Internet usage elevates elderly obesity: evidence from a difference-in-differences analysis of the broadband China policy

Abstract

Background

The global aging population is rapidly increasing, which has led to a growing prevalence of obesity among the elderly. Body mass index (BMI) is a crucial measure of obesity and is linked to an increased risk of various chronic diseases. At the same time, the widespread use of the internet and digital technologies has significantly influenced the health behaviors and outcomes of the elderly.

Objective

This study aims to examine the causal relationship between internet usage and BMI among the elderly, addressing a gap in existing research and providing evidence for the development of health policies targeted at the elderly population.

Methods

Utilizing China’s “Broadband China” strategy as a quasi-natural experiment, we employed a difference-in-differences (DID) approach to analyze panel data from the CHARLS covering the years 2011–2015. By comparing the treatment and control groups before and after the policy’s implementation, we identify causal effects.

Results

Our findings indicate that the “Broadband China” strategy significantly increased BMI among the elderly. The mechanisms underlying this effect include reduced sleep duration, decreased physical activity levels, and worsened mental health. Furthermore, the impact of internet usage on obesity is particularly pronounced among urban residents, those without chronic diseases, and individuals with fewer surviving children.

Conclusions

Policy recommendations include promoting healthy internet usage practices, enhancing community-based activity facilities, and providing comprehensive mental health support to mitigate obesity rates and improve health outcomes among the elderly.

Text box 1. Contributions to the literature
• This study uncovers the causal link between internet use and elderly BMI, solving prior endogeneity issues.
• It applies a robust quasi-natural experiment, surpassing cross-sectional methods.
• The research fills a gap in developing country data, informing China’s health policy.

Introduction

With the accelerating aging of the global population, health issues, particularly obesity among the elderly, are receiving increasing attention [1]. Obesity not only increases the risk of chronic diseases but also may lead to reduced mobility, diminished quality of life, and an increased risk of premature death [2]. Body mass index (BMI), a key indicator for measuring individual obesity levels, is directly linked to the risk of chronic diseases, including cardiovascular disease and diabetes [3]. Therefore, understanding and managing elderly BMI is essential for enhancing their overall health status [4].

In recent years, the widespread adoption of the internet and digital technologies has not only transformed people’s lifestyles but has also profoundly influenced health behaviors and outcomes, especially among the elderly [5]. Although existing studies indicate a significant association between internet usage and BMI, particularly among younger and middle-aged populations [6, 7], research examining the impact of internet usage on elderly BMI remains scarce. Given the substantial impact of elderly health on medical resources and social welfare, investigating the influence of internet usage on their BMI is essential. To fill this research gap, this study aims to investigate the relationship between internet usage and BMI among the elderly.

The primary challenge in this study is effectively identifying the causal relationship between internet usage and BMI among the elderly. A typical issue of reverse causality exists: it is unclear whether internet usage influences BMI or vice versa. Most previous studies have conducted merely correlation analyses, which fail to effectively identify the causal relationship between internet usage and BMI, potentially leading to biased estimates and unreliable conclusions. Fortunately, the gradual implementation of the “Broadband China” strategy (BCs) since 2013 presents a unique opportunity to address this challenge. The BCs is designed to accelerate the construction of information infrastructure, increase broadband network coverage, and improve access speeds, thus fostering the widespread adoption of information technology across various socioeconomic sectors. The implementation of BCs significantly increases the likelihood of internet usage among the elderly in China and can be considered an ideal exogenous shock to internet usage. By employing the difference-in-differences (DID) method, we can effectively identify the causal relationship between internet usage and BMI among the elderly.

Specifically, this study uses BCs implementation as a quasi-natural experiment, which creates a control group and a treatment group through a phased rollout. Using the DID method, the study compares the differences between the control and treatment groups before and after the policy implementation. This approach controls for time-invariant individual characteristics and common trends, thereby enhancing the reliability of causal inferences. It also helps to address endogeneity issues and ensures the external validity of the policy effects, providing broad policy implications. Finally, through empirical analysis of three-wave longitudinal panel data from the China Health and Retirement Longitudinal Study (CHARLS) spanning 2011–2015, this study overcomes the limitations of cross-sectional data in establishing causality. Through this design, the study aims to uncover both the causal relationships and potential mediating mechanisms by which internet usage affects the health of middle-aged and elderly individuals, thereby providing empirical evidence from a developing country and offering a scientific basis for policies that promote healthy internet usage among the elderly.

