The Effect and Mechanism of Agricultural Informatization on Economic Development: Based on a Spatial Heterogeneity Perspective

Abstract

As the future direction of modern agriculture, agricultural informatization (AI) and its economic effects are worth exploring because they can well promote digital agriculture. Based on the existing research results, we propose six hypotheses around the economic benefits of AI and construct the geographically and temporally weighted regression (GTWR) model with a sample of 30 Chinese provinces over the period 2001–2020 for validation. Specifically, starting from farmers’ income and regional economic growth, the entropy value method is used to construct the AI indicators, and the GTWR model is constructed to analyze the effect of AI. Furthermore, the transmission mechanism of AI was explored from the perspective of agricultural industry structure upgrading. The following conclusions were drawn. First, the level of AI in China has increased significantly, and meanwhile its spatial correlation has also strengthened year by year. Second, AI demonstrates a positive correlation with farmers’ income growth and regional economic development, which means that it has become an important contributing factor of rural economic output. Third, agricultural industry structure upgrading is one of the important ways for AI to leverage its economic effect. Hence, improving the informatization level in the rural and agricultural sectors through multi-dimensionality is of positive and pragmatic significance for the rural economy.

1. Introduction

The information revolution, which began in the 1940s and 1950s, has intensified in recent years, driving profound changes in agriculture and rural areas and forming many new industries and new forms and models of business. Indeed, informatization has affected agriculture and many other industries [1]. Agricultural informatization (AI) refers to the development and application of modern information technology in agriculture in a comprehensive manner, so that it penetrates into the whole process of agricultural production, market, consumption and all specific aspects of rural society, economy and technology. Apart from providing services for agriculture, AI has also enhanced the value of agricultural information resources and has played an important role in narrowing the digital divide between urban and rural areas [2,3]. As a major component of modernization, exploring the economic effects of AI and its mechanism is highly important, which is especially true in China. The Chinese government has always put emphasis on AI. After the victory in poverty reduction, the focus of the “three-rural-issue” work has shifted to rural revitalization, which is also the general grasp of the “three-rural-issue” work in the new era [4]. Moreover, AI is an important direction for rural revitalization. The report of the 19th National Congress of China proposes to promote the simultaneous development of new industrialization, informatization, urbanization and agricultural modernization. The Strategic Plan for Rural Revitalization (2018–2022) clearly proposes to improve the level of AI and consolidate the infrastructure for rural informatization so as to implement the digital rural strategy. In addition, in the Outline of the Digital Countryside Development Strategy, it is clearly put forward that the digital countryside will be taken as an important aspect of constructing digital China, and that the development of informatization will be accelerated so as to drive the modernization of agriculture and rural areas.

As China is now implementing a digital countryside strategy and its modern agriculture is developing towards informationization, exploring the economic effects of AI can help promote the healthy development of digital agriculture. Informatization has become a new driving force for rural economic development. However, the actual impact of AI on agricultural economy is closely related to the conditions of the rural areas themselves, which manifest distinct spatial and temporal heterogeneity [5]. A scientific assessment of regional differences in AI and its impact on agricultural economy is important to promote the construction of a digital countryside.

Although informatization has been massively penetrating China’s agriculture as it develops, there is still a lack of a convincing and comprehensive review to assess the impact of AI on farmers’ income and regional economic development, especially in the context of the widely discussed “productivity paradox of information technology”. Therefore, this study will answer the realistic questions that plague the development of modern agriculture from two dimensions: the effect of increasing farmers’ income and the effect of enhancing regional economic growth. More specifically, this study is going to answer the following questions: Is AI spatially dependent? Does AI have positive effects on the economy? Are these effects on farmers’ income or on regional economic development? And in what ways do these effects work?

This study is innovative mainly in the following three aspects: (1) to address the lack of a comprehensive review of the economic effects of AI, we look into the effects on farmers’ income growth and regional economic development respectively, to provide reliable evidence for the development of AI in China; (2) the spatial and temporal heterogeneity of AI is included in the study’s scope to bring the research results more in line with the development status quo and to improve the scientificity of the estimation results of the model; (3) from the perspective of upgrading the agricultural industry structure, the mediating effect model is also applied to explore the mechanism of AI’s economic effects, which is conducive to a better understanding of the mechanism of AI.

2. Literature Review and Hypothesis

With the popularity of the Internet, exploring the relationship between the informatization level and the economy has gradually become one of the important topics of academic interest. Li found that the development and application of Internet-related technologies has increased labor productivity in manufacturing and agriculture [6]. Using the United States county-level data, Atasoy found that the informatization level is positively correlated with employment, and this relevance is stronger in rural and remote areas [7]. However, there have also been views on the “IT productivity paradox” in academia. For example, Jorgenson conducted a series of studies on economy and productivity growth in the United States since 1990, which led them to the conclusion that the rapid growth and prosperity of the United States economy was driven by the substitution of computer capital for non-computer capital, and to the acknowledgement of the paradox [8]. Some other developed countries also observed such a phenomenon in their informatization progress [9,10].

