Does internet use make farmers happier? Evidence from Indonesia

Abstract The literature on the role of Internet use on subjective well-being has grown over the years. However, evidence on how Internet use affects farmers’ subjective well-being (happiness and life satisfaction) remains scarce. This research fills the gap by analyzing the effects of Internet use on farmers’ subjective well-being. This study provides two essential contributions. First, we provide novel empirical evidence on the association between Internet use and farmers’ subjective well-being. Second, we employ national representative data from Indonesia, consisting of 7221 urban and rural farmers. A two-stage residual inclusion (2SRI) strategy was used to tackle the endogeneity problems in the decisions to use the Internet. The finding indicates that Internet use positively and significantly affects farmers’ happiness and life satisfaction. The disaggregate analysis also reveals that Internet use positively and significantly affects happiness at income quintiles 1 and 3. However, the effect of Internet use on life satisfaction was only significant at the lowest income (farmers’ income quintile 1). Our study suggests developing Internet use in agricultural sectors to boost farmers’ subjective well-being.


Introduction
The role of the Internet in developing the agricultural sectors has become increasingly crucial and necessary.Research has shown how farmers can benefit from the Internet.First, farmers can find solutions to their problems dealing with weather, agricultural input, and technological innovation (Ma & Zheng, 2021;Zheng & Wanglin, 2021).Second, farmers can connect more easily with other farmers, agricultural institutions, and academics as the Internet is more affordable and interactive than other means of communication, such as the telephone (Barton, 2003).Third, the Internet improves farmers' marketing opportunities by opening up potential markets with better prices for their agricultural products (Owusu et al., 2018;Rolfe et al., 2003).For instance, Internet use has helped banana farmers in Uganda to obtain higher sales (Muto & Yamano, 2009).In other words, Internet use can potentially develop the agriculture sector.Therefore, researchers have focused studies on the impact of Internet access on economic performance.
The literature has documented the effects of Internet use on agricultural sectors, which can be grouped into two strands of literature.The first strand of literature focuses on the economic effects of Internet use, such as income (Ma et al., 2020;Siaw et al., 2020;Zhang et al., 2021) and technical efficiency (Zheng et al., 2021).For example, a study by Siaw et al. (2020) estimates the effect of Internet access on farmers' welfare, indicating a positive and significant effect on both farm and household incomes.Similarly, Ma et al. (2020) and Zhang et al. (2021) argued that Internet use is significantly associated with higher farm income.Meanwhile, Zheng et al. (2021) analyzed crosssectional data from a survey in China and revealed that Internet use increased farmers' technical efficiency.
The second strand of literature investigates the relationship between Internet access and sustainability practices in agriculture (Ma & Wang, 2020), such as fertilizer and pesticide use (Yuan et al., 2021;Zhao et al., 2021).For example, a study by Yuan et al. (2021) estimates the impact of Internet access on the use of chemical fertilizer in China.They found that Internet access significantly reduced chemical fertilizer use by 12.6%.Meanwhile, employing Probit and bootstrap methods, Zhao et al. (2021) discovered that Internet access increased pesticide use.Ma and Wang (2020) found that Internet use enables farmers to adopt higher sustainable agriculture practices, such as organic fertilizer, pesticides, modern variety, and integrated pest management.
Besides the economic impacts, Internet access improves subjective well-being, proxied by happiness and life satisfaction.Subjective well-being is an essential well-being dimension, referring to the personal cognition and emotive assessment of life quality (Diener & Ryan, 2009;Park, 2004;Z. Zhu et al., 2021).Past research has explored the association between Internet use and subjective well-being, but the findings are mixed.Past studies have shown a positive association between Internet use and subjective well-being.For instance, Lu and Kandilov (2021) investigated the association between Internet use and the subjective well-being of older Chinese adults.They revealed that mobile Internet use was significantly and positively associated with higher subjective well-being.Likewise, Hong and Chang (2020) compared households with and without Internet access, revealing that those with Internet access had 10% higher life satisfaction.Meanwhile, Yang et al. (2021) used propensity score matching and revealed that Internet use promoted personal skills and human capital, helping rural households improve life satisfaction and happiness.Lam et al. (2020) analyzed data from a survey of older adults in England and revealed that Internet access increased communication, information access, and electronic commerce, lowered depression, and improved life satisfaction.Similarly, Nie et al. (2020) indicated the positive impacts of Internet use on happiness and life satisfaction in rural China.The Internet use frequency can also improve subjective well-being (Cuihong and Chengzhi, 2019).
