Can Internet Use Improve Farmers' Welfare Effect ——A Case Study of Chinese Vegetable Growers

： This article uses the data of 797 vegetable growers in Shouguang, Shandong Province, 8 and the endogenous conversion model to investigate the impact of Internet use on household welfare. 9 We select the per-mu vegetable yield, net income, and per capita net income of households as 10 welfare indicators. The results show: ① Internet use can significantly improve the farmers' welfare 11 effect. ② Under counterfactual assumptions, when farmers who use the Internet do not use it, the 12 farmers' vegetable yield per mu, net income and household per capita net income will drop by 13 10.88%, 13.96% and 9.46%. When farmers who do not use the Internet use it, the farmers' vegetable 14 output, net income and family per capita net income will rise by 13.62%, 16.66% and 11.64%. 15 Internet use has the most excellent effect on the net income of vegetables, followed by the yield per 16 mu, and the net income per household is the lowest. ③ Compared with small-scale farmers, 17 Internet use has a better impact on the welfare of large-scale farmers, which also widens the welfare 18 gap between farmers to a certain extent. Based on this, make suggestions to strengthen information 19 infrastructure, improve information technology training, and adjust support policies promptly. 20


Introduction 24
The national nature of socialism with Chinese characteristics needs to guarantee farmers' 25 welfare to achieve social harmony and stability and long-term stability. The welfare effect is an 26 important indicator to measure economic development, social progress and people's living standards. 27 2 Therefore, exploring the driving factors of the improvement of farmers' welfare effect has important 28 practical significance for improving farmers' quality of life and improving farmers' economic status. 29 At present, agricultural informatization is a vital engine driving rural economic development (Yunis, 30 et al.,2018). Existing literature has carried out detailed research on the relationship between 31 informatization and social production and found that information technology not only affects 32 economic output and farmer welfare, but also plays a crucial role in the national development farmers' welfare is conducive to exploring the "hematopoietic" mechanism of Internet use and 51 provides useful explorations for rural informatization, modernization, and farmers' changes welfare. 52 At present, there are many controversies about whether the impact of Internet use on the 53 welfare of farmers is "information welfare" or "information gap," and there is still no consensus 54 among the theoretical circles. Some scholars believe that Internet users can obtain and share the 55 latest knowledge and information (Guo et al.2017) and improve farmers' economic status and market 56 3 competitiveness (Dimaggio & Bonikowski,2008). The information economics argument also 57 believes that the Internet's information can reduce individual information search costs, thereby 58 profoundly affecting its economic development, technological progress, and production 59 efficiency (Stigler,1961). That is, Internet use has "information benefits." However, some scholars 60 have found that while the use of the Internet improves farmers' production efficiency and economic 61 development, it may also produce inequality in human society (Tan et al.,2017). Because people 62 with different income levels or education levels have large differences in the degree of acceptance 63 and use of the Internet, which leads to large differences in the economic benefits of recipients and 64 exclusions, thus widening the gap between the rich and the poor (Heinz,2002). That is, the 65 "information gap" formed by the use of the Internet. Based on the above analysis, although scholars 66 have conducted a lot of empirical analysis and theoretical discussion on the economic effects of 67 Internet use on farmers, there are still problems that need to be studied in depth. It mainly includes: 68 First, whether the impact of Internet use on farmers' welfare increases or decreases has not yet 69 reached a consistent conclusion. Second, most studies did not include Internet use in the entire link 70 of agricultural production, nor did they explore the economic effects of Internet use from the 71 perspective of farmers' welfare. Third, most existing studies use the instrumental variable method 72 and propensity score matching method to test the impact of Internet use on agricultural production. 73 The former ignores the heterogeneity of the treatment effect, while the latter does not consider its 74 endogeneity, both of which may cause deviations in the estimated results. 75 Given this, the article uses the per-mu vegetable yield, net income, and household income 76 indicators as household welfare indicators. Based on a sample of 797 vegetable growers in 77 Shouguang, Shandong Province, the endogenous conversion model was used to analyze the impact 78 of Internet use on farmers' welfare and the differences in the welfare effects of farmers of different 79 sizes. Compared with the previous literature, the article incorporates the Internet's use into the 80 overall analysis framework of vegetable production, price and income. It pays attention to the 81 changes in its welfare from the production link to the sales link and supplements the previous use 82 of a single indicator to measure farmers' welfare. To ensure the article's accuracy and stability, it 83 chooses the endogenous transformation model (ESR) to solve sample endogeneity and uses the 84 seemingly uncorrelated regression model to conduct robustness tests. 85 Table 1 The scale of Chinese rural netizens from 2014 to 2020 87 1 Theoretical analysis and model setting 88