Subsequently, the development of AI started to gain attention from researchers. Grimes believes that ICT can help reduce the rural-urban divide, bringing both development and challenges to rural economies [11,12]. Whether rural areas can benefit from telecommunications technology is a question worth thinking about. Malecki found that only some rural areas gained economic development through AI, and that informatization is not a quick solution for rural development in the United States [13]. Some studies in China pointed out that AI will have a direct impact on agricultural production. For example, Zhang and Zhang analyzed China’s rural information and rural economic growth from 1993 to 2002 and concluded that the level of rural informatization directly affects rural economy and serves as a powerful drive for growth [14]. In addition, Zhao and Wen found a positive correlation between rural informatization index and agricultural economic growth in the Ningxia irrigation area, China, suggesting that rural informatization in Ningxia is gradually showing an economic promotion effect [15]. The classic literature on the role of information technology for economic development is summarized in

Table 1. Summary of existing literature studies.

Ref.	Country	Year	Result
[16]	22 OECD countries	2002–2007	There is a significant causal positive link between informatization and economic growth.
[17]	OECD	1996–2007	For every 10 percentage point increase in information technology infrastructure development, the per capita economic growth rate will increase by 0.9–1.5 percentage points
[18]	U.S.	1999–2006	There is a positive relationship between informatization and local economic growth.
[19]	201 countries	1988–2010	A 10 percentage point increase in information technology penetration rate raises real GDP per capita by 0.57 to 0.63 percentage points.
[20]	U.S.	2001–2010	High levels of information adoption are causally associated with higher incomes.
[21]	12 countries	2006–2014	Informatization has a significant positive effect on employment rates, thus increasing the average income of society as a whole.

.

The mechanism of AI is rarely reported in the literature. Nevertheless, Ye and Ren [22], and Lv and Chen [23] proposed that telecommunication technology has a significant structural adjustment effect. It accelerates the transfer of secondary industries to tertiary industries and greatly contributes to the upgrading and optimization of the industrial structure. Industrial structure upgrading is actually a selection process of production factors flowing from low-productivity sectors to high-productivity sectors, in which industries with higher productivity will further boost economic development after obtaining development priority [24,25]. Similarly, we wonder if agricultural industry structure upgrading plays an intermediary role for AI to exert its economic effects.

Based on the above findings, the following research hypotheses are proposed in this study and will be tested in the subsequent analysis.

Hypothesis 1 (H1):

China’s AI development has spatial relevance.

Hypothesis 2 (H2):

AI technology can exert a positive economic enhancement effect.

Hypothesis 3 (H3):

AI improves farmers’ income, that is, AI has the effect of increasing farmers’ income.

Hypothesis 4 (H4):

AI promotes regional economic development, that is, AI exerts a regional economic growth effect.

Hypothesis 5 (H5):

AI plays its effect of increasing farmers’ income through the upgrading factor of the agricultural industry structure.

Hypothesis 6 (H6):

AI exerts its regional economic growth effect through agricultural industrial structure upgrading factors.

3. Methodology, Variables and Data

3.1. Model Establishing

3.1.1. GTWR Model

The economic development in China’s provinces is not random, and the provinces independent from each other in space. Instead, the development shows an obvious spatial correlation and agglomeration. Since the data exhibit a certain spatial correlation and complexity in space, the interaction between the independent and dependent variables of the model may also differ in different regions when constructing regression models, so it is necessary to take into account the influence of spatial factors in the study.

Generally, studies that need to consider spatial factors mainly sample data by the geographic location of the sample, and the relationship between the independent and dependent variables in the study varies with the spatial location of the sample, i.e., spatial non-stationarity. Both the ordinary linear regression and the spatial panel data models are global regression analyses that explain only the relationships among factors that do not vary by location, so they cannot capture the spatial non-stationarity of the data. To effectively solve this problem, Fortheringham et al. came up with the concept of the geographically weighted regression (GWR) model [26]. Huang et al. (2010) introduced a time factor based on the GWR model and proposed a geographically and temporally weighted regression (GTWR) model that considers both temporal and spatial non-stationarity [27]. The GTWR model is widely used to measure the influence on explanatory variables. Compared with traditional statistical models such as the spatial econometric model, the GTWR model is obviously superior in reflecting spatial-temporal heterogeneity across regions. Specifically, the GTWR model can more directly demonstrate the geostatistical relationships between variables within each sample area, thus effectively reflecting the evolutionary relationships between variables in spatial-temporal scenarios [28,29,30]. Therefore, in this study, the GTWR model is adopted to explore the economic effects of AI in China (Equation (1)):

$$y_i={\beta }_0\left(u_i,v_i,t_i\right)+{{\displaystyle \sum }}_{k=1}^p{\beta }_k\left(u_i,v_i,t_i\right)x_{ik}+{\epsilon }_i i=1, 2, 3, \dots , \mathrm{n}$$

(1)

where $y_i$ is the dependent variable; $\left(u_i,v_i,t_i\right)$ is the spatial-temporal coordinates of the i-th sample point; ${\beta }_0\left(u_i,v_i,t_i\right)$ is the spatial-temporal intercept of the i-th sample point; P is the number of independent variables; ${\beta }_k\left(u_i,v_i,t_i\right)$ is the k-th regression parameter of the i-th sample; $x_{ik}$ is the value of the kth independent variable of the i-th sample point; and ${\epsilon }_i$ is the residual.