However, studies also showed that Internet use may be negatively associated with subjective well-being.Koç (2017) showed that excessive Internet use negatively affected farmers' subjective well-being because it may trigger negative emotions.Büchi et al. (2019) claimed a similar finding, stating that the overuse of digital media strongly correlates with low subjective well-being.
It should be noted that the studies above involve rural residents in general (Cuihong & Chengzhi, 2019;Nie et al., 2020;Yang et al., 2021), students (Koç, 2017;Lei et al., 2020), and older adults (Lam et al., 2020;Lu & Kandilov, 2021).No studies have examined the effect of Internet use on subjective well-being, proxied by happiness and life satisfaction, involving farmers.The findings may differ because farmers and non-farmers have varying economic circumstances, educational levels, and the purpose of Internet use.To fill this gap, this study investigates the effects of Internet use on farmers' subjective well-being.The findings should provide valuable insights for agriculture development policy in Indonesia and other countries.
In this study, happiness as short-term subjective well-being is defined as the immediate joy and contentment derived from daily experiences on the farm.It includes finding delight in successful harvests, nurturing livestock, and being connected to nature (Rahman et al., 2023;Wijayanto et al., 2022).Meanwhile, life satisfaction represents farmers' long-term subjective well-being and reflects the overall evaluation of their lives, including financial stability, job security, family support, and the fulfillment of farming-related goals and aspirations (Rahman et al., 2023;Wijayanto et al., 2022).These two factors contribute to farmers' subjective well-being, shaping their perceptions of contentment and fulfillment in farming.In addition to subjective well-being, this study also investigates the differences in characteristics among farmers using and not using the Internet in Indonesia.
This study contributes to the literature on agriculture development and farmers' subjective well-being from three angles.First, this study involves farmers in rural and urban areas to estimate the effects of Internet use on farmers' subjective well-being.Farmers are interesting respondents because they have different characteristics (i.e., economic conditions, education levels, and access to the Internet) from other groups, such as rural and urban residents, students, and older adults, which have become the focus of past studies.Farmers also tend to fall behind in technology and information development.Second, we used national-level data from Indonesia, which can be generalized for all farmers in Indonesia.The finding of this study can be applied to other developing countries to inform agriculturerelated policymaking.For example, the Internet can revolutionize the delivery of agriculture development policies.With its widespread accessibility, the Internet can help disseminate crucial information effectively and bridge policymakers and farmers.Real-time updates on weather patterns, market trends, and innovative farming techniques can be accessed through digital platforms anywhere.Therefore, policymakers can reach a broader audience and ensure that agriculture development policies are communicated efficiently and transparently.The Internet also facilitates knowledge sharing among farmers and fosters collaboration and exchange of best practices.As such, agriculture development policies can be implemented more effectively to achieve more sustainable and inclusive growth in the agricultural sector.Finally, this study estimates the disaggregate effects of Internet use on happiness and life satisfaction by household income quantile, so it captures the heterogeneous effects.The result of this study is novel because previous studies only investigated the homogenous effects of Internet use on subjective well-being (Cuihong & Chengzhi, 2019;Koç, 2017;Lam et al., 2020;Lei et al., 2020;Lu & Kandilov, 2021;Nie et al., 2020;Yang et al., 2021;Z. Zhu, Wanglin, Sousa-Poza, et al., 2020).
The remaining article is organized as follows.Section two presents the estimation strategy.Section three describes the research data, variables measurements, and descriptive statistics.Section four presents the results and discussion.Section five concludes the study and provides the policy implication.