The selection of middle-aged and elderly individuals in China as the focus of this research is highly important and offers several advantages. First, China, once the most populous country in the world, now faces a rapidly aging population. According to the latest data from the National Bureau of Statistics of China, the population aged 60 and above has reached nearly 300 million, representing 21.1% of the total population [8]. This substantial demographic group is of considerable practical importance for health research. Second, China has experienced rapid internet adoption in recent decades. In particular, driven by the “Broadband China” policy, internet usage among middle-aged and elderly individuals has significantly increased. The CNNIC reported that by December 2022, the proportion of internet users aged 60 and above had risen to 14.3%, with the number of elderly internet users reaching 153 million, indicating that at least one in two elderly individuals now has access to the internet [9], thus offering a unique opportunity to study its health impacts. Furthermore, compared with developed countries, China’s middle-aged and elderly population has unique characteristics in terms of socioeconomic status, healthcare coverage, and cultural background, making the study of this group valuable for generating comparative data and providing new insights into global elderly health research.

Compared with existing studies, this research offers the following contributions. First, it makes both theoretical and empirical contributions. This study goes beyond existing correlation analyses by uncovering the causal relationship between internet usage and BMI among the elderly, thereby deepening our understanding. Additionally, it explores the specific mechanisms through which internet usage affects BMI, enriching the theoretical framework surrounding the relationship between internet usage and health outcomes. Second, in terms of empirical contributions, by utilizing BCs as a quasi-natural experiment for the first time, this study employs the DID method to provide robust causal inferences. This rigorous research design serves as a methodological reference for similar studies, enhancing the credibility and reliability of causal inferences in this field. Third, the findings provide the first empirical evidence from a developing country to inform policies that promote healthy internet usage among the elderly, offering important insights for addressing the global challenges of aging.

Literature review

In recent years, the widespread adoption of the internet and digital technologies has attracted significant global attention because of their health impacts. This section reviews relevant literature to explore the complex relationship between internet usage and BMI, focusing on findings, methodologies, and limitations of various studies, and proposing directions for future research.

Direct association between internet use and BMI

Research has consistently identified a significant association between internet usage and BMI. For example, Vandelanotte et al. (2007) [6] conducted multiple logistic regression analyses on Australian adults and discovered that increased computer and internet usage during leisure time was linked to overweight and obesity, without corresponding changes in physical activity levels. Similarly, Canan et al. (2014) [7] reported a significant association between internet addiction and increased BMI among Turkish adolescents. Further research by Shen et al. (2021) [10] highlighted that the relationship between digital technology use and BMI is partially mediated by sleep deprivation in adolescents. Yen et al. (2010) [11] reported that in Taiwanese adolescents, high levels of television watching and internet usage were associated with increased BMI, although physical exercise could mitigate these effects. Collectively, these studies indicate that sedentary digital media use significantly impacts weight gain, although specific mediating mechanisms and causal relationships still require further investigation. Park and Lee (2017) [12] analyzed data from Korean middle and high school students and reported that problematic internet usage was associated with inappropriate weight control behaviors. Similarly, Sari and Aydin (2014) [13] reported a significant relationship between BMI and problematic internet usage among Turkish university students. These findings suggest a universal impact of internet usage on weight management across various age groups and cultural backgrounds.

Longitudinal study and cross-cultural comparison

Longitudinal studies provide clearer causal evidence regarding the relationship between internet usage and BMI. For example, Barrense-Dias et al. (2016) [14] conducted a longitudinal study on Swiss adolescents and reported that spending more than two hours on the internet during weekends significantly increased the risk of being overweight. Similarly, Tsitsika et al. (2016) [15] reported a significant association between problematic internet usage and the risk of obesity among European adolescents, particularly boys. These studies, which track data over extended periods, provide more robust evidence supporting the relationship between internet usage and BMI. Moreover, Qiu et al. (2021) [16] examined adults in China and reported a negative correlation between computer usage time and BMI, whereas mobile phone usage time was not significantly related to BMI, indicating that different types of internet usage may have varying effects on body weight. Subu et al. (2021) [17] examined Indonesian middle school students and reported that high levels of internet gaming addiction were associated with a tendency toward being overweight, although the relationship was relatively weak. These cross-cultural comparative studies reveal both the generalizability and the variability of the impact of internet usage on BMI. In the context of China, Tang et al. (2023) [18] utilized data from the 2020 China Family Panel Studies (CFPS) and reported that internet usage among adult netizens was associated with BMI. Specifically, behaviors such as playing online games and watching short videos were positively correlated with BMI growth, whereas engaging in online learning was negatively correlated with BMI growth. This finding suggests that different types of internet usage behaviors may exert distinct effects on weight management, warranting further investigation.

Mediating effect of internet usage on mental health

Research suggests that the relationship between internet usage and BMI might be mediated by psychological factors. Faith et al. (2016) [19] reported that obese individuals are more likely to use the internet to seek health information, potentially because of their heightened concern about their health. Gentile et al. (2021) [20] emphasized that BMI influences internet addiction through social stigma and self-esteem, suggesting that a complex psychological mechanism requires further exploration. Novaković et al. (2023) [21] reported that among children, internet usage time is associated with obesity, psychological distress, and poor social adjustment. These studies indicate that the impact of internet usage on BMI encompasses not only physiological factors but also significant psychological and social dimensions.