Farmer welfare effect 89
Welfare is the degree of satisfaction that an individual consumes goods or services. However, 90 because the "degree of satisfaction" is difficult to measure, individual welfare is usually expressed 91 by economic practice. The higher the economic level, the better the personal welfare status. Farmers 92 are rational economic people whose ultimate goal is to maximize profits. Therefore, this article 93 selects economic welfare as its research object. The theory of social practice believes that: livelihood 94 capital is the sum of resources and abilities owned by an individual or family, and it is also the basis 95 for affecting individual practical activities and business performance (Bourdieu,1986 knowledge and skills. Improving human capital is a crucial way to increase learning ability, 99 knowledge accumulation and innovation awareness (Fleisher,2010). Natural capital (such as land) 100 is the fundamental guarantee for rural social stability and farmers' survival. Material capital is a 101 condition for farmers to maintain good initiative and strong resistance to market competition 102 pressure. Simultaneously, physical capital is also an indirect force that widens the gap in farmers' resources based on the same cognition (Bourdieu,1986), which can provide individuals with 107 The size of rural netizens in China from 2014-2020 5 instrumental and emotional support (Lin,2001). 108 In addition to the traditional livelihood capital, farmers' welfare is also affected by the prices welfare analyzes the whole process from the most basic cost input to the output and income of 120 agricultural products and then to farmers' income. Therefore, this article uses per-mu yield, net 121 income, and household net income per capita as indicators to explore the impact of Internet use on 122 farmers' welfare. 123 Internet use provides farmers with comprehensive agricultural information services. 124 Simultaneously, the Internet user's accuracy and convenience can also reduce the procedures and 125 costs of obtaining information, allowing farmers to make optimal production decisions 126 (Fafchamps&Minten,2012). The accurate market information brought by Internet use can reduce 127 agricultural transaction costs, stimulate agricultural production and increase agricultural income 128 (Aker,2014). Mainly manifested in the following aspects: (1) The use of the Internet helps improve 129 the ability to collect and use information so that the supply of agricultural products can effectively 130 connect the demand to guide production better and increase output (Baorakis,2002). The Internet 131 provides accurate, dynamic and scientific all-around information services for agricultural 132 production and management. The intelligent management and precise services it brings to 133 agricultural production effectively increase agricultural products' output (Garrett,2013). From a 134 long-term perspective, using the Internet can change the agricultural planting structure and introduce 135 new varieties, thereby increasing agricultural productivity (Nakasone,2014).
(2) The Internet use 136 can also solve information asymmetry, optimize the allocation of traditional production factors such 137 6 as land, capital, and labor, thereby changing the current "high input, high consumption, and high 138 pollution" economic development mode. It can reduce excessive dependence on resource 139 consumption in agricultural modernization (Varian et al.,2004;Sun&Li,2018). Also, the use of the 140 Internet has a significant positive effect on increasing the sales of agricultural products in the market, 141 increasing the sales prices of agricultural products, and improving the welfare of farmers 142 However, some researchers believe that Internet use has a significant negative impact on 152 welfare effects. The Internet use creates a "digital gap" between the information-rich and the 153 information-poor, leading to an increase in the income gap between individuals (Bonfadelli,2002； 154 Wouterlood,2012). The development of the Internet will only benefit those wealthy people. The 155 wealth accumulation of the rich will continue to be higher than that of the low-income people, use may also lead to Internet addiction users, addicted to the virtual world, lower social trust between 160 reality, which have a significant negative impact on the welfare effects(Abatini,2017). 161