Internet use for farmers' subjective well-being
In this study, subjective well-being indicators, i.e., happiness and life satisfaction, are assumed to be affected by Internet use and other control variables, including age, gender, vehicle ownership, education, health, family size, cooperative membership, and television ownership.Age is considered a significant predictor of subjective well-being as it reflects the changing individuals' circumstances and priorities across stages of life (Bussière et al., 2021;Rahman et al., 2023).Regarding gender, research has also shown that men and women have varying experiences and factors influencing their well-being (Ma et al., 2022).Vehicle ownership has an influence, too, as it makes commuting more convenient and reduces commute-related stress.Access to a reliable and efficient vehicle ownership system positively affects individuals' well-being (Nugroho et al., 2022).Education is frequently linked to subjective well-being, with higher levels of education being associated with greater life satisfaction.Education equips individuals with knowledge and skills and opens opportunities that can enhance their overall well-being (Rahman et al., 2022).Health condition is a crucial determinant of subjective well-being.Poor health can significantly impact one's quality of life and overall happiness.Conversely, good health is often associated with higher levels of subjective well-being (Dury et al., 2021;Schneider et al., 2004).The number of family members may influence subjective well-being due to family relationships and support systems dynamics.Social connections and a supportive family environment can impact well-being positively (Li & Cheng, 2015).Cooperative membership reflects participation in collective activities and community engagement.Research suggests that participating in cooperative or similar organizations can enhance social capital, sense of belonging, and overall well-being (Qi et al., 2022).Television ownership represents access to media and entertainment.Research findings on the relationship between television ownership and subjective well-being vary, but some have shown that excessive television viewing can be associated with decreased well-being (Rieger et al., 2014).Meanwhile, the formulation is specified as follows. Description: (1) S j i represents the i th farmers' ordinal subjective well-being, including happiness and life satisfaction for j = 1 and j = 2, respectively (2) I i represents a dummy variable of Internet use (1 for Internet users and 0 for others) (3) X i is a vector of control variables (4) @ i and a i are parameters to estimate, and e i is an error term Equation 1 could be executed by an ordered Probit to estimate the subjective well-being effects of Internet use on farmers' well-being.The estimation requires that the Internet use variable should be exogenous.However, the Internet use variable in Equation 1 is potentially endogenous because farmers decide whether to use the Internet depending on the observable and unobservable factors.Observable factors refer to directly measurable variables (observed) in a study, such as income, education level, or age.These factors are typically included in the statistical model and are considered exogenous if not affected by the endogenous variables of interest.On the other hand, unobservable factors, also known as confounders, are not directly measurable or observed.They represent factors that may affect the model's dependent and independent variables but are not included as explanatory variables.Unobservable factors can cause a biased estimate of the relationships between variables, resulting in endogeneity.Endogeneity refers to a situation where the treatment variable is correlated with the error term in the regression model, leading to biased and inconsistent estimates.In particular, the unobservable variable (i.e., farmers' motivation) may simultaneously influence farmers' subjective well-being and decision to use the Internet.In this case, Equation 1 will fail to address the endogeneity problem, and the estimation will be biased.Following previous studies (Liu & Gang, 2021;Ma & Zhu, 2020;Z. Zhu et al., 2021), this study employs a two-stage residual inclusion (2SRI) approach to address the endogeneity problem of a binary treatment variable (i.e., Internet use).The 2SRI approach can address the endogeneity problem in econometric analysis's causal-effect estimation of a treatment variable.

Two-stage residual inclusion (2SRI)
The first stage of the 2SRI procedure estimates the likelihood of Internet users using a Probit model as follows. Description: (1) I � i represents the likelihood of Internet use and is measured by a dichotomous variable (I i = 1 for Internet users and I i = 0 otherwise) (2) X i is a vector of the control variables, as mentioned earlier (3) IV i is an instrumental variable (4) b i and φ i are variables to be assessed, and u i is an error term Regarding the instrumental variable (IV), this study uses the average household members with access to the Internet by region (measured by person).The IV is expected to influence farmers' choices to utilize the Internet but insignificantly affects farmers' happiness and life satisfaction.The IV in this study is validated using the Pearson correlation test (see Table 1).The findings indicate that the IV is significantly and positively correlated with Internet use but insignificantly correlated with happiness and life satisfaction.Hence, the findings suggest that the average household member with access to the Internet is appropriate as the IV of Internet use.
The second stage of the 2SRI approach estimates the effects of Internet use on farmers' subjective well-being by regressing the model in Equation 1 and including the residual term.The residual term is predicted after estimating the Probit model in the first stage of the 2SRI (Equation 2) and was employed as an additional variable in the following model.