The spatial-temporal weight matrix is set as $\mathrm{W}\left(u_i,v_i,t_i\right)=\mathrm{diag}\left(w_{i1},w_{i2},\dots ,w_{in}\right),$ where the diagonal element $w_{ij}$ is obtained by the Gaussian function (Equation (2)):

$$W_{ij}=\exp (-{(\frac{d_{ij}^{ST}}{h})}^2)$$

(2)

where h is the bandwidth whose optimal value is determined by minimizing the AIC; $d_{ij}$ is the spatial-temporal distance between i and j; $\mathsf{\lambda }$ and $\mu$ are the spatial scale parameter and the temporal scale parameter, respectively. The spatial-temporal distance function is constructed (Equation (3)):

$$d_{ij}^{ST}=\sqrt{\lambda \left[{\left(u_i-u_j\right)}^2+{\left(v_i-v_j\right)}^2\right]+\mu {\left(t_i-t_j\right)}^2}$$

(3)

The above analysis shows that the GTWR model can not only embed the temporal and spatial characteristics of economic development data into the econometric regression model, but also analyze the non-stationarity of economic development in each province more objectively, thus better reflecting the real characteristics of the data and providing strong evidence for the effect of AI on economic development enhancement. Therefore, this study adopts the GTWR model to explore the economic effect of AI and to grasp its mechanism through the perspective of agricultural industry structure upgrading.

3.1.2. Mediating Effect Model

The mediating effect in agricultural industry structure upgrading is tested by examining regression coefficients in sequence. The following model is constructed.

$$y_{it}={\alpha }_1{x}_{it}+{\displaystyle \sum }_{k=1}^{p-1}{\theta }_k{x}_{it}^{con}+{\mu }_i+{\epsilon }_{it}$$

(4)

$$m_{it}={\alpha }_2{x}_{it}+{\displaystyle \sum }_{k=1}^{p-1}{\theta }_k{x}_{it}^{con}+{\mu }_i+{\epsilon }_{it}$$

(5)

$$y_{it}={\alpha }_3{x}_{it}+{\beta }_1{m}_{it}+{\displaystyle \sum }_{k=1}^{p-1}{\theta }_k{x}_{it}^{con}+{\mu }_i+{\epsilon }_{it}$$

(6)

where x_con is the controlled variable. The test of mediating effect follows the stepwise regression method: First, whether the estimated coefficients ${\alpha }_2$ and ${\beta }_1$ pass the significance test is tested, respectively. If they both pass the test, then whether ${\alpha }_3$ passes the significance test is tested further. If ${\alpha }_3$ also passes the test, the partial mediating effect is significant; and the opposite testing results indicate that the full mediating effect of economic development is significant. If at least one of the estimated coefficients ${\alpha }_2$ and ${\beta }_1$ is insignificant, a Sobel test will be performed. If the test result is significant, it indicates that the partial mediating effect of economic development is significant; otherwise, it indicates that the mediating effect is insignificant.

3.2. Variable Design

3.2.1. Explained Variables

This study focuses on the economic effects of AI in two dimensions, i.e., farmers’ income and regional economic development. The farmers’ income (FI) variable is expressed by the per capita disposable income of rural residents, and the regional economic development (RED) level is expressed by the per capita GDP. Meanwhile, in order to honestly reflect economic growth, this paper deflates prices of the above nominal economic variables using the 2001 rural consumer price index (CPI) as the base period [31].

3.2.2. Core Explanatory Variable

Based on the current situation of AI construction in China, and considering the availability of relevant data, this paper draws on the study of Li and Zhang to measure the level of AI in the following approach [32]. This paper believes that the construction of the indicator system of the level of AI should take into account three aspects, i.e., AI hardware facilities, AI service facilities and AI subjects. AI hardware facilities are the basis for the development of AI, and are the main way for AI subjects to obtain information. AI service facilities are another important aspect of AI infrastructure, and the external guarantee for the information subjects to access to information. As the users of agricultural information and the developers of AI, the subjects’ income and culture directly determine their demand for and ability to search and use agricultural information. From these three aspects, this paper constructed an AI readiness evaluation indicator system containing 3 primary indicators and 12 secondary indicators (

Table 2. Agricultural informatization evaluation indicator system and the weight of each indicator.

Primary Indicators	Weight of Primary Indicators	Secondary Indicators	Weight of Secondary Indicators
Hardware facilities	0.298	Color TV sets	0.184
		Cell phones	0.304
		Computers	0.512
Service facilities	0.449	Comprehensive population coverage rate of broadcasting programs	0.155
		The comprehensive TV programs coverage rate	0.132
		The household cable radio and TV coverage rate	0.192
		The proportion of administrative villages with postal service	0.139
		The proportion of administrative villages with Internet broadband service	0.181
Information subjects	0.253	Rural broadband users	0.201
		agricultural technicians in publicly owned enterprises and institutions	0.227
		The proportion of rural residents’ household labor force with college education or above	0.349
		The per capita net income of rural residents	0.424

).

