional status of rural residents primarily face obstacles such as income constraints, limited food accessibility, and insufficient nutritional literacy. Existing research has demonstrated that internet usage can enhance dietary diversity among rural residents by increasing market accessibility (Xue et al., 2021) and other means. Building upon prior studies, this research posits that the internet may also influence rural residents’ nutrition through the following two pathways. Improving the nutritional status of rural residents primarily faces obstacles such as income constraints, limited food accessibility, and insufficient nutrition literacy. Existing research has demonstrated that internet usage can enhance dietary diversity among rural residents by increasing market accessibility and other means. Building upon prior studies, this research posits that the internet may also influence rural nutrition through the following two pathways.

In terms of direct effects, the internet can improve rural residents’ dietary diversity by enhancing nutrition literacy. Nutrition literacy, a specialized form of health literacy, encompasses knowledge and concepts related to food nutrition, along with the skills required to obtain, select, and prepare food. It serves as a bridge connecting individuals, food, and their environment, playing a crucial role in dietary behavior change and improving dietary quality. Simultaneously, nutrition literacy primarily functions through nutrition education, profoundly influencing residents’ ability to make scientifically informed nutritional decisions in both cognitive and skill-based dimensions. Traditionally, China’s rural society has been structured around kinship and geographical ties, forming a social network system. Villagers’ activities are constrained by three geographical boundaries: their living sphere, production sphere, and social sphere. Traditional customs and agrarian culture heavily influence dietary habits and patterns, often leading to a lack of rational understanding of dietary quality. Examples include eating “leftovers” due to thriftiness and preferences for staple foods with heavy oil and salt consumption stemming from high-intensity physical labor. Due to traditional beliefs and limited access to nutritional knowledge, nutritional literacy remains significantly underdeveloped. According to statistics from the National Health Commission, the health literacy rate among urban residents nationwide reached 33.25% in 2023, while that of rural residents stood at 26.23%, indicating a persistent gap in health literacy between urban and rural populations.

The widespread adoption of the internet has profoundly influenced and transformed rural residents’ nutritional literacy and behavioral patterns. Generally speaking, rural residents acquire nutritional knowledge through two primary channels. First, they rely on mutual learning and imitation of nutritional knowledge among community members. However, within this model, the relatively closed nature of rural communities leads to limited and homogeneous sources of nutritional information. This approach has limited effectiveness in improving dietary quality, and the sharing of experiences within communities may even perpetuate numerous misconceptions about nutrition. Second, they acquire nutrition knowledge through traditional media such as television, radio, and newspapers. However, in these channels, rural residents are passive recipients of nutrition information. The nutrition knowledge disseminated by these official media often diverges significantly from the realities of rural life, making it difficult to generate an effective “demonstration effect.” In contrast, internet communities leverage new media platforms like social networks for knowledge dissemination. Knowledge sharing is primarily driven by individuals who provide abundant scientific nutrition information and health advice closely aligned with rural residents’ daily lives. This approach carries greater emotional resonance, readily forming a “demonstration effect” on rural residents’ nutritional literacy and profoundly influencing their dietary perceptions and behaviors. Furthermore, rural residents can utilize search engines and other platforms to access cooking techniques and nutritional knowledge anytime, anywhere, according to their specific needs. This significantly enhances their nutritional literacy and improves dietary diversity. Based on the above discussion, the following hypothesis is proposed.

In terms of indirect impact, internet use can improve nutritional levels by increasing rural household income. Building upon Barnum and Squire (1979) and Argaw et al. (2021), this paper analyzes the impact of internet use on dietary diversity among farming households within a semi-commercialized family farm behavior model framework featuring competitive labor markets. Semi-commercial farms represent a transitional model between fully commercial farms—which hire labor and sell all agricultural products—and purely subsistence farms—which rely solely on household labor and sell only a portion of their output. All households participate in the labor market as either buyers or sellers of labor. The fundamental assumptions of the Chinese agricultural model analyzed here align closely with this framework. Within this model, labor hours and output are determined by market wages and prices, both treated as exogenous variables.

Farmers aim to maximize utility under constraints, specifically production function constraints, time constraints, and budget constraints. To link this to dietary diversity, we assume separate utility functions for agricultural and non-agricultural goods, expressed as:

In Equation 1, U is the household utility function, DD ( q 1 , … , q n ) is defined as the sub-utility function for food consumption and also represents the consumption of non-agricultural goods (dietary diversity), q 0 denotes the consumption of agricultural products, and Z is the fixed household characteristic.

In Equation 2, F is the production constraint function, L is labor input, V i is other variable inputs, and X is the household structure characteristic.

In Equation 3, Y represents total income, T denotes total available time, p L , p , v i , and p L are the prices of various factors, p 0 , p D is the product price, and E is exogenous income.