Model setting 162
Assuming that the net income obtained by farmers using the Internet is * , the net income 163 obtained by not using the Internet is * , and the difference between the two (the difference between 164 the net income of users and non-users) is * . When * >0, farmers who use the Internet have higher 165 net income than non-users, and farmers choose to use the Internet. But * is a latent variable and 166 cannot be directly observed, so it is expressed as a function composed of observable variables, such 167 7 as the following latent variable model: 168 In the formula (1), represents the behavioral decision of whether to use the Internet, = 1 170 represents that the farmer uses the Internet, and = 0 represents that the farmer does not use the 171 Internet. Construct the impact of farmers' use of the Internet on farmers' welfare: 172 In the formula (2), represents the welfare effect of farmers, is the external environmental 174 characteristic variables such as personal characteristics, family characteristics, and village 175 characteristics that affect the use of the Internet by farmers. ¢ and ¢ are the coefficients to be 176 estimated, and is the random interference term. 177 The article uses the endogenous transformation model (ESR) proposed by Maddala (1983). Step 1: The selection equation of whether to use the Internet: 190 Step 2: The welfare level equation of farmers who use and not use the Internet: 192  Welfare expectations of farmers who do not use the Internet: 228 Welfare expectations of use group of farmers when they are not using the Internet: 230 Welfare expectations of non-use group farmers when using the Internet: 232 The average treatment effect of the welfare level of farmers who have used the Internet, that is, 234 the average treatment effect (ATT) of the treated group is expressed as the difference between 5a 235 and 5c: 236 Similarly, the average treatment effect of the welfare level of farmers who have not used the 238 Internet, that is, the average treatment effect on the untreated (ATU) of the control group, is 239 expressed as the difference between 5d and 5b: 240

Variable description 255
Internet use refers to using the latest information technology to make information exchange 256 10 between people more rapid and accurate and continuously promote information technology's rapid 257 development (Yan,2010). This article uses mobile phones and computers as Internet representatives. 258 However, rural households in Shouguang, Shandong, have a high adoption rate of mobile 259 communication, and most people use it for entertainment and communication. Therefore, to measure 260 the impact of Internet use on agricultural production, this paper selects whether to actively use 261 mobile phones and computers to query agricultural information as an indicator (Sheng et al.,2017).  Judging from household heads' characteristics in Table 1, the younger the age, the more 283 educated farmers are more likely to use the Internet. In terms of family characteristics, the more 284 muscular the rural households' economic strength, the higher the degree of specialized production, 285 the more government subsidies, and the more active they are to participate in rural professional 286 11 cooperatives, the more likely they are to use the Internet. The amount of financial loans of the user 287 group is much higher than that of the non-use group, indicating that the former may have a higher 288 degree of risk appetite. From the input perspective, the material input, labor input, and land input of 289 the non-use group are higher than those of the user group. In contrast, the mechanical input and 290 technical input are the opposite. From the perspective of village characteristics, farmers who are far 291 away from the trading market and have received information technology training are more inclined 292 to use the Internet (see Table 1). 293

Empirical result analysis 297
The Internet use decision-making results and the impact of Internet use on the three welfare 298 indicators are shown in Table 2 to Table 5 Table 2 to Table 5, and the resulting equation is in the 301 third and fourth columns of Table 2 to Table 5. 302  Note: ***, **, * indicate that the estimated results are significant at the statistical levels of 1%, 5%, and 10%, 304 respectively. The standard errors are in parentheses, and the same applies below.    From the third and fourth columns of Table 2, there is a significant positive correlation between 328 land input, material data input, labor input and yield per mu. Experience shows that: farmer's factor 329 input is the most effective and direct measure to increase agricultural production and income. 330

306
Although the marginal income will continue to decrease with the increase of factor input, it still has 331 a significant role in improving farmers' output in the short term. The larger the land for vegetable 332 cultivation, the more likely it is to enjoy the intensive advantages of agricultural specialization, 333 mechanization, and labor division, positively impacting agricultural output. Also, there is a 334 significant negative correlation between the proportion of non-agricultural income and vegetable 335 production. The higher the non-agricultural income of farmers, the lower their emphasis on 336 agricultural production, and the less willing to invest more time and financial resources in 337 agricultural production. It is worth noting that machinery input is positively correlated with farmers 338 in the user group but not related to farmers in the non-use group. It shows that compared with non-339 use group farmers, using group farmers' mechanization impacts farmers' vegetable output. 340

