Digital Economy and Farmers’ Price Information Responsiveness

Abstract

Based on a survey of 1220 kiwifruit growers in Shaanxi and Sichuan provinces, China, this study employs the Lewbel instrumental variable approach with 2SLS regression to examine the impact of digital economy participation on farmers’ information capabilities. The results reveal the existence of a capability trap in digital development. Initial participation significantly improves information acquisition capacity, but deeper engagement impairs information judgment ability due to information overload. The study also finds unexpected effects of traditional advantage factors. Higher education weakens the positive effect of digital participation on information acquisition. Extensive social networks intensify the negative impact of digital use. Only income demonstrates a protective effect, mitigating the damage to judgment ability from digital engagement. Spatial analysis indicates no significant spillover effects from digital economy development. Importantly, only information judgment, not mere information acquisition, significantly improves agricultural income. These findings highlight the need for differentiated digital strategies that move beyond basic access promotion to enhance farmers’ information processing capabilities.

1. Introduction

In traditional agricultural economies, information asymmetry has long constrained farmers’ pricing decision-making abilities. This asymmetry makes it difficult for farmers to access real-time market prices, forcing them to rely on localized or outdated information for production decisions, thereby exacerbating market risks and income volatility [1]. Information scarcity further leads to supply–demand mismatches in agricultural markets [2]. Additionally, farmers’ limited information-processing capacity weakens their bargaining power. Due to fragmented operations and insufficient analytical capabilities regarding supply–demand dynamics and consumer preferences, farmers often find themselves at a disadvantage in transactions, allowing intermediaries to extract “information rents” [3]. The root causes of these information constraints lie in weak rural information infrastructure and high market information costs. However, the rapid development of the digital economy offers technological solutions to these challenges. According to the China Digital Economy Development Research Report (2024) [4] released by the China Academy of Information and Communications Technology, China’s digital economy has entered an accelerated growth phase, with its role in the national economy becoming increasingly prominent. The digital economy now accounts for 42.8% of GDP, an increase of 1.3 percentage points from the previous year, contributing 66.45% to GDP growth. Digital technologies are reshaping farmers’ information ecosystems, enabling them to access vast amounts of market data beyond spatial and temporal constraints. By the end of 2023, China had 192 million rural broadband users, with 5G networks covering 96% of townships (China Internet Society 53 times China Internet network development state statistic report [EB/OL]. Retrieved from http://www.cnnic.net.cn/. Accessed 6 March 2025). Digital financial platforms have expanded farmers’ access to economic information; e-commerce platforms have reduced information search costs; and social media has facilitated information sharing among farmers.

The rise of the digital economy has profoundly affected farmers’ price information responsiveness. On one hand, digital economy platforms have broken traditional spatiotemporal constraints by integrating massive information, significantly enriching farmers’ information sources and enhancing their information acquisition capabilities [5]. For instance, real-time market price data and targeted supply–demand trend analyses enable farmers to more accurately assess market dynamics, thereby optimizing sales decisions and improving income levels [1]. This precise information dissemination allows farmers to actively participate in market competition rather than relying on singular or delayed channels [6]. However, on the other hand, improved information accessibility may not necessarily translate to higher transaction prices for farmers [7,8,9]. Excessive information on digital economy platforms may also overwhelm farmers [10]. When confronted with complex data and diverse interpretations, farmers lacking sufficient information filtering skills and professional knowledge may experience information overload, potentially leading to misjudgments or decision-making delays [11,12]. This dual effect of digital economy information empowerment reflects both its transformative impact on agriculture and the existing contradictions between digital economic development and farmers’ information responsiveness.

The existing literature on the digital transformation of agriculture primarily evolves along two prominent research streams. The first stream extensively investigates the effects of the digital economy on the primary sector, particularly in agriculture. A consensus confirms that digitalization significantly enhances agricultural production and sales by improving market access and operational efficiency [5,13,14]. Concurrently, a second, closely related stream of research focuses on producer behavior, specifically their ability to analyze market signals and adapt their strategies accordingly. Scholars within this field have universally emphasized the critical role of information as a cornerstone for farmers’ production and pricing decisions [1,15,16]. While these two streams establish a foundational understanding, a critical gap persists at their intersection.