Therefore, the relationship between farmers’ demand for dietary diversity and various influencing factors can be further expressed as:

In Equation 4, dietary diversity is primarily influenced by product prices, household fixed characteristics, and income. Since product prices are exogenous and household fixed characteristics are constant, dietary diversity is mainly affected by income. Furthermore, the impact of the internet as an exogenous variable on dietary diversity is primarily transmitted through income. In traditional rural societies, farmers often face high and incomplete costs in accessing market information. The use of the internet plays a crucial role in reducing information asymmetry and transaction costs for farmers in areas such as agricultural input procurement, agricultural production, and agricultural product sales.

First, in the procurement of agricultural inputs, farmers traditionally rely on input dealers to purchase production factors, lacking diversity in procurement choices. The internet not only provides farmers with more procurement channels to obtain agricultural inputs at lower prices but also enables them to bypass distributors and purchase directly from manufacturers, securing even lower costs. Second, in agricultural production, the internet enhances total agricultural output by incorporating total factor productivity into production functions. It enables farmers to conveniently access critical production information such as pest forecasts and weather conditions, allowing more precise guidance on planting, fertilization, irrigation, and other processes. This reduces delays, losses due to missed planting seasons, and damages from weather disasters. Third, in the agricultural product sales phase, smallholder farmers in many developing countries typically sell their produce to middlemen between villages and markets. Farmers lack autonomous pricing power and can only passively accept prices. The internet not only helps narrow the gap in access to market price information for farmers, enabling them to understand price differences across markets and thereby improve bargaining power to raise agricultural product prices, but also helps farmers sell directly to consumers online, reducing distribution channels and granting them autonomous pricing power to increase sales prices. Based on the above discussion, the following hypotheses are proposed.

The data used in this study are selected from the China Health and Nutrition Survey (CHNS). This survey covers 15 provinces and municipalities: Beijing, Liaoning, Heilongjiang, Shanghai, Yunnan, Jiangsu, Henan, Hubei, Zhejiang, Hunan, Guangxi, Guizhou, Chongqing, and Shaanxi. To date, the CHNS official website has released data from 11 rounds of surveys conducted between 1989 and 2015. The survey employs a multistage, stratified, random sampling method, collecting information on socioeconomic status, individual dietary patterns and nutritional status, and household income. CHNS began collecting individual internet use data in 2004, and its food consumption and nutritional data are only updated through 2011. Therefore, this study selected data from four rounds (2004, 2006, 2009, 2011) as the sample dataset. First, nutritional data were matched with individual and household data. After retaining rural resident samples and cleaning missing values, an unbalanced panel dataset with a sample size of 27,012 was obtained. According to a report by the China Internet Network Information Center (CNNIC), internet penetration in rural areas rose from 0.076% in 2002 to 11.7% in 2008, reaching 20.7% by 2011. The rural internet usage sample in this study accounted for 15.1% of the total sample, which generally aligns with the rural internet penetration rates published by CNNIC. Considering the trend in internet penetration, the sample data still holds certain significance for current policy formulation.

The benchmark regression model is specified in Equation 5:

DD S it denotes the dietary diversity score of rural resident i during the period t . Interne t it represents the internet use status of rural resident i during period t , with a value of 1 for internet use and 0 otherwise. X it denotes the control variables across three levels: individual characteristics, household characteristics, and community characteristics. σ t and ω i denote the time fixed effect and individual fixed effect, respectively. ε it represents the random error term.

The dependent variable in this study is dietary diversity. Food types in the CHNS dietary data are classified and coded according to the Chinese Food Composition Table, divided into 21 major categories such as cereals and products, tubers, starches and products, and dried beans and products. A 6-digit coding system is used: the first three digits represent the food category code, the third digit indicates the food subcategory code, and the last three digits denote the sequential number within the subcategory. To calculate dietary diversity scores, this study combined the food classification standards of the Food and Agriculture Organization of the United Nations (FAO) and China’s balanced diet pyramid, grouping the 21 major food categories into 9 categories: cereals, vegetables, fruits, meat, eggs, fish, dairy, nuts, and oils. The dietary diversity score is calculated based on the total number of distinct food categories consumed by each resident over a 3-day period. Each food category consumed is assigned a weighting of 1 point, with a maximum score of 9 points. Consumption frequency and quantity are not counted. Foods outside the above 9 categories—including infant formula, snacks, cookies, instant foods, beverages, sugar, dried fruits, candied fruits, and honey—are excluded from scoring.

The core independent variable in this study is whether or not individuals use the internet. Use of the internet is assigned a value of 1, and non-use is assigned a value of 0. Figure 1 presents the cumulative distribution of dietary diversity scores among rural residents in the internet-using and non-internet-using groups. It shows that the dietary diversity scores of the internet-using group are significantly higher than those of the non-internet-using group. Figure 2 illustrates the annual changes in dietary diversity scores among rural residents. Overall, dietary diversity scores exhibit a year-on-year increasing trend, with the internet-using group consistently maintaining higher scores than the non-internet-using group.