The impact of internet usage on net vegetable income 341
From the third and fourth columns of Table 3, it can be obtained that the proportion of non-342 agricultural income, material input and net vegetable income have a significant negative correlation. 343 It shows that the larger the proportion of farmers' non-agricultural income, the more material input, 344 the lower the net income of vegetables. Higher non-agricultural income represents an increase in 345 farmers' non-agricultural employment, which intensifies the transfer of high-quality labor from 346 agriculture to non-agricultural industries. It leads to weaker and weaker enthusiasm for vegetable 347 production and negatively affects net income. The more material input, the higher the production 348 cost and the lower the net income. Land input, labor input, and technology input significantly 349 correlate with farmers' net income. The increase in vegetable planting area will obtain the benefits 350 of large-scale operation and obtain more agricultural subsidies. The more labor input, the more likely 351 it is to realize intensive land cultivation and obtain more benefits. The investment in agricultural 352 technology can optimize agricultural products' quality, liberate labor productivity, increase 353 agricultural productivity, etc. Also, vegetable cultivation's length has a significant positive 354 correlation with farmers' net income in the user group. In contrast, the impact on farmers in the non-355 use group is not significant. It may be because the farmers can get more agricultural information by 356 using the group. The longer the vegetable planting period, the stronger the use group farmers' ability 357 to use various information. The easier it is to increase the farmers' income. 358

Impact of Internet use on per capita net income of rural households 359
From the third and fourth columns of Table 4. For the use group of farmers, the household's 360 per capita net income has a significant positive correlation with the length of education, the 361 proportion of non-agricultural income, land input, and machinery input, while a significant negative 362 correlation with the length of vegetable cultivation. Therefore, use group farmers mainly rely on 363 expanding the planting area, increasing agricultural mechanization to supplement the family income. 364 Internet users can break down "knowledge barriers," increase farmers' opportunities to acquire new 365 technologies and new knowledge and guide them to large-scale and mechanized production. 366 Simultaneously, Internet use reduces the cost of communication between farmers and the outside 367 world, making it easier to obtain market employment information, thereby increasing family income 368 from non-agricultural employment. For the non-use group, land input, government subsidies, 369 whether to join a cooperative, material data input and labor input are positively correlated with 370 household income. The non-use group farmers mainly rely on the traditional "extensive" economic 371 growth method of increasing material input, land, capital and other factors to increase their income. 372 Table 5 shows the average treatment effect of farmers' Internet use on the three welfare 374 indicators of vegetable yield per mu, net income and household per capita net income. Specifically: 375 when farmers who use the Internet do not use it, their per-mu yield, net income, and family per 376 capita net income will drop by 10.88%, 13.96% and 9.46%. When farmers who do not use the 377

Average treatment effect of farm household welfare indicators 373
Internet use it, the per-mu yield, net income, and per-capita net income of households will rise by 378 13.62%, 16.66% and 11.64%. So Internet use has a significant positive impact on improving the 379 farmers' fare of d the magnitude of the impact in net income, yield per mu, and net income per 380 household. 381 results' robustness. In order to avoid collinearity, the multicollinearity test was performed. The test 388 results were: the variance inflation factor (VIF) was less than 2, indicating no multicollinearity issue 389 among the variables. In order to reduce the influence of heteroscedasticity on the data, all dependent 390 variables are processed logarithmically. 391 SUR regression must satisfy the hypothesis that the equation's disturbance term is related to 392 the same period. Therefore, the three regression equations' disturbance terms need to be tested for 393 "no synchronization correlation": the chi-square value is 454.996, P=0.0000<0.001. Therefore, the 394 null hypothesis that the disturbance terms of each equation are mutually independent can be rejected. 395 The empirical results are shown in Table 6. 396 From the empirical results in Table 6, the use of the Internet by rural households has a 398 significant positive correlation with per-mu yield, net income, and household per capita net income. 399 That is, the use of the Internet by farmers can improve the welfare of farmers. In addition, it can be 400 concluded from the coefficient value that the greatest positive impact of the use of the Internet by 401 farmers is the net income of vegetables, followed by the yield per mu, and the least impact on the 402 net income per capita of the household. This conclusion supports the robustness of the above 403 empirical analysis. 404

Analysis of heterogeneity 405
The article divides the sample farmers into small-scale farmers group (a vegetable area less 406 than 5 acres) and large-scale farmers group (a vegetable area more than 5 acres (inclusive) according 407 to the area of vegetable production land. Calculate the treatment effects of these two groups of 408 farmers' use of the Internet on the per-mu vegetable output, net income, and per capita net income 409 of the family. 410 Table 7 shows that the treatment effect (ATT) of per-mu vegetable yield, net income, and per 411 capita net income of households using the Internet are 1.093, 1.458, and 1.005 for the small-scale 412