Existing studies have predominantly assumed a linear and positive relationship between digital access and improved market outcomes. However, they have largely overlooked a fundamental deconstruction of how digital tools influence farmers’ core competence: their responsiveness to price information. It remains theoretically and empirically unclear whether digital empowerment systematically enhances decision-making rationality or merely expands information quantity without a corresponding improvement in decision-making quality. Furthermore, the potential for a non-linear trajectory—where initial benefits may diminish or reverse with deeper engagement—and the moderating role of key farmer characteristics remain underexplored.

This study bridges these gaps by positioning its contribution within this theoretical framework. We move beyond the question of whether digitalization affects farmers to investigate how it reshapes their information capabilities. Using survey data from kiwifruit growers in major production regions, this research empirically addresses three pivotal questions: First, does digital economy participation enhance price responsiveness by improving judgment, or does it merely inflate information access? Second, how does this relationship evolve with the depth of digital participation, testing the notion of a ‘capability trap’? Third, how do traditional determinants like education, income, and social networks moderate these effects? By answering these questions, this study provides a more nuanced theoretical understanding of the mechanisms and boundaries of digital empowerment in smallholder agriculture.

Currently, China’s rural digital economy is still in its early stages. The promotion and application system of digital technology in smart agriculture has not yet been fully developed, resulting in insufficient coverage and depth of the entire agricultural industry chain. The widespread application of the Internet has fundamentally changed the social and economic system, profoundly impacting the organization of social and economic spaces. Therefore, the spatial spillover effects of the Internet economy cannot be ignored. Existing studies mainly discuss the spatial spillover effects of the digital economy at the macro level [17,18,19]. The development of the rural digital economy takes time to generate macro-level and broader spillover effects. In rural areas, small farmers act as economic decision-making units, both independent and closely connected. At the present stage, it is more appropriate to discuss the spillover effects of small farmers participating in the rural digital economy at the village and township level. Therefore, we use a spatial effects model to examine whether the participation of small farmers in the digital economy affects the price information responsiveness of other farmers in the same village or town.

This study selects kiwifruit growers as research subjects for two primary reasons: First, kiwifruit is a high-value cash crop characterized by significant price elasticity in trading markets and relatively competitive market conditions, where variations in farmers’ price responsiveness substantially influence their sales prices. Second, China is the world’s largest producer and consumer of kiwifruit, and the kiwifruit industry is undergoing a technological transformation. With the rapid development of the digital economy, digital technologies are being deeply integrated into every stage of kiwifruit production and sales. Therefore, exploring the impact of the digital economy on the price responsiveness of kiwifruit growers is highly representative. The mechanisms explored in this paper may not only apply to kiwifruit growers but could also extend to other similar cash crops, thereby providing an important reference framework for research in related fields.

This study makes four key contributions to the literature. First, it identifies and empirically validates the existence of a “capability trap” in agricultural digitalization, revealing the complex relationship between digital participation and information capabilities. While initial engagement enhances information acquisition, deeper involvement impairs judgment ability due to information overload, challenging the prevailing assumption in existing literature that the digital economy uniformly benefits farmers. Second, the research demonstrates the paradoxical effects of traditional advantage factors in digital contexts. The findings reveal that education weakens information acquisition benefits, social networks amplify negative impacts, while only income provides protective value. These findings necessitate a reconceptualization of human and social capital theories in digital agriculture research. Third, the study provides novel insights into spatial mechanisms by demonstrating the absence of digital economy spillover effects. Information acquisition exhibits spatial dependence, whereas judgment capability shows no such spatial correlation. This indicates that critical thinking skills resist spatial diffusion and must be cultivated individually. Finally, the research establishes that information judgment, rather than mere acquisition, drives agricultural income growth. This crucial distinction provides theoretical grounding for shifting policy focus from information access to cognitive skill development, offering valuable insights for developing countries pursuing digital agricultural transformation. The empirical evidence from China’s specialized kiwifruit growers presents a relevant case for understanding digital transformation in smallholder agriculture, particularly regarding how digital tools influence price responsiveness and decision-making quality in commercialized farming systems.

