Digital Infrastructure and Grain Production:Empirical Evidence Based on Deep Learning
The availability of digital infrastructure in rural areas has led to a significant increase in farmers'access to information.As of December 2021,the number of Internet users in China had reached 1.032 billion,with an Internet penetration rate of 73.0％,among which the number of rural Internet users was 284 million,and the Internet penetration rate in rural areas was 57.6％,having important implications for farmers'livelihoods and agricultural production.From a theoretical level,the use of digital infrastructure may affect the agricultural production activities of rural households in two ways.The first is the information effect of the use of digital infrastructure.The Internet,as an important carrier of information dissemination in the digital age,can facilitate users'convenient access to relevant information and greatly reduce their information search costs.Therefore,the popularization and use of rural digital infrastructure can bring a large amount of nonfarm employment information to rural laborers and reduce information search costs in the labor market,thus facilitating nonfarm employment of rural residents.The second is the technological effect of digital infrastructure use.From the perspective of technological progress,the popularization of digital infrastructure has promoted the digital transformation of relevant enterprises and accelerated the process of agricultural digitization in China.The widespread use of digital infrastructure has also promoted the deep integration of emerging digital technologies with traditional agriculture,providing a guarantee for new business models such as"smart agriculture"and"e-commerce agriculture."In the era of big data,agricultural digitization to promote modern agriculture in China is key to improving the income of agricultural production and stimulating the expansion of agricultural production scale.Therefore,the overall impact of digital infrastructure use on agricultural production depends on the results of the combined effects of information and technology,and it is necessary to conduct rigorous empirical tests on the effects of rural digital infrastructure use.Using data from the 2009-2017 National Rural Fixed Observation Point Micro Household Survey,this study estimates the impact of digital infrastructure on rural households'total food production and sown area based on a deep neural network instrumental variable(DNN-Ⅳ)approach.The comparison of the two-stage least square method and DNN-Ⅳ reveals that the latter has better performance in prediction and identification,implying that DNN-Ⅳ can provide more accurate results in treatment effects estimation.Using this method,we found that digital infrastructure reduces households'total food production and total sown area.Further heterogeneity analysis found that the negative impact of digital infrastructure on rural households'agricultural production had a U-shaped relationship with the education level of the head of a household,while rural households with more arable land per capita in a household were more negatively affected by digital infrastructure in their food production.In addition,the mechanism analysis in this study reveals that digital infrastructure has increased the probability of rural households working outside the home,raised the size of land transfers,and improved the proportion of nonfood crops sown.Thus,we can conclude that digital infrastructure reduces households'food production activities mainly through intra-household labor transfer effects and nonfood crop substitution effects.As food security is the foundation of national security,this study argues that the negative impact of rural digital infrastructure penetration on the total amount of food production and sown area needs to be given great attention.However,this study does not consider the construction of rural digital infrastructure as detrimental.In the context of the digital transformation of agriculture,it is necessary to build digital infrastructure to ensure an increase in agricultural production in rural areas.Further,from a micro perspective,under the premise of treating rural laborers as rational individuals,agricultural production decisions depend on the returns from growing food.When the returns from engaging in agricultural production activities are higher than those from working outside,agricultural production activities would become the primary choice of households.Therefore,it is more important to encourage farmers to engage more in agricultural production activities and achieve high-quality development of agriculture through rational policy guidance in the context of high Internet penetration rates.The policy recommendations of this study include the following:increase agriculture subsidies to improve farmers'income from grain cultivation and stimulate household food production,establish a big data dynamic monitoring system to improve agriculture laws enforcement in rural areas,ensure that the size of cultivated land is maintained within a reasonable range,and promote the construction of new business models such as"smart agriculture"and"e-commerce agriculture"to ensure the efficiency of agricultural production.
Digital InfrastructureGrain ProductionDeep LearningAllocation EffectSubstitution Effect
NETL
NSTL
万方数据