In summary, existing research consistently demonstrates a significant association between internet usage and BMI, with this relationship possibly influenced by various mediating mechanisms. However, the majority of studies are cross-sectional, limiting the ability to establish causality. Although some studies utilize longitudinal panel data, issues of endogeneity remain inadequately addressed. Additionally, theoretical frameworks for analyzing the relationship between internet usage and BMI remain underdeveloped. This study aims to construct a more comprehensive theoretical framework and provide rigorous causal identification to advance our understanding of how internet usage influences BMI.

Mechanism hypotheses

Reduced sleep duration

Research consistently shows a significant association between digital technology use and sleep deprivation. Excessive internet usage can adversely affect sleep quality among the elderly by reducing total sleep time and increasing sleep latency [10]. Additionally, reliance on social networking sites has been strongly linked to poorer sleep quality, a finding that also applies to the elderly [22]. Sudharkodhy et al. (2023) [23] further demonstrated that internet addiction is strongly and negatively correlated with sleep quality, particularly because nighttime internet usage leads to difficulties falling asleep, fragmented sleep, and a reduction in overall sleep duration among the elderly. Overall, excessive internet usage—especially increased screen time at night—may substantially impair sleep quality in the elderly.

Sleep deprivation has numerous adverse effects on elderly health, particularly on BMI. Research indicates that inadequate sleep can increase appetite in the elderly, leading to increased consumption of high-calorie foods and consequent weight gain [24]. Additionally, studies have shown that sleep deprivation is associated with reduced physical activity, potentially leading to increased weight and BMI [25]. Moreover, sleep deprivation can decrease the metabolic rate and insulin sensitivity—physiological changes that may also lead to weight gain [26]. In summary, reduced sleep duration in the elderly not only directly impacts daily energy expenditure but also significantly increases BMI by affecting appetite and metabolic function.

Hypothesis 1

Internet usage increases BMI among the elderly by reducing sleep duration.

Decreased physical activity levels

Extended internet usage often encroaches on time that could be devoted to physical activities, leading to a reduction in overall activity levels. Vandelanotte et al. (2007) [6] reported that individuals who spent more than three hours per week on computers and the internet had a significantly greater risk of overweight or obesity. This finding indicates a strong association between increased internet usage and reduced physical activity. Additionally, internet usage not only affects the time allocated for physical activity but also influences the intensity and frequency of such activities. Moreno et al. (2013) [27] further demonstrated that internet usage, particularly for social networking, is significantly associated with fewer days of vigorous physical activity. These results suggest that increased internet usage not only reduces the total time spent on physical activities but also decreases the frequency of high-intensity activities, thereby further diminishing energy expenditure.

The negative health impacts of reduced physical activity are particularly pronounced, especially in relation to BMI. A decrease in physical activity results in insufficient energy expenditure and increased fat accumulation. Vandelanotte et al. (2009) [28] reviewed the relationship between sedentary behavior and health outcomes and reported close associations between sedentary behavior and increased obesity and BMI. Reduced physical activity lowers the metabolic rate and increases fat accumulation, leading to elevated BMI. Riiser et al. (2014) [29] conducted a longitudinal study on adolescents and reported a significant correlation between a lack of physical activity and increased BMI, with the risk of BMI increase being greater with less activity. These studies suggest that reduced physical activity not only lowers energy expenditure but also alters the metabolic rate, contributing to weight gain and increased BMI.

Hypothesis 2

Internet usage increases BMI among the elderly by lowering their level of physical activity.

Worsened mental health

Research indicates that internet usage can adversely impact the psychological health of the elderly. While the internet can enhance social connections, it also introduces psychological stress and anxiety [30]. For example, excessive internet usage is strongly associated with symptoms of depression among the elderly, particularly women, those with higher incomes, and those with higher levels of education [31]. Choi and DiNitto (2013) [32] also reported that a higher frequency of internet usage among the elderly was linked to more pronounced symptoms of depression and anxiety. Overall, despite providing more social opportunities, excessive internet usage may lead to a deterioration in mental health among the elderly.

Deterioration in mental health significantly affects physical health, particularly BMI. Research shows that psychological issues, such as depression and anxiety, are associated with weight gain and increased BMI [33]. For example, psychological stress and depressive symptoms can increase appetite, leading to increased consumption of high-calorie foods and subsequent weight gain and increased BMI [34]. Additionally, mental health problems can lead to decreased physical activity, further contributing to weight gain and increased BMI [35]. In summary, deterioration in mental health significantly increases BMI through various mechanisms. On this basis, the following hypothesis is proposed:

Hypothesis 3

Internet usage increases BMI among the elderly by worsening their mental health.