Table 2 is the evaluation indicator system of China’s AI level. This system is divided into three dimensions (primary indicators), i.e., hardware facilities, service facilities and information subjects. Among them, hardware facilities (three secondary indicators) include color TV sets, cell phones and computers; service facilities (five secondary indicators) include the comprehensive population coverage rate of broadcasting programs, the comprehensive TV programs coverage rate, the household cable radio and TV coverage rate, the proportion of administrative villages with postal service and the proportion of administrative villages with Internet broadband service; information subjects (four secondary indicators) include rural broadband users, agricultural technicians in publicly owned enterprises and institutions, the proportion of rural residents’ household labor force with college education or above, and the per capita net income of rural residents.

Since this evaluation system contains multiple indicators and the importance of each indicator is different, i.e., since the weights are different, this paper adopts the entropy method for weight assignment. The entropy method is an objective assignment method, which determines the weight of an indicator by calculating its information entropy and then measuring the influence magnitude of its relative change on the system as a whole. The smaller the information entropy, the lower the disorder degree of the information, and thus the greater the utility value of the information, and the greater the weight of the indicator that should be assigned. The weights obtained by the entropy method are more objective and accurate than the subjective assignment method. Using the data in the period 2001–2020 in China, the weights of each dimension of the evaluation indicator system of the AI level constructed in this paper are shown in Table 1.

3.2.3. Mediating Variable

Through the analysis of previous literature, the indicator of agricultural industry structure upgrading is selected as the mediating variable, using ISU as abbreviation. The agricultural structure upgrading is measured by the agricultural structure optimization index, which is synthesized by the rationalization index and the advanced index of agricultural structure.

The rationalization of the industrial structure is a reflection of the structural coupling degree of production factors’ input and output and is usually measured by the degree of structural deviation. However, this indicator does not consider the importance of each industry in the economy, and calculating absolute values also brings inconveniences. Hence, the Thiel (TL) index is introduced in this study, as shown in Equation (7). The TL index takes into account the relative importance of industries and avoids the calculation of absolute values, thus retaining the theoretical basis and economic meaning of structural deviation. Therefore, it is a good indicator to measure the rationality of the industrial structure.

$$TL={\displaystyle \sum }_{i=1}^n\frac{Y_i}{Y}\ln \left(\raisebox{1ex}{$\frac{Y_i}{L_i}$}\!\left/ \!\raisebox{-1ex}{$\frac{Y}{L}$}\right.\right)$$

(7)

where $\frac{Y_i}{Y}$ is the output structure; and $\frac{Y}{L}$ is the production efficiency. If the agricultural economy is in equilibrium, the TL index is zero; and if the TL index is not zero, then the agricultural industrial structure deviates from equilibrium and rationality. Due to the special nature of agriculture, the number of employees in the planting, forestry, animal husbandry and fishery industries is not available at present, and thus $\frac{Y_i}{L_i}$ cannot be calculated. Given that $\frac{Y_i}{L_i}$ is used to represent the production efficiency of an industry, this paper uses the value added by per unit of intermediate consumption (i.e., value added/intermediate consumption) in planting, forestry, animal husbandry and fishery to measure the production efficiency of each industry, then substitutes $\frac{Y_i}{L_i}$ with the proportion of each industry output to total agricultural output, and finally calculates the TL index of agricultural industry, which also ranges from 0 to 1. The smaller the TL index, the smaller the difference in production efficiency, and vice versa.

Most of the established literature directly uses the TL index to measure agricultural industry structure upgrading [33]. The TL index takes the value of [0, 1], and a smaller index indicates a more reasonable agricultural industry structure; however, for the industrial structure advancement index H used below, a larger value indicates a more advanced agricultural industry structure; therefore, the TL index is in the opposite direction of H. Therefore, in order to synthesize the index of agricultural industry structure optimization, an improvement is made to the TL index so that it will fall in the same direction with the H index. To be specific, the TL is subtracted from 1, and then the value of the TL index would be in the same direction with the H index (the larger the value, the higher the degree of agricultural industry structure upgrading and optimization), then the obtained index can be used to measure the industry structure rationalization:

$$T{L}^{\prime }=1-TL$$

(8)

The industrial structure advancement includes the evolution of the proportional relationship and the improvement of labor productivity, which are the quantitative and qualitative connotations of the industrial structure advancement, respectively. We constructed an index of industrial structure advancement based on Liu et al.:

$$H=\sum v_{it}\times L{P}_{it}$$

(9)

where $v_{it}$ is the proportion of the output of Industry i at Time t; and $L{P}_{it}$ is the labor productivity of the Industry i at Time t. Similarly, the value added by per unit of intermediate consumption is used to express production efficiency, and $v_i$ is the proportion of industry output in total agricultural output. Since labor productivity has a scale but output weight has not, this paper adopts the “minimum-maximum standardization” method to standardize the labor productivity index. The equation is:

$$L{P}_{it}^N=\frac{L{P}_{it}-\min \left(L{P}_{it}\right)}{\max \left(L{P}_{it}\right)-\min \left(L{P}_{it}\right)}$$