The remainder of this paper is organized as follows: Section 2 introduces the data sources and the current status description. Section 3 describes the main empirical methods used in this study. Section 4 presents the key empirical results and analysis. Section 5 summarizes the paper and provides policy recommendations.

2. Data Source and Status Description

2.1. Data

In November 2023, we conducted a household survey of kiwifruit growers in Shaanxi and Sichuan provinces, two leading production regions in China. The questionnaire was structured into five coherent sections to systematically capture key variables and their relationships: (1) farmers’ basic characteristics and household demographics; (2) input-output data related to kiwifruit cultivation; (3) participation in and intensity of engagement with the digital economy; (4) information acquisition and judgment capabilities regarding price dynamics; and (5) sales channel diversification and postharvest commercialization behavior. This design ensures that each module aligns with the study’s analytical objectives and allows for a logical flow from digital exposure to information capacity and eventual economic outcomes.

The selection of Shaanxi and Sichuan as study areas was based on two key considerations. First, Shaanxi and Sichuan are China’s primary kiwifruit production regions, ranking first and second in cultivation scale, respectively. They serve as exemplary models for the development of China’s kiwifruit industry. Second, as major production hubs, both provinces exhibit significant potential for digital transformation in the kiwifruit sector. Rapid advancements in information technology have positioned digitalization as a critical component of agricultural modernization in these regions.

The survey employed a combination of stratified and random sampling methods. Data were collected through face-to-face interviews between enumerators and household respondents, resulting in 1245 distributed questionnaires. After excluding invalid samples, 1220 valid responses were retained, yielding an effective response rate of 97.99%. Geographically, Shaanxi accounted for 60% (732 questionnaires) and Sichuan for 40% (488 questionnaires) of the sample.

2.2. Farmers’ Price Information Responsiveness

This study employs factor analysis to construct two latent variables: farmers’ price information acquisition capability and price information judgment capability. The mathematical model for factor analysis is expressed as:

$$X_{ij}={\lambda }_{j1}{F}_{i1}+{\lambda }_{j2}{F}_{i2}+\cdots +{\lambda }_{jm}{F}_{im}+{\epsilon }_{ij}$$

(1)

where $X_{ij}$ represents the standardized score of the $i$ farmer on the $j$ observed variable, ${\lambda }_{jm}$ denotes the loading coefficient of the $j$ variable on the $m$ common factor, $F_{im}$ indicates the factor score, and ${\epsilon }_{ij}$ is the random error term. The specific observed variables include: (1) Price information acquisition capability, measured by four items: communication frequency (1 = very rarely, 5 = very frequently), and the accuracy, timeliness, and convenience of price information acquisition (1 = strongly disagree, 5 = strongly agree); (2) Price information judgment capability, comprising four dimensions: ability to discern information authenticity and capacity to utilize information for decisions regarding selling price, timing, and channels (all measured on 1–5 scales).

To test the validity of the theoretical model, we first conducted a reliability analysis on the overall sample data. The reliability of the scales for each latent variable and the overall scale was assessed using Cronbach’s α. The Cronbach’s α values for the latent variables ranged between 0.7961 and 0.9077, all exceeding the minimum threshold of 0.7 (

Table 1. Indicators of farmers’ price information responsiveness and factor analysis test.