Description:
(1) S j i is a vector of subjective well-being variables, including happiness and life satisfaction, and is mentioned in Equation 1.
(2) I i and X i are defined earlier (3) Residual i represents the residual term predicted after estimating Equation 2. The endogeneity issue of the Internet use variable is identifiable if the residual term shows a significant effect in Equation 3 (Ma & Zhu, 2020).(4) ω i , σ i , and τ i are parameters to estimate, and 2 i represents the error term

Research data
This study employs open-access data from the Indonesian life survey family 5 (IFLS-5), collected from 13 provinces in Indonesia.This behavior and outcomes study was initiated by RAND Corporation in collaboration with Survey Matters and Indonesian universities.The IFLS sample represents 83% of the Indonesian population.Stratified sampling based on the province scheme was used in selecting 321 areas randomly from 13 provinces (Sujarwoto, 2021).The IFLS-5 surveys were conducted from October 2014 to August 2015, 15067 households.As this study focuses on farmers in rural and urban areas, non-farmer samples were excluded from the empirical analysis, resulting in 7,585 farmers' household.Then, selected respondents with missing data were excluded from the estimation, resulting in 7,221 farmers.The sample was distributed across 13 provinces in Indonesia, with 5,462 farmers (75.64%) living in rural areas and 1,759 (24.36%) in urban areas.The detailed information on the IFLS survey can be accessed at https://www.rand.org/well-being/social-and-behavioral-policy/data/FLS/IFLS.html.
The IFLS data contains much information, such as respondents' demographics, socio-economics, food and non-food consumption, social capital, health, economic, and subjective well-being.However, this study necessitates data sifting, focusing only on subjective well-being, i.e., happiness and life satisfaction, Internet access, and socio-demographic condition, i.e., age, gender, vehicle ownership, education health condition, number of families, cooperative membership, and television ownership.
This study assesses farmers' subjective well-being using two indicators, i.e., happiness and life satisfaction, following the previous studies (Nie et al., 2020;Sujarwoto et al., 2018;Z. Zhu et al., 2021;Rahman et al., 2023).First, based on the IFLS questionnaire, the farmers were asked, "All things considered, how would you say things are these days-would you say you were very happy, happy, unhappy, or very unhappy?"They answered using a four-scale rating system, where 1 = very unhappy, 2 = unhappy, 3 = happy, and 4 = very happy.For the life satisfaction measurement, the respondents were asked, "Considering your life as a whole, how satisfied are you?"They answered using a five-scale rating system, where 1 = not satisfied at all, 2 = not very satisfied, 3 = somewhat satisfied, 4 = very satisfied, and 5 = completely satisfied.Previous studies have pointed out that happiness is a short-term subjective well-being indicator that measures the emotional state and perception of life daily.Meanwhile, life satisfaction measures long-term thoughts and feelings about life as a whole (Z.Zhu et al., 2021).Lastly, the treatment variable of this study is Internet use, represented as a dummy variable, scoring 1 if the farmer uses the Internet and 0 otherwise.In addition, the average number of household members with access to the Internet by region was taken from the 2014 data by Statistics Indonesia or Badan Pusat Statistik (BPS).

Descriptive statistics
This section explains the measurement and descriptive statistics of the research variables (Table 2).The results showed that the average value of the happiness variable was 2.970, and the life satisfaction variable was 3.191, which suggests Indonesian farmers' subjective well-being is good.The Internet use variable shows that 12.8% of farmers in this study used the Internet, and 86.8% did not.The average age was 43 years old.Gender-wise, 40% were female and 60% male.Most farmers (81.5%) were married.Most (64.3%) also have private vehicles like cars, boats, bicycles, and motorbikes.Table 2 also presents the farmers' education level measured as dummy variables.The data show that 8.7% of farmers have no formal education, 4.92% finished primary school, 20% junior high school, and 12.8% senior high school.A minority (1%) earned an associated degree, and 2.3% a university degree.The average health conditions of the respondents were 2.926.On average, the respondents had four family members.Only a few (3%) of them were cooperative members.Most (83.6%) have a television.Finally, the average income was 844,593.5 Rupiah/ month.
Table 3 presents the differences in characteristics between Internet-user farmers and non-Internet-user farmers.The results indicate that the two groups were systematically different.For instance, Internet-user farmers were often young, more educated, healthy, male, earned higher incomes, and lived in urban areas.