To provide a more intuitive presentation of the three theoretical mechanisms and the core theoretical framework of this study, we have created the following causal diagram (Fig. 1):

Firstly, the Broadband China strategy (BCs) is used as the independent variable throughout the paper, representing internet usage through the exogenous policy shock of the Broadband China pilot program. BMI serves as the outcome variable, indicating the level of obesity among older adults. The mediator represents the intermediate variables, including sleep duration, physical activity, and mental health. The covariate includes a series of control variables, such as alcohol consumption, socioeconomic status, years of education, residential location, gender, marital status, and age. The error term captures all unobservable factors influencing BMI.

Secondly, BCs affects BMI through two pathways. One is direct, represented by BCs → BMI, and the other is indirect, represented by BCs → Mediator → BMI. Covariate includes a series of control variables that directly influence BMI but are assumed to be independent of the BCs pilot policy, consistent with the exogeneity of BCs. Including these covariates in the regression model helps control for their direct effects on BMI, ensuring that the estimated impact of BCs on BMI is unbiased. Error represents all unobservable factors outside the econometric model that influence BMI, adhering to the classical assumption that error terms are uncorrelated with explanatory variables. Thus, the error term is assumed to affect only BMI and not BCs.

Fig. 1

Causal diagram illustrating the theoretical framework of the study on the Broadband China strategy and obesity among older adults in China (2011–2015)

Empirical design

Data sources

This study uses data from the China Health and Retirement Longitudinal Survey (CHARLS). CHARLS is a large, interdisciplinary survey project managed by the National School of Development at Peking University and jointly executed by the China Social Survey Center and the Youth League Committee at Peking University. The survey aims to collect detailed information on households and individuals aged 45 and older in China. It employs a probability proportional to size (PPS) sampling method at multiple levels—including county, village, household, and individual. The baseline survey was conducted in 2011 and included 450 villages in 150 counties (districts) nationwide. A total of 10,257 households and 17,708 individuals were surveyed, providing a comprehensive representation of the elderly population in China. The CHARLS questionnaire encompasses personal demographic information, family structure, work and retirement details, health status, and health behaviors. Since 2011, CHARLS has conducted nationwide surveys every two years, resulting in data collection spanning a decade, including the years 2011, 2013, 2015, 2018, and 2020. Currently, CHARLS is the most authoritative and largest-scale representative microlevel database available for research on elderly health in China. Since BMI measurement data are available only from the 2011, 2013, and 2015 surveys, this study uses panel data from these three waves (2011–2015) for empirical analysis.

The data cleaning process was conducted as follows: (1) Merging longitudinal data: The survey data from the three waves (2011, 2013, and 2015) were merged to form a longitudinal dataset with a total of 57,405 observations. (2) Removal of observations with missing dependent variable values: observations with missing values for the dependent variable (BMI) were excluded, reducing the dataset to 39,464 observations. (3) Exclusion of samples under age 55: Samples younger than 55 years [36] were further excluded, reducing the dataset to 26,608 observations. (4) Removal of observations with missing key explanatory and control variables: Observations with missing values for key explanatory and control variables were removed, leaving a final dataset of 18,482 observations.

Econometric specification

To accurately assess the impact of internet usage on BMI among the elderly, a difference-in-differences (DID) estimation model utilizing a two-way fixed effects approach was constructed:

$$\:BM{I_{it}} = {\beta _0} + {\beta _1}BC{s_{ct}} + X'\theta + {\delta _i} + {\gamma _t} + {\varepsilon _{it}}$$

(1)

where $\:{\text{BMI}}_{\text{it}}$ represents the body mass index of individual $\:i$ at time $\:t$, which serves as an indicator of obesity among the elderly. $\:{\text{BCs}}_{\text{c}\text{t}}$ is a binary variable representing the Broadband China strategy, capturing the exogenous policy shock of increased internet usage, where $\:c$ denotes the city. $\:{\text{X}}^{{\prime }}\theta$ is a vector of control variables that account for other factors influencing BMI, with $\:\theta\:$ being the vector of coefficients for these control variables. $\:{\text{δ}}_{\text{i}}$ represents individual fixed effects, controlling for differences between individuals. $\:{\text{γ}}_{\text{t}}$ denotes time fixed effects, controlling for unobserved shocks affecting all individuals in a given year. $\:{\text{ε}}_{\text{it}}$ represents the error term. The coefficient of primary interest, $\:{\text{β}}_{\text{1}}$, measures the causal effect of increased internet usage (attributed to the BCs) on BMI.