Variable	Indicator Name	Indicator Description	Std. Loading	Std. Dev	p-Value	Cronbach’s α	CR
farmers’ price information acquisition capability	Communication frequency	You often exchange price information with other farmers	0.544	NA	NA	0.8997	0.881
	Information accuracy	Be able to obtain price information accurately	0.832	0.086	0.000 ***	0.8007
	Information timeliness	Be able to obtain price information in a timely manner	0.888	0.093	0.000 ***	0.7961
	information convenience	It is very easy to obtain the price information	0.879	0.093	0.000 ***	0.7966
farmers’ price information judgment capability	The authenticity of the price information	Be able to accurately determine the authenticity of price information	0.703	NA	NA	0.9077	0.901
	Sales price decision	Be able to determine the appropriate selling price by using the obtained price information	0.902	0.044	0.000 ***	0.8460
	Sales time decision	Be able to use the obtained price information to determine the appropriate sales time	0.878	0043	0.000 ***	0.8573
	Sales channel decision	Be able to use the obtained price information to determine the appropriate sales channels	0.845	0.042	0.000 ***	0.8640

Notes: *** indicate significance at the 1%, levels, respectively.

). Additionally, composite reliability (CR) was employed to evaluate the reliability of the observed variables for their respective latent variables. The calculated CR values, derived from standardized factor loadings, were all significantly greater than the critical value of 0.7. This indicates good internal consistency among the observed variables for each latent construct, confirming the strong reliability and stability of the survey questionnaire.

In factor analysis, the KMO (Kaiser-Meyer-Olkin Measure of Sampling Adequacy) is an indicator used to assess whether data are suitable for factor analysis. It measures the partial correlations among variables to evaluate the adequacy of common factors shared between them. Factor analysis of farmers’ price information acquisition capability yielded a KMO value of 0.804 and Bartlett’s test statistic of 2682.569 (p = 0.000). For farmers’ price information judgment capability, the KMO was 0.833 with Bartlett’s test of 3161.728 (p = 0.000), indicating the sample data’s suitability for factor analysis. Farmers’ price information acquisition capability showed only one eigenvalue with a variance contribution rate of 71.63%, while farmers’ price information acquisition capability similarly exhibited one eigenvalue accounting for 76.91% of variance.

2.3. Participation of Farmers in the Digital Economy

Examining farmers’ engagement in digital economic activities at the micro level provides the most direct reflection of the current state of rural digital economic development. Drawing on the County-Level Digital Rural Index (2020) [20], this study explores the rural digital economy from a micro perspective. It defines farmers’ participation in the digital economy as their involvement in digitally enabled economic activities, encompassing the digitization of production, sales, supply, and financial services within rural industrial chains. This participation is characterized across four dimensions: digital production, digital sales, digital supply, and digital finance. Specifically, farmers engage with the digital economy primarily through: Digital production (e.g., adopting drones and other digital technologies to optimize planting processes); Digital sales (e.g., selling agricultural products via social platforms like WeChat, QQ, JD.com, and Taobao, or through livestreaming on platforms such as Kuaishou and Douyin); Digital supply (e.g., using smart logistics systems for precise transportation and distribution); Digital finance (e.g., utilizing third-party payment platforms like WeChat Pay and Alipay, or investing in financial products such as funds, stocks, and bonds via Yu’e Bao or online banking).

As noted earlier, the digital economy is categorized into these four components. A farmer is considered a participant if engaged in any component; otherwise, they are classified as non-participants. Statistical analysis of famers participation reveals stark disparities (

Table 2. Participation of farmers in the digital economy.

Participation Content	Number	Percentage	Degree of Participation in Digital Economy	Number	Percentage
Digital production	64	5.25%	0	265	21.72%
Digital sales	417	34.18%	1	502	41.15%
Digital supply	418	34.26%	2	130	10.66%
Digital finance	857	70.25%	3	298	24.43%
Digital economy	955	78.28%	4	25	2.05%

): only 5.25% adopted digital production, whereas 34.18% utilized digital sales, 34.26% engaged in digital supply, and 70.25% participated in digital finance. Overall, 78.28% of famers were involved in the digital economy. We measure the extent of farmers’ participation in the digital economy by the number of its four components they engage in. As shown in Column 6 of Table 2, 21.72% of farmers did not participate in any digital economy activities, 41.15% participated in one activity, 10.66% in two activities, 24.43% in three activities, and 2.05% in all four activities.