Determining farmers' decision to use the Internet: The first stage of the 2SRI model
Farmers' decision to use the Internet was estimated by Equation 2 using a Probit model in the first stage of the 2SRI, as presented in Table 4.The study estimation indicated that farmers' decision to use the Internet was significantly affected by age, gender, marital status, vehicle ownership, education, cooperative membership, TV ownership, income, and urban domicile.However, health conditions and family size had no significant effect.
Age negatively and significantly affects farmers' decisions to use the Internet, with younger farmers being more likely to use the Internet than older farmers.Older farmers are usually left behind in technology, including the Internet, and they are not always aware of the advantages.This result aligns with Ma and Zhu (2021), who reported a negative correlation of age on Internet use among Chinese citizens.Meanwhile, the negative and significant effects of the gender indicate that men farmers use the Internet more frequently than women farmers.In Indonesia, men are often the decision-maker in the agricultural sectors, so they are more likely to use the Internet to inform their decision-making.Z. Zhu, Wanglin, and Leng (2020) also pointed out a positive relationship between men and Internet use.The marital status variable also negatively and significantly affects the adoption of information and communications technology, such as the Internet.
Unmarried farmers tend to use the Internet more intensively than married farmers.In terms of education, low education levels-no education, primary and junior high school graduates-negatively and significantly affect farmers' decisions to use the Internet.Accordingly, higher education levels-associate and university degrees-positively and significantly affect farmers' Internet decisions.This finding suggests that education encourages farmers to seek information and equips them with the knowledge to adopt technology quickly.This finding agrees with Mittal and Mehar (2016), claiming a positive association between education level and Internet adoption.
The coefficient of cooperative membership was positively and statistically significant.Cooperative members were positively associated with Internet adoption because cooperatives enabled them to adopt technological innovation more easily, such as the Internet.This finding is consistent with Manda et al. (2020) and Zhang et al. (2021), revealing a positive association between cooperative membership and technology adoption in the agricultural sectors.Likewise, television ownership showed a positive and statistically significant coefficient, suggesting it encourages Internet use.This may suggest that the entertainment and information enjoyed by television could motivate more technology adoption, such as the Internet.Finally, the statistically significant and positive income coefficient indicates that higher-income farmers were more likely to adopt the Internet.Higher-income means higher purchasing power, including purchasing devices and data packages.This finding aligns with (Ramírez-Hassan & Carvajal-Rendón, 2021), revealing the positive effect of income on Internet adoption.Finally, the IV showed a positive and statistically significant coefficient, suggesting that the average number of household members with access to the Internet improves a farmer's probability of accessing the Internet.They may be motivated by family members who are also avid Internet users.The household Internet access variable was utilized as an instrumental variable (IV) of Internet users in this study.Hence, we did not expect the household Internet access variable to significantly affect subjective wellbeing, such as happiness and life satisfaction.

Effects of internet use on subjective well-being
Table 5 presents the results of the second stage of the 2SRI estimation.The ordered Probit model was employed in this stage based on Equation 3. We mentioned earlier that Equation 3 includes the residual predicted from the Probit model (the first stage of the 2SRI) to address the endogeneity problems linked with the unobservable variables.The result indicated that the residual variable significantly impacts the dependent variables: happiness and life satisfaction.This result suggests an endogeneity problem associated with the Internet use variable.Therefore, this result confirms the appropriate use of 2SRI in estimating the effects of Internet use on subjective well-being.
Table 5, columns 2 and 3, show that Internet use positively and significantly affects happiness and life satisfaction, implying that the Internet could improve farmers' subjective well-being.Farmers using the Internet are happier and more satisfied with life than those who do not.This finding confirms the results of past studies by Lu and Kandilov (2021), Hong andChang (2020), andYang et al. (2021) in China; Lam et al. (2020) in England;and Omar et al. (2019) in Malaysia.This study adds new evidence that Internet use increases Indonesian farmers' subjective well-being.The ability to access agricultural information to solve farm-related problems and other needs can make farmers happier and more satisfied with life.
This study also shows the marginal effects of the ordered Probit model in the second stage because the coefficients in Table 3 are unsuitable for interpretation.To simplify, Table 6 only presents the marginal effects of our key variable, namely Internet use.The results of the control variables' marginal effects are available upon request.They showed that, on average, the effect of Internet use on happiness and life satisfaction is positive and statistically significant.For example, Internet use improves the probability of farmers' feeling "very happy" by 5.9% and "very satisfied" by 5.8%.It also reduces farmers' The feeling "unhappy" by 1.0% and "not very satisfied" by 1.0%.