Definition of variables

Dependent variable (BMI)

In the context of obesity measurement, BMI is widely recognized in the international literature as the standard metric. The formula used is “$\:\text{B}\text{M}\text{I}={\text{w}\text{e}\text{i}\text{g}\text{h}\text{t}\:\left(\text{k}\text{g}\right)/\text{h}\text{e}\text{i}\text{g}\text{h}\text{t}\:\left(\text{m}\right)}^{2}$”. This index serves as the primary measure of obesity in the elderly throughout this study. Nonetheless, alternative obesity measurement indicators proposed by some scholars were considered in the sensitivity analysis of this study.

Key explanatory variable (BCs)

To address the endogeneity of internet usage, this study employs BCs as an exogenous policy shock, utilizing it as a proxy for internet usage. The BCs initiative was officially launched in 2013, with the number of pilot cities reaching 68 by 2015. If city $\:c$ implemented BCs in year $\:t$, it is coded as 1 for year $\:t$ and all subsequent years and 0 otherwise. This coding scheme establishes the treatment and control groups for this study.

Control variables

In line with the literature, this study’s regression analysis controls for the influence of various factors, including alcohol consumption [37], socioeconomic status [38], years of education [39], residential location [40], gender [41], marital status [42], and age [43]. The key variables for this study are defined in Table 1 as follows.

Table 1 Variable definitions for the study on the Broadband China strategy and obesity among Chinese older adults (2011–2015)

Descriptive statistics

Table 2 provides the descriptive statistics for the key variables used in this study. The mean BMI was 3.143, with a standard error of 0.157, reflecting a reasonable level of variability within the data. The mean value of the BCs is 0.139, indicating that 13.9% of the samples belong to the treatment group, whereas the remaining samples form the control group. Other variables are not discussed further.

Table 2 Descriptive statistics for key variables in the study on the Broadband China strategy and obesity among Chinese older adults (2011–2015)

Results and discussion

Benchmark regression

This section empirically examines the theoretical analysis presented earlier, with detailed results provided in Table 3. Column (1) presents the baseline regression results, showing that the coefficient of BCs on BMI is 0.0188, which is statistically significant at the 1% level. This suggests that the implementation of the Broadband China strategy resulted in a 1.88% increase in the obesity index among the elderly, confirming that internet usage significantly contributes to an increase in obesity within this demographic, consistent with theoretical expectations. When the effects of the control variables are examined, wages exhibit a significant positive effect on BMI, indicating that higher socioeconomic status is associated with a greater likelihood of obesity among the elderly [44]. Urban residence also has a significant positive effect on BMI, suggesting that elderly individuals residing in urban areas are more susceptible to obesity [45]. Additionally, age shows a significant positive relationship with BMI, implying that advanced age increases the likelihood of obesity among the elderly [46]. The other control variables are not statistically significant and are therefore not discussed further.

To validate the robustness of the baseline regression results, a series of sensitivity tests was performed. Column (2) adopts the methodology of Wang et al. (2024) [47] by employing the C-index (conicity index) as an alternative measure of obesity among the elderly, replacing BMI. The C-index measures the distribution of body fat, particularly the accumulation of abdominal fat, with higher values indicating a greater concentration of abdominal fat. The calculation formula is as follows:

$${\text{C-index = 0}}.{\rm{10}}{{\rm{9}}^{{\rm{ - 1}}}}{\rm{*Waist}}\left( {\rm{m}} \right)*{\left[ {\frac{{{\rm{Weight}}\left( {{\rm{kg}}} \right)}}{{{\rm{Height}}\left( {\rm{m}} \right)}}} \right]^{\frac{{\rm{1}}}{{\rm{2}}}}}$$

(2)

The revised regression results indicate that the coefficient of BCs on the C-index is 0.0185, which is statistically significant at the 5% level, suggesting that the implementation of the Broadband China Strategy resulted in a 1.85% increase in elderly obesity. This finding remains robust. Column (3) similarly adopts the methodology of Wang et al. (2024) [47] by employing the relative fat mass (RFM) index as an alternative measure of obesity among the elderly instead of BMI. Compared with BMI, RFM better reflects the percentage of total body fat and provides a more accurate representation of body fat distribution. The calculation formula is as follows:

$${\text{RFM-men = 64-}}\left[ {{\rm{20*}}\left( {\frac{{{\rm{Height}}\left( {\rm{m}} \right)}}{{{\rm{Waist}}\left( {\rm{m}} \right)}}} \right)} \right]$$

(3)

$${\text{RFM-women = 76-}}\left[ {{\rm{20*}}\left( {\frac{{{\rm{Height}}\left( {\rm{m}} \right)}}{{{\rm{Waist}}\left( {\rm{m}} \right)}}} \right)} \right]$$