The control variables are based on the research of Huang and Tian (2019) and Deng et al. (2024), and include data from three levels: individual, family, and community. The individual-level variables are age and level of education. The level of education is calculated based on years of schooling. Those who did not attend school are assigned a value of 0, while those who graduated from elementary school are assigned a value of 6. Graduates of middle school are assigned a value of 9, graduates of high school (including general high school, technical school, and vocational school) are assigned a value of 12, graduates of college are assigned a value of 16, and graduates of graduate school are assigned a value of 19. The family-level variables are household size, household net income, whether or not the household owns a three-wheeled vehicle, a motorcycle, or a car, and whether or not the household owns a refrigerator. The community-level variables are urbanization rate and community population density.

Table 1 presents descriptive statistical analysis results. Regarding the dependent variable, the dietary diversity of the internet-using sample was approximately 1.286 higher than that of the non-internet-using sample. This mean difference passed the 1% significance test, preliminarily indicating that internet use positively promotes dietary diversity among rural residents. Regarding control variables, the mean differences between internet-using and non-internet-using samples all passed the 1% significance test, indicating the necessity of further controlling these variables in regression analysis.

Table 2 presents the results of the impact of internet use on dietary diversity among rural residents. Column (1) shows the univariate regression results. Controlling for individual fixed effects, the coefficient for internet use is significantly positive at the 1% level, indicating that internet use significantly promotes dietary diversity among rural residents. Column (2) adds control variables to the univariate regression, and the coefficient for internet use remains significantly positive at the 1% level. Column (3) controls for time fixed effects and other variables on top of the univariate regression. The coefficient for internet use remains significantly positive at the 1% level. The final column simultaneously controls for individual and time fixed effects, yielding robust regression results. These findings align with the conclusions of the studies by Deng et al. (2024) and Yang et al. (2023).

Among the control variables, when controlling for both individual and time fixed effects, the coefficients for household size and motorcycles are significantly positive at the 5% level. Specifically, larger households may exhibit greater dietary diversity due to differing preferences among members. Motorcycles, as transportation tools, can shorten the distance between rural households and nearby markets or farmers’ markets, thereby indirectly enhancing dietary diversity. Compared to other modes like tricycles or cars, motorcycles offer lower travel costs. The coefficients for refrigerators and household income were significantly positive at the 1% level, indicating that these variables exert a significant positive influence on rural residents’ dietary diversity. Income stands as one of the primary factors enabling dietary diversity among rural residents, primarily by alleviating budget constraints. Karlsson and Subramanian (2023) note that refrigerator use mitigates nutrient loss in food, reduces exposure to harmful pathogens from improper storage, and enables purchasing larger quantities of food to improve nutrition while lowering costs.

Although this paper controls for individual, household, and community-level covariates in the benchmark regression model and incorporates fixed effects for both individuals and time to address unobservable factors, endogeneity issues may still arise. First, unobserved variables are unavoidable. Second, rural residents’ internet use may exhibit sample self-selection due to differences in individual, household, and community characteristics. To address these issues, this study primarily employs the instrumental variables method and the Heckman two-step approach.

The socio-ecological framework emphasizes the complexity of food behavior and the interaction of individual and environmental factors in various circumstances (Story et al., 2008). Individual factors include biological and demographic characteristics, as well as behavioral abilities. Food affordability, availability, accessibility, and quality are all aspects in the food environment (Lytle and Sokol, 2017). Food affordability considers household income and the ability to purchase food; food availability considers the availability of food channels in the community; food accessibility also considers geographical distance and the cost of obtaining food; and food quality primarily considers the richness and variety of food types. According to the preceding discussion, the impact of internet use on dietary diversity among rural residents may range depending on individual, family, and community variables, as well as food type. As a result, this section focuses on heterogeneity from the perspectives outlined below.

The theoretical analysis section indicates that internet usage influences dietary diversity among rural residents through both direct and indirect pathways. First, via the direct pathway, the internet can improve the nutritional environment by enhancing nutritional literacy. Nutrition literacy is measured using a proxy variable—whether individuals are aware of the Dietary Guidelines for Chinese Residents. Indirectly, internet use enhances nutritional status by increasing rural household income. The impact of household income on dietary diversity has already been validated in the baseline regression section and is not revalidated here. Table 9 presents the results of the mechanism analysis. Columns (1) and (2) validate the direct pathway mechanism. In Column (1), the coefficient for internet usage is significantly positive at the 1% level, indicating that internet usage has a significant positive impact on rural residents’ dietary knowledge. In Column (2), even after controlling for dietary knowledge, internet usage still exerts a significant positive effect on dietary diversity. These findings suggest that internet usage improves dietary diversity among rural residents by enhancing dietary knowledge and household income.