(10)

Industrial structure advancement is calculated through Equation (11). The optimization of the agricultural industry structure includes two dimensions, i.e., rationalization and advancement. This paper synthesizes the optimization index of the agricultural industry structure, with each of the two dimensions occupying a weight of 50%.

$$ISU=0.5\times T{L}^{\prime }+0.5\times H$$

(11)

The index of upgrading and optimization of the agricultural industry structure is a comprehensive representation of the rationalization and advancement of the agricultural industry structure. The greater the value of the index of upgrading and optimization of the agricultural industry structure, the more advanced China’s agricultural industry structure is. Compared with agricultural industry growth, the index of agricultural industry structure upgrading does not indicate agriculture industrial improvement in quantitative terms, but interprets the rational and advanced changes of the agricultural industry structure from the perspective of industrial structure optimization. That is, whether the structure of agricultural industries is becoming more balanced, and whether the share of the dominant production sectors in the overall agricultural economy is gaining more advantage.

3.2.4. Control Variables

The proportion of investment in fixed assets in GDP (INV): As one of the three driving forces of economic growth, investment has a direct impact on economic development. Therefore, this paper sets INV as a controlled variable that affects economic growth.

The proportion of employment in total population (EP): Labor is an important factor affecting economic growth, and the actual number of people invested in the production process is generally used as a quantitative indicator to study the labor factor. This paper uses EP to characterize the actual use of labor resources over a certain period of time.

Human capital input (EE): The Solow growth model suggests that in addition to the absolute quantity of the labor force, human input should also include “the experience, knowledge, and skills that the labor force has after careful investment,” namely, the quality of the labor force. Drawing on the quality of labor force presented in previous studies, and considering the reliability and availability of data, selecting the proportion of education expenditure in residents’ disposable income can reflect the regional level of investment in human capital to a certain extent.

Proportion of industrial added value in GDP (IND): The growth rate of industrial added value reflects the fundamentals of a country’s economy, and the degree of industrialization has become an important factor in determining the speed of regional economic growth. This paper used IND as an indicator of the role of industrialization in economic growth.

3.3. Data

This paper conducted an empirical study using balanced panel data of 30 Chinese provinces, excluding Tibet, Hong Kong, Macau and Taiwan, with a sample size of 600 over the period 2001–2020. The original data are obtained from CEIC and the China Statistical Yearbook. In order to eliminate the problem of heteroskedasticity among the data, all variables have been logarithmicized, and the results of descriptive statistics of the variables are presented in

Table 3. Descriptive statistics of variables.

Variable	Obs	Mean	Std. Dev.	Min	Max
LnFI	600	8.788	0.725	7.277	10.461
LnRED	600	10.195	0.818	8.006	12.013
LnAI	600	7.509	0.699	6.045	9.111
lnIND	600	−1.063	0.281	−2.334	−0.556
lnEE	600	−0.694	0.272	−1.258	0.125
lnINV	600	−0.617	0.339	1.429	0.496
lnEP	600	−2.264	0.403	−3.281	−0.984
LnISU	600	−0.329	0.292	−2.681	−0.018

.

4. Results and Discussion

4.1. Spatial Agglomeration Characteristics of AI

Considering the spatial agglomeration of telecommunication technology, the spatial correlation analysis on the AI level of 30 Chinese provinces was conducted and the spatial Moran index is shown in

Table 4. Global Moran index of AI in China in the period 2001–2020.

Year	2001	2002	2003	2004	2005	2006	2007	2008	2009	2010
Moran’s I	0.105 *	0.106 **	0.114 *	0.114 *	0.114 *	0.118 **	0.118 **	0.121 **	0.122 **	0.125 **
Z value	1.796	1.811	1.909	1.906	1.918	1.965	1.963	2.010	2.044	2.053
Year	2011	2012	2013	2014	2015	2016	2017	2018	2019	2020
Moran’s I	0.125 **	0.130 **	0.135 **	0.135 **	0.138 **	0.138 **	0.138 **	0.140 **	0.143 **	0.143 **
Z value	2.107	2.167	2.176	2.189	2.226	2.231	2.246	2.291	2.309	2.401

Note: * indicates p < 0.1; ** indicates p < 0.5.

. China’s AI passes the spatial correlation test at the 10% level during 2001–2020. Moran’s I fluctuates but maintains an overall increasing trend, indicating that the spatial correlation of China’s AI strengthens year by year, and also that ignoring the spatial characteristics of AI would increase the error margin of model estimation.

ArcGIS software was used to spatially visualize the degree of AI in 30 Chinese provinces in the period 2001–2020 and to show the level of AI in 2001, 2006, 2011, 2016 and 2020 (Figure 1). During 2001–2020, the level of AI in 30 Chinese provinces has a significant increase, and the eastern coastal provinces have been in the lead. In addition, the level of AI in each province of China shows significant spatial agglomeration characteristics. Therefore, H1 is confirmed to be correct.