(4)

The revised regression results show that the coefficient of BCs on the RFM is 0.0162, which is statistically significant at the 1% level, indicating that the implementation of the Broadband China Strategy resulted in a 1.62% increase in elderly obesity. This finding remains robust. Column (4) adopts the approach of Potter et al. (2024) [48] by employing body fat percentage (BF%), a measure of the percentage of body fat associated with overweight and obesity, as an alternative to BMI for assessing obesity in elderly individuals. The calculation method is as follows:

$$ {\rm{BF\% }} = {\rm{(1}}{\rm{.20}} \times {\rm{BMI)}} + {\rm{(0}}{\rm{.23}} \times {\rm{age)}} - {\rm{(10}}{\rm{.8}} \times {\rm{gender)}} - {\rm{5}}{\rm{.4}} $$

(5)

With males coded as 1 and females as 0, the revised regression results indicate that the coefficient of BCs on BF% is 0.0157, which is statistically significant at the 10% level, suggesting that the implementation of the Broadband China Strategy resulted in a 1.57% increase in elderly obesity. This finding remains robust. In Column (5), the model controls for the Smart City Strategy, which significantly overlaps with the Broadband China Strategy in both the implementation period and policy content, as both strategies aim to promote the adoption and development of new technologies. This overlap could potentially confound the study’s results. After controlling for the Smart City Strategy (Smart), the revised regression results show that the coefficient of BCs on BMI is 0.0187, which slightly decreases but remains statistically significant at the 1% level, confirming that the findings are robust.

In addition, we conducted a series of other sensitivity tests, such as incorporating city-level economic and health variables, using the PSM method, employing various imputation methods to fill in missing values¹, introducing interaction terms between BCs and covariates, and using random effects and GEE models. The results remained robust. For details, please refer to Appendix Tables A, B, C, and D. We also included the confidence intervals and p-values for the BCs coefficients, which can be found in Appendix Table E.

Table 3 Benchmark regression results examining the impact of the Broadband China strategy on obesity among Chinese older adults (2011–2015)

Parallel trend test

The application of DID regression relies on the assumption that the treatment and control groups satisfied the parallel trends condition prior to the implementation of the Broadband China Strategy. This section tests this assumption via two approaches: first, by plotting parallel trends, and second, by conducting a placebo test.

• (1) Parallel Trends Analysis.

Using data from multiple years for the treatment and control groups, this study calculates the average BMI for both groups, as illustrated in Fig. 2. Prior to the implementation of the Broadband China Strategy—in 2011 and 2013—the treatment group (red line) and the control group (blue line) exhibited parallel trends. However, after the policy’s implementation in 2015, the treatment group deviated from the control group’s trend and continued to rise. This preliminary evidence supports the conclusion that the implementation of the Broadband China Strategy contributed to an increase in the BMI of the elderly.

Fig. 2

Testing the parallel trends assumption for the effect of the Broadband China strategy on obesity among Chinese older adults (2011–2015)

• (2) Placebo test.

To assess whether the baseline regression results stem from random group assignment rather than the effect of the Broadband China Strategy, a placebo test was conducted. In this procedure, the treatment and control groups were randomly assigned, and 500 placebo regressions were performed. The resulting placebo regression coefficients were plotted on a density graph. Theoretically, these coefficients should cluster around zero and differ significantly from the true regression coefficient. As illustrated in Fig. 3, the 500 placebo regression coefficients are concentrated around zero and are significantly distinct from the true value (dashed line), indicating that the baseline regression results are indeed driven by the Broadband China Strategy rather than by random factors.

Fig. 3

Placebo test validating the effect of the Broadband China strategy on obesity among Chinese older adults (2011–2015)

Mechanism test

The previously proposed hypotheses identify three potential mechanisms. This section empirically tests these mechanisms, with the specific results presented in Table 4.

• (1) Reduced sleep duration.

To measure sleep duration, this study uses two indicators: nighttime sleep duration (hours) and daytime nap duration (minutes). The regression results are presented in Columns (1) and (2). The coefficient for the effect of the Broadband China strategy on nighttime sleep duration is -0.0254, and the coefficient for daytime nap duration is -0.124. These results indicate that the implementation of the BCs led to reductions in nighttime sleep duration and daytime nap duration of 2.54% and 12.4%, respectively. These findings support Hypothesis 1, suggesting that internet usage significantly decreases sleep duration among the elderly.

• (2) Decreased physical activity levels.

To measure physical activity levels, this study aggregates the number of days in the past week that respondents engaged in heavy, moderate, or light physical activities for more than 10 min, as recorded in the CHARLS survey. The regression results are displayed in Column (3). The coefficient for the effect of the Broadband China strategy on physical activity levels is -0.0983, indicating that its implementation led to a 9.83% decrease in physical activity time among the elderly. These findings support Hypothesis 2, suggesting that internet usage significantly reduces physical activity levels among the elderly.