2.4. Selection of Control Variables

Based on existing literature and the need to account for potential confounding factors, we include the control variables summarized in

Table 3. Main variable definitions and descriptive statistics results.

Variable	Variable Definition	Mean	Std. Dev.
Participation in digital economy	Whether farmers participate in the digital economy: Yes = 1; No = 0	0.783	0.413
Degree of participation in digital economy	The number of digital economy projects participated by farmers: 0–4	1.439	1.137
Age	Age of head of household (years)	59.299	9.466
Education	Years of schooling for head of household (years)	7.669	3.288
Gender	Gender of head of household: male = 1; Female = 0	0.898	0.303
Health status	Perennial illness = 1; General = 2; Good = 3	2.635	0.600
Mobile phone contact	The number of contacts in the respondents’ mobile phone address book	131.175	185.417
Per capita household income	Per capita household income (Yuan/person)	34,609.842	54,244.333
Work outside	Is there anyone in the family working outside	0.774	0.419
Land circulation	Whether to transfer land: Yes = 1; No = 0	0.808	0.394
Cultivated land area	Family-run cultivated land area (mu)	7.499	11.506
Distance to main road	Distance between house and main road (miles)	2.081	3.784
Cooperative participation	Whether to participate in a cooperative: Yes = 1; No = 0	0.492	0.500
Number of cold stores	Number of village cold storage	4.714	10.013
The number of new agricultural business entities	Number of specialized cooperatives and family farms in villages (per unit)	2.174	2.952
Parcel hub	Whether the village has a Parcel hub: yes = 1; No = 0	0.765	0.424
E-commerce service station	Whether the village has an e-commerce service station: yes = 1; No = 0	0.437	0.496
Financial service point	Whether the village has a financial service point: yes = 1; No = 0	0.302	0.460
Distance to provincial capital	Spherical distance from sample village to provincial capital city (km)	80.361	17.842
Region	Sichuan = 1; Shaanxi = 0	0.400	0.490
N	1220

. These cover four categories: (1) Individual characteristics—age, education, gender, health status, and mobile contacts of the household head, which may influence cognitive capacity and digital adoption; (2) Family characteristics—per capita income, off-farm work, land transfer, cultivated land area, road accessibility, and cooperative membership, as these affect resource allocation and market engagement; (3) Regional characteristics—number of cold stores, agribusiness entities, parcel hubs, e-commerce and financial service points, and distance to the provincial capital, capturing local digital and market infrastructure; (4) A provincial dummy to control for broader regional heterogeneity. This selection aims to mitigate omitted variable bias and better isolate the net effect of digital economy participation.

3. Empirical Model

3.1. Benchmark Regression Model

To analyze the impact of digital economy on famers’ ability to obtain and evaluate price information, the following model was constructed:

$$info\_acquisitio{n}_i={\alpha }_1+c_{1}digita{l}_i+{\delta }_1{X}_i+{\epsilon }_{1i}$$

(2)

$$info\_judgmen{t}_i={\alpha }_2+c_{2}digita{l}_i+{\delta }_2{X}_i+{\epsilon }_{2i}$$

(3)

where $info\_acquisitio{n}_i$ denotes the farmers’ price information acquisition capability; $info\_judgmen{t}_i$ denotes farmers’ price information judgment capability; $digita{l}_i$ denotes participation in digital economy or the degree of participation in the digital economy; $X_i$ denotes the control variable; and $\epsilon$ denotes a random disturbance term.

Farmer engagement in the digital economy is influenced by various factors, including their price information responsiveness. Consequently, model specification may face potential issues related to selection bias and reverse causality. To address these concerns, we employed a composite instrumental variable (IV) that meets the exclusion restriction requirements. When no ideal exogenous instrumental variables are available, Lewbel [21] proposed a method to construct instruments using heteroscedasticity. The approach suggests that if the residuals from regressing endogenous variables on other exogenous variables in the model exhibit heteroscedasticity, the product of these residuals and the demeaned exogenous variables can serve as valid instruments. This method has been widely applied in empirical research.