4.2. Economic Effect of AI

To ensure the robustness of the GTWR estimation results, OLS estimation was first performed, and the results are shown in

Table 5. OLS estimation results.

Variables	Model 1		Model 2
	LnFI		LnRED
	Coefficients	p-Value	Coefficients	p-Value
LnAI	1.029	0.000	0.964	0.000
LnIND	−0.015	0.000	0.045	0.097
LnEE	0.003	0.426	−0.143	0.000
LnINV	0.013	0.000	0.122	0.000
LnEP	0.015	0.000	0.422	0.000
C	1.093	0.000	3.939	0.000
R2	0.564		0.536
R2 Adjusted	0.561		0.533

, below. The degree of AI shows a significant positive correlation with farmers’ income and regional economic development. The estimation results of the OLS model indicate that H2 holds. Since OLS estimation does not consider spatial distance and the variability among observations, the results fail to reflect the spatial instability.

When using the GTWR model to analyze the economic effects of agricultural information, both the spatial correlation of informatization and the heterogeneity of individual cities are included in the research framework. The estimated results are listed in

Table 6. Estimated results of the economic effects of AI based on the GTWR model.

	Variables	Average	Min	25%	50%	75%	Max	IQR
LnFI Model3	LnAI	1.024	0.896	1.018	1.027	1.032	1.069	0.014
	lnIND	−0.010	−0.061	−0.024	−0.013	0.003	0.043	0.027
	lnEE	0.008	−0.059	−0.006	0.005	0.022	0.085	0.028
	lnINV	0.011	−0.039	−0.002	0.009	0.018	0.089	0.020
	lnEP	0.021	−0.071	0.007	0.019	0.032	0.143	0.025
	C	0.000	0.114	−0.004	0.000	0.004	0.028	0.008
	Bandwidth		0.115	Sigma		0.010	R2	0.999
	Residual Squares		0.055	AICc		−3660.86	R2 Adjusted	0.999
	Spatio-temporal Distance Ratio					0.542
	Variables	Average	Min	25%	50%	75%	Max	IQR
LnRED Model 4	LnAI	0.970	0.533	0.807	0.963	1.110	1.535	0.303
	lnIND	−0.043	−0.444	−0.176	−0.061	0.055	0.922	0.231
	lnEE	−0.320	−1.376	−0.515	−0.286	−0.082	0.200	0.433
	lnINV	0.073	−0.322	−0.025	0.071	0.168	0.477	0.193
	lnEP	0.400	−0.593	0.299	0.392	0.472	1.196	0.173
	C	0.011	−0.290	−0.028	0.008	0.053	0.220	0.081
	Bandwidth		0.115	Sigma		0.072	R2	0.992
	Residual Squares		3.138	AICc		−1224.61	R2 Adjusted	0.992
	Spatio-temporal Distance Ratio					0.642

Note: IQR indicates interquartile range.

. Compared with the OLS estimation results, the models estimated by GTWR all achieve a fitting degree of over 0.8, which is much higher than that of the model R² estimated by OLS. To further illustrate the applicability of the GTWR model, the estimated residuals were tested for spatial correlation. The estimated residuals of both Model 3 and Model 4 fail the spatial correlation test, which means that the omitted variables are in random distribution rather than spatially correlated, which increases the reliability of the GTWR estimation results.

From the above estimation results, it can be seen that AI exerts a positive effect on enhancing farmers’ income and regional economic development during 2001–2020. Hence, H3 and H4 are correct. In other words, AI enhances farmers’ income and regional economic development. By comparing the estimated coefficients of AI variables in Model 3 and Model 4, it can be seen that the effect of AI on farmers’ income is more pronounced. Moreover, its interquartile range is relatively small, indicating that this effect of increasing farmers’ income appears to be concentrated. Although AI also plays an incentive role for regional economic development, its estimated coefficient interquartile range is relatively large, indicating that the effect of this factor is more dispersed. Therefore, the economic effects of AI are analyzed from the time dimension and the regional dimension, respectively.

Spatial and temporal differences exist in the impact of AI on farmers’ income and regional economic development. In order to analyze the temporal characteristics of this impact, Figure 2 and Figure 3 show box plots of the changing trends of the estimated results of AI variables over time in Model 3 and Model 4, respectively. On the whole, the economic enhancement effect of AI varies widely from year to year due to influence factors such as labor force, investment and degree of industrialization in each province.

As illustrated in Figure 2 and Figure 3, the economic effects of AI fluctuate as time goes by. The average level of AI’s effect on increasing farmers’ income remains stable during the sample observation period. From 2001 to 2009, the positive contribution of AI to farmers’ income increased unevenly, and the mean value of the regression coefficient reached a maximum value of 1.031 in 2009. After 2009, the role of AI on farmers’ income decreased until the coefficient reached 1.017 in 2020. The relationship between AI and farmers’ income fluctuated significantly at the beginning of the study, and then the dispersion of the coefficients gradually decreased, indicating that the spatial variability of the impact of AI on farmers’ income decreased between 2001 and 2010. The coefficient dispersion showed a weak increase in the later part of the study.