• (3) Worsened mental health.

This study utilizes depression levels as an indicator of mental health. Depression levels are assessed via the 10-item Center for Epidemiologic Studies Depression Scale (CESD-10), which comprises 10 questions. Each question is rated on a 4-point scale, with values ranging from 1 to 4. Questions 5 and 8 are reverse-scored. The scores from all 10 questions are summed to calculate the elderly depression scale, where higher scores indicate a greater risk of depression among the elderly. The regression results are displayed in Column (4). The coefficient for the effect of the Broadband China strategy on depression is 0.0314, indicating that its implementation led to a 3.14% increase in depression scores among the elderly. These findings support Hypothesis 3, suggesting that internet usage significantly worsens mental health among the elderly.

In addition, confidence intervals and p-values for the BCs regression coefficients are provided; specific results can be found in Appendix Table F.

Table 4 Mechanism test results for the impact of the broadband China strategy on sleep, physical activity, and mental health among Chinese older adults (2011–2015)

Heterogeneity analysis

• (1) Types of Household Registration.

Elderly individuals in urban areas, due to greater digital literacy and better access to internet resources, experience a more pronounced impact of internet usage on BMI. Research indicates a significant positive correlation between internet usage and health behaviors, as well as social participation, among the elderly—particularly among urban residents, who use the internet more frequently for health management and social interaction [49]. The health information and social platforms accessible via the internet may lead to increased dietary intake and reduced physical activity, a trend that is more prominent among the urban elderly. Additionally, urban elderly individuals typically have higher education levels and socioeconomic status, which enhances their ability to benefit from internet usage but may also increase their intake of high-calorie foods and sedentary time [32]. In contrast, elderly individuals in rural areas experience a smaller impact on BMI due to the digital divide and limited internet access. Studies reveal that rural elderly people use the internet less frequently, primarily due to a lack of necessary equipment and skills [50]. Therefore, it is anticipated that the impact of internet usage on BMI will be more pronounced among those with urban household registration.

The elderly population is divided into urban and rural groups on the basis of household registration type, and separate regressions are conducted. The results, displayed in Columns (1) and (2) of Table 5, show that the coefficient for the Broadband China strategy in the urban household sample is greater than that in the rural household sample (0.021 > 0.0175), indicating that the impact of internet usage on BMI is more pronounced among elderly individuals with urban household registration, thereby confirming the hypothesis.

• (2) Number of Chronic Diseases.

For elderly individuals with a greater number of chronic diseases, internet usage can provide enhanced health management resources and support, aiding them in better managing their conditions and indirectly influencing their BMI. For example, the internet can offer detailed information and advice on managing chronic conditions such as hypertension, dyslipidemia, and diabetes, aiding these elderly individuals in adopting effective dietary and exercise plans to maintain or reduce their BMI [32]. Moreover, internet usage can enhance the health knowledge and self-management capabilities of elderly individuals, particularly those with multiple chronic diseases [51]. Conversely, for elderly individuals with fewer chronic diseases, the impact of internet usage on BMI is more closely related to increased social interaction and recreational activities, such as staying connected with friends and family via social media, watching videos, or engaging in online games. However, this can also lead to increased sedentary time and reduced physical activity, potentially resulting in a higher BMI. Thus, the direct impact on BMI may not be as significant as the influence of health management resources for those with multiple chronic diseases [52]. Therefore, it is anticipated that the impact of internet usage on BMI will be more pronounced among elderly individuals with fewer chronic diseases.

On the basis of the responses to 14 chronic disease questions in the CHARLS survey, the elderly population is divided into groups with and without chronic diseases. Separate regressions are conducted for each group. The results, displayed in Columns (3) and (4) of Table 5, indicate that the coefficient for the Broadband China Strategy in the sample without chronic diseases is greater than that in the sample with chronic diseases (0.0201 > 0.00461), confirming that the impact of internet usage on BMI is more pronounced among elderly individuals without chronic diseases.

• (3) Number of surviving children.

The impact of internet usage on the BMI of the elderly may be less significant among those with more surviving, healthy children. Elderly individuals with healthier children often receive increased social support and care. These children may teach their parents how to use the internet and encourage them to utilize online health information and management resources [52]. They often monitor their parents’ health by offering nutritional advice and exercise guidance, which assists the elderly in maintaining a healthy weight and reducing BMI. Furthermore, healthy children may regularly remind and encourage their parents to adhere to a healthy lifestyle, thereby mitigating the adverse health behaviors that may arise from internet usage. For example, despite the convenience of online food delivery and social platforms, under the supervision of healthy children, elderly individuals are more inclined to choose healthier diets and engage in moderate exercise rather than adopt a sedentary lifestyle. This supervision and encouragement can partially offset the negative effects of internet usage, aiding in the maintenance or reduction of BMI among the elderly. Therefore, it is anticipated that the impact of internet usage on BMI will be weaker among elderly individuals with more surviving children.