The results of the control variables
In addition to the Internet use variable, the estimation of the 2SRI model in the second stage reveals the effects of the control variables (farmers' characteristics) on subjective well-being, as shown in Table 5, columns 2 and 3. Age negatively and significantly influences happiness and life satisfaction, suggesting that young farmers are happier and more satisfied than old farmers.Young farmers tend to have more power, less stress, and more attractiveness, so their emotional state tends to be better.Sujarwoto (2021) supports this finding, highlighting a negative relationship between age and the subjective well-being of Indonesian residents in general.Meanwhile, gender positively and significantly affects life satisfaction (column 2) but insignificantly affects farmers' happiness (column 3).This indicates that women farmers are more satisfied with their lives than men farmers.Likewise, marital status positively and significantly affects subjective well-being in happiness and life satisfaction.In other words, marriage helps farmers improve their subjective well-being.In Indonesia, marriage is often considered a primary need, especially for Muslims.Being married is often associated with happiness and life satisfaction.Anna et al. (2019) showed similar findings in the Indonesian context, highlighting the positive relationship between marital status and subjective well-being.
Vehicle ownership shows positive and statistically significant coefficients, suggesting that farmers with private vehicle ownership tend to be happier and more satisfied with life.This finding is not surprising because vehicle ownership helps farmers improve their mobilization.Owning a vehicle also enables farmers to sell agricultural inputs or outputs more easily, making them happier and more satisfied.The health condition variable has a positive and significant relationship with happiness and life satisfaction.The healthier the farmers, the happier and more satisfied they become (Z.Zhu et al., 2021).Similarly, television ownership positively and significantly affects farmers' happiness but insignificantly affects life satisfaction.The finding suggests that television only improves subjective well-being in the short term, i.e., happiness.Finally, the income variable is positively and significantly associated with happiness and life satisfaction.This result suggests that farmers with better incomes tend to be happier and more satisfied than those with lower incomes.This result agrees with previous studies by Sohn (2013) and Sujarwoto (2021) in Indonesia, Z. Zhu et al. (2021) in China, andCuong (2021) in Vietnam.

The disaggregated analysis
This study employs a disaggregate analysis to provide heterogeneous results of the Internet use effects on subjective well-being by considering farmers' income quantiles.The estimation results are presented in Table 7. Internet use at income quantiles 1 and 3 positively and significantly affects happiness.However, with increasing income, the effect of Internet use on happiness decreases.We also found that Internet use has a statistically significant effect on life satisfaction at the lowest income level (quantile 1).However, this is not the case among farmers with higher income levels in quantiles 2, 3, and 4. Internet use is usually associated   with higher income levels.When farmers with lower incomes use the Internet, their subjective well-being tends to rise.

Conclusion
This study provides new empirical evidence of the effects of Internet use on Indonesian farmers' subjective well-being, proxied by happiness and life satisfaction.The data were based on a survey with 7,221 farmers from the Indonesian Life Survey Family 5 (IFLS-5).A two-stage residual inclusion (2SRI) model was used to address the endogeneity problem of Internet adoption.In addition, this study analyzes the heterogeneous effects of Internet use on subjective well-being by considering income levels.
Our study highlights that the characteristics of Internet users and non-users are systematically different.Farmers who use the Internet tend to be young, more educated, healthier, and mostly male.The empirical results also indicate that Internet use positively and significantly affects both subjective well-being measurements, i.e., happiness and life satisfaction.Furthermore, the heterogeneous effects of Internet use show a positive and statistically significant effect on happiness at all income levels from quantiles 1 to 4. However, this positive effect is greater in the lowest income level (quantile 1).Meanwhile, Internet use's positive and significant effects on life satisfaction only apply to the lowest income level (quantile 1) but not to the higher income levels (quantiles 2 to 4).This study also points out that vehicle ownership, education, cooperative membership, television ownership, income, and urban residency positively and significantly affect Internet use.By contrast, age, gender, and marital status negatively and significantly affect their decisions.The subjective well-being is also affected by farmers' basic characteristics.Gender (male), marital status, vehicle ownership, health, and income positively and significantly affect farmers' subjective well-being.Meanwhile, age has a negative effect on farmers' subjective well-being Since Internet use positively affects farmers' subjective well-being, this study suggests that further research aims to assist farmers in improving their subjective well-being.Therefore, the government must improve the Internet networks, access, and speed in rural areas so that farmers can access the Internet to improve their welfare.In addition, the government can provide Internet optimization training, especially for the older farmers living in rural areas, to improve their understanding and digital skills.In doing so, the government can collaborate with Internet providers, extension agents, rural cooperatives, and farmer groups.

Table 4 . Results from the first stage estimation of the 2SRI model Variable Internet usage (coefficient)
Note: ** and *** denote significance on 5% and 1 %, respectively.The values in parentheses represent the standard errors.