Table 4. 2SLS regression result.

	(1)	(2)	(3)	(4)
	info_acquisition	info_judgment	info_acquisition	info_judgment
Participation in digital economy	0.296 ***	0.318 ***
	(0.089)	(0.095)
The degree of participation in the digital economy			−0.086	−0.194 **
			(0.084)	(0.084)
Age	0.005	0.002	0.001	−0.004
	(0.004)	(0.003)	(0.004)	(0.004)
Education	0.038 ***	0.008	0.047 ***	0.021 *
	(0.010)	(0.010)	(0.011)	(0.011)
Gender	0.046	0.117	0.054	0.125
	(0.115)	(0.106)	(0.120)	(0.122)
Health status	0.227 ***	0.204 ***	0.240 ***	0.218 ***
	(0.053)	(0.054)	(0.057)	(0.056)
Mobile phone contact	0.000 ***	0.000 *	0.001 ***	0.001 ***
	(0.000)	(0.000)	(0.000)	(0.000)
Per capita household income (After HIS conversion)	0.205 ***	0.230 ***	0.241 ***	0.290 ***
	(0.049)	(0.055)	(0.049)	(0.054)
Work outside	0.168 **	0.139 *	0.167 **	0.140
	(0.079)	(0.078)	(0.083)	(0.087)
Land circulation	0.030	0.018	0.019	0.005
	(0.096)	(0.099)	(0.096)	(0.100)
Cultivated land area	0.002	0.000	0.002	−0.000
	(0.002)	(0.003)	(0.002)	(0.003)
Distance to main road	−0.001	0.011	−0.000	0.012
	(0.006)	(0.008)	(0.006)	(0.009)
Cooperative participation	−0.001	0.035	0.027	0.085
	(0.067)	(0.074)	(0.068)	(0.083)
Number of cold stores	−0.000	−0.004	−0.000	−0.003
	(0.003)	(0.004)	(0.003)	(0.004)
The number of new agricultural business entities	−0.007	−0.004	−0.006	−0.002
	(0.008)	(0.007)	(0.007)	(0.008)
Parcel hub	0.063	0.030	0.061	0.026
	(0.096)	(0.105)	(0.098)	(0.109)
E-commerce service station	−0.054	−0.027	−0.044	−0.003
	(0.084)	(0.090)	(0.088)	(0.098)
Financial service point	−0.074	0.226 ***	−0.072	0.218 **
	(0.078)	(0.086)	(0.077)	(0.090)
Distance to provincial capital	0.005	0.000	0.003	−0.002
	(0.003)	(0.003)	(0.003)	(0.003)
Region	0.080	−0.057	−0.034	−0.269 **
	(0.120)	(0.098)	(0.123)	(0.110)
Kleibergen–Paap rk Wald F statistic	2718.594	2718.594	84.630	84.630
Breusch-Pagan test p-value	0.000	0.000	0.000	0.000
R2	0.1456	0.1379	0.1059	0.0459
F	11.72	8.95	10.36	9.20
N	1220	1220	1220	1220

Notes: ***, **, and * indicate significance at the 1%, 5%, and 10% levels, respectively. Standard errors are shown in parentheses.

reports the results of the 2SLS regression using instruments constructed by this approach. The Breusch-Pagan test rejects the null hypothesis of homoscedasticity, and the first-stage F-statistics of the 2SLS regression exceed the critical value, confirming the validity and relevance of the instruments.

3.2. Spatial Lag Model (Slm)

To further explore the spatial spillover effects of digital economy participation on farmers’ price information response capability, the following Spatial lag model is constructed:

$$y_i=\rho {\displaystyle \sum _{j=1}^n{w}_i}{y}_j+\beta x_i+{\mu }_i$$

(4)

We adopt the extended Queen spatial weight to set the spatial weight matrix. We construct two spatial weight matrices based on village and township codes. Specifically represented as:

$$W_{ij}=\left\{\begin{array}{l}1, \mathrm{Household}i\mathrm{and}j\mathrm{reside}\mathrm{in}