On the basis of the median number of surviving children in the CHARLS data, the elderly population is divided into two groups: those with many surviving children and those with few surviving children. Separate regressions are conducted for each group. The results, displayed in Columns (5) and (6) of Table 5, indicate that the coefficient for the Broadband China strategy in the sample with fewer surviving children is greater than that in the sample with more surviving children (0.02 > 0.0121), suggesting that the impact of internet usage on BMI is stronger among elderly individuals with fewer surviving children, thereby confirming the hypothesis.

In addition, to ensure significant statistical differences between the group regression coefficients, an interaction term test was conducted, incorporating the interaction between BCs and the group dummy variables. The specific results are presented in Appendix Table G.

Table 5 Heterogeneity analysis of the Broadband China strategy on obesity among Chinese older adults by household registration, chronic diseases, and surviving children (2011–2015)

Conclusion

This study leverages China’s “Broadband China” strategy as a quasi-natural experiment and employs a DID approach to examine the causal relationship between internet usage and elderly BMI. The results indicate that internet usage significantly increased BMI among the elderly. This effect is particularly pronounced among elderly individuals with urban household registration, those without chronic diseases, and those with fewer surviving children. The mechanism analysis indicated that internet usage significantly increased BMI among the elderly through reduced sleep duration, decreased physical activity, and worsened mental health. These findings not only elucidate the causal relationship between internet usage and elderly BMI but also provide rigorous empirical evidence, offering robust scientific support for public policy formulation. Therefore, targeted public health interventions are recommended to help elderly individuals use the internet more judiciously, control obesity rates, and improve overall health.

Based on the findings of this study, three policy recommendations are proposed. First, public health education programs related to internet usage should be promoted for the elderly. The government should encourage community and healthcare institutions to implement internet health education programs tailored specifically for the elderly. Such programs should instruct seniors on how to use the internet effectively to access health information, manage their diet and exercise routines, and avoid sedentary behavior and insufficient sleep resulting from excessive internet usage. These programs could combine online and offline delivery methods to ensure broader reach among elderly populations. Second, community activity facilities and programs for the elderly should be enhanced. To mitigate the decline in physical activity due to internet usage, communities should develop and improve facilities and activities tailored for the elderly. This includes establishing dedicated fitness facilities and organizing regular health-focused activities that encourage outdoor exercise and social interaction. Such initiatives can improve physical activity levels, promote social engagement, and positively impact mental health, thereby reducing the adverse effects of internet usage on BMI. Third, psychological health support services should be provided for the elderly. Given the potential negative impact of internet usage on mental health and its subsequent effect on BMI, offering targeted psychological support services within communities and healthcare settings is crucial. These services should include psychological counseling, depression screening, and appropriate intervention measures to identify and address mental health issues related to internet usage, ultimately helping elderly individuals develop healthier lifestyles.

This study has two limitations. First, it utilizes three waves of longitudinal panel data covering a relatively short period, which may not fully capture the long-term impact of internet usage on elderly BMI. Future research could address this limitation by extending the data collection period and conducting longer-term tracking studies to gain a more comprehensive understanding of the sustained effects of internet usage on elderly health. Second, while this study focuses on the overall impact of internet usage, it does not examine how different types of internet activities (such as social networking, online shopping, and health information searching) may differentially affect BMI. Different forms of internet usage might have varied effects on elderly health. Future research should investigate these specific behaviors to uncover their distinct mechanisms and impacts on BMI.

Appendix

Table A Exogeneity test of the Broadband China strategy on obesity among Chinese older adults (2011–2015) using city variables and propensity score matching

Table B Balance test of sample characteristics between treated and control groups among Chinese older adults (2011–2015)

Table C Imputation regression results for the Broadband China strategy effect on obesity among Chinese older adults (2011–2015) using different missing value strategies

Table D Sensitivity test of model specifications for the Broadband China strategy effect on obesity among Chinese older adults (2011–2015) using interaction terms, random effect, and generalized estimating equation models

Table E Supplementary confidence intervals and p-values for baseline regression results on body mass index, conicity index, relative fat mass, and body fat percentage among Chinese older adults (2011–2015)

Table F Supplementary confidence intervals and p-values for the mechanism test on sleep, physical activity, and depression among Chinese older adults (2011–2015)

Table G Interaction term test for heterogeneity analysis of the Broadband China strategy effect on obesity among Chinese older adults (2011–2015)