In terms of the average effect, the positive contribution of AI to the regional economy shows a decreasing trend with the deep diffusion of telecommunication technology, i.e., the average level of this estimated coefficient gradually decreases from 2001 to 2020. The average enhancement coefficient of AI for regional economic development decreases from 1.215 to 0.779 in 2020. This indicates that the development of AI has had a diminishing effect on the average improvement of the national regional economy. In addition, the dispersion of the regression intensity coefficients of AI on regional economic development in each province gradually decreased from 2001 to 2012, and the dispersion of the coefficients gradually increased since 2013, indicating that the spatial variability of AI on regional economic development is more obvious. This indicates that although AI still shows a positive enhancing effect on regional economic development, the actual effect has a large gap due to spatial differences.

The concentration degrees of the estimated results of AI are different when these samples are under observation, which indicates that significant geographical differences exist in the promoting effect of AI on farmers’ income and regional economic development. In order to present directly the spatial and temporal differences in the impact of transportation on the economy of each province, the spatial heterogeneity of the estimated coefficients of AI variables was explored by combining spatial visualization (Figure 4 and Figure 5).

The effect of AI on increasing farmers’ income is stronger in developmentally backward regions than in more developed regions. For example, AI increases farmers’ income more notably in Heilongjiang, Jilin, Shaanxi, Gansu, Qinghai, Xinjiang and Yunnan than in provinces such as Liaoning, Inner Mongolia, Hebei, Beijing, Jiangsu, Anhui, Zhejiang and Shanghai. This means that for regions with a backward economic development, deepening AI is an important way to enhance farmers’ income and thus lift them out of poverty. In provinces such as Heilongjiang, Gansu and Guangdong, AI’ beneficial effect on regional economic growth is more prominent, followed by coastal provinces such as Shandong, Jiangsu, Zhejiang, Shanghai and Fujian.

Synthesizing the results of the above analysis, we believe that the findings are similar to those of Almalki et al. [34] and Nukala et al. [35], that is, information technology has led to an accelerated modernization process in the agricultural sector. IoT platforms help farmers to predict farm environmental data, and this information can effectively improve crop productivity and help to enhance farmers’ income and regional economic development.

4.3. Mediating Effect in Agricultural Industry Structure Upgrading

The estimation results of the GTWR model indicate that AI has a positive impact on both farmers’ income and regional economic development in China. So, what are the reasons for this phenomenon? In other words, what is the transmission mechanism by which AI exerts its economic effects? Based on the previous analysis, this section will investigate this transmission mechanism from the perspective of the upgrading path of the agricultural industry structure. To verify H5 and H6, a spatial mediating effect model was established with farmers’ income and regional economic development as the explained variables, AI as the core explanatory variable and agricultural industry structure upgrading as the mediating variable.

Table 7. AI and farmers’ income: transmission mechanism of agricultural industry structure upgrading.

Variables	Model 5.1	Model 5.2	Model 5.3
	LnFI	LnFI	LnISU
LnAI	1.034 ***	1.034 ***	0.045 **
	(684.73)	(680.41)	(2.25)
LnISU		0.005 ***
		(4.78)
lnIND	−0.006 **	−0.006 **	−0.238 ***
	(−3.62)	(−3.73)	(−4.50)
lnEE	0.002 *	0.002 *	0.034
	(2.98)	(2.99)	(0.65)
lnINV	0.012 ***	0.012 ***	−0.010
	(4.61)	(4.59)	(−0.27)
lnEP	−0.003 **	−0.003 **	0.160 ***
	(−3.75)	(−3.84)	(3.37)
C	1.021 ***	1.020 ***	−0.539 **
	(57.13)	(56.70)	(−2.28)
N	600	600	600
R2	0.998	0.997	0.308

Note: z statistics in parentheses; * indicates p < 0.1; ** indicates p < 0.5; *** indicates p < 0.01.

reports the estimation results of the model of farmers’ income growth effect of AI with agricultural industry structure upgrading as the mediating variable. In these results, the coefficient of the agricultural industry structure upgrading variable in Model 5.2 is significantly positive, which indicates that agricultural industry structure upgrading contributes the farmers’ income growth effect; the coefficient of the AI variable in Model 5.3 is significantly positive, which indicates that AI accelerates, to a considerable extent, agricultural industry structure upgrading. By comparing the estimation results of Models 5.1–5.3, it is found that the process of AI accelerates the upgrading of agricultural industrial structure, which then exerts a significant effect on the growth of farmers’ income, so this empirical result verifies H5.

Table 8. AI and regional economic development: transmission mechanism of agricultural industry structure upgrading.

Variables	Model 6.1	Model 6.2	Model 6.3
	LnRED	LnRED	LnISU
LnAI	1.075 ***	1.072 ***	0.045 **
	(93.26)	(94.63)	(2.25)
LnISU		0.112 ***
		(4.91)
lnIND	0.485 ***	0.518 ***	−0.238 ***
	(15.86)	(16.93)	(−4.50)
lnEE	−0.198 ***	−0.208 ***	0.034
	(−6.64)	(−7.09)	(0.65)
lnINV	0.114 ***	0.113 ***	−0.010
	(5.58)	(5.66)	(−0.27)
lnEP	0.229 ***	0.209 ***	0.160 ***
	(8.32)	(7.65)	(3.37)
C	3.093 ***	3.131 ***	−0.539 **
	(22.61)	(23.26)	(−2.28)
N	600	600	600
R2	0.816	0.801	0.308

Note: z statistics in parentheses; ** indicates p < 0.5; *** indicates p < 0.01.

reports the estimation results of the model of regional economic development effect of AI with agricultural industry structure upgrading as a mediating variable. In these results, the coefficient of agricultural industry structure upgrading variable in Model 6.2 is significantly positive, which indicates that agricultural industry structure upgrading greatly contributes to the local economy. A comparison of the estimation results of Models 6.1–6.3 reveals that AI relies on the industrial structure upgrading to play its role in enhancing regional economies, so H6 is correct.

In China, the emergence of information technology has blurred industrial boundaries, and there is even a situation where information technology dominates the development of the agricultural sector, which has promoted the upgrading of the agricultural industrial structure. AI is characterized by high growth, high efficiency and high value-added, which has a strong correlation with the upgrading of the industrial structure. Moreover, the advantages of a strong penetration and deep impact of AI can also strongly promote the upgrading of the industrial structure [36,37,38]. Similar to Yu [39], the same phenomenon of information technology accelerating industrial structure upgrading exists in China. Therefore, while AI plays a direct role in promoting farmers’ income and regional economic development, it will also play an indirect role by promoting the upgrading of the agricultural industry structure.

5. Conclusions

At present, China’s agriculture is transcending its traditional form towards a modern one. Information technology, which is scientific, fast-evolving, highly penetrative and influential, provides support for the development of modern agriculture. Exploring the economic effect of AI from the perspectives of enhancing farmers’ income and regional economic development is conducive to the sound development of digital agriculture. In this study, the entropy method is adopted to establish AI indicators. Considering the spatial and temporal heterogeneity of AI, the GTWR model is constructed to analyze the effect of AI on farmers’ income and regional economic development. Besides, the transmission mechanism of AI is explored from the perspective of agricultural industry structure upgrading. The estimation results of these empirical models confirm that all six hypotheses hold, and the following conclusions are drawn.

First, during 2001–2020, the level of AI in China increased significantly and showed typical spatial agglomeration characteristics, with spatial correlation strengthening year by year. Meanwhile, there is significant spatial heterogeneity in the impact of AI on economic development.

Second, AI has shown significant economic enhancement—specifically, AI has enhanced farmers’ income while also promoting regional economic growth. This suggests that Hypotheses 2, 3, and 4 are all valid.

Third, agricultural industry structure upgrading is one of the important ways for AI to play its economic role. Specifically, AI not only directly affects farmers’ income and regional economic development, but also indirectly promotes the upgrading of the agricultural industry structure. That is, H5 and H6 are valid.

6. Policy Implications and Outlook

6.1. Policy Implications

Along with the rapid development of China’s agriculture and rural economy, the demand for information in rural areas is growing stronger and stronger. Hence, increasing the informatization level in rural areas through various methods is of positive and pragmatic significance for rural economy. Based on the above findings, the following policy suggestions are proposed.

(1)

To improve the efficiency of investment and allocation of AI infrastructure. Demonstration bases of AI should be created. Governments at all levels and agricultural science and technology departments should act on the principle of “pilot, demonstrate and then promote” to build rural informatization demonstration bases, and create a hierarchy of informatization demonstration bases according to local conditions for the goal of sustainable development.

(2)

To train technical talents for AI. Farmers are the main force of agricultural production. We make efforts to provide basic skills training and fundamental education to ensure that farmers learn the knowledge of informatization, which then enables them to use the informatization platform to search for information related to the production and sale of agricultural products. Moreover, informatization can also be utilized to transfer the knowledge and skills of rural production and life, cultivate new farmers for a new era, provide more jobs, ensure that all the remaining rural laborers are fully employed and gradually improve farmers’ own informatization awareness and ability.

(3)

To strengthen the integration of information technology and regional special agriculture. It is suggested to (1) take advantage of local features and leverage information technology to pave the road for special AI; (2) use advanced information technology, such as 5G technology, cloud computing and big data, to build “smart agriculture” and achieve on-demand agricultural production; (3) develop the “Internet + rural E-commerce” mode to further expand the sales channels of agricultural products; (4) create an intelligent system for product quality supervision; and (5) connect “the last mile” in rural areas with the help of intelligent logistics technology, thus solving the difficulty in transporting agricultural products.

6.2. Research Outlook

Our study constructs AI indicators with China as an example, uses the GTWR model to verify the positive relationship between AI with farmers’ income and regional economic development, and further explores these mechanisms. This study provides new paths for emerging economies similar to China to explore rural economic development, while providing fresh and reliable evidence for the development of informatization. Not only that, this paper provides new ideas for other scholars’ research in the field of agricultural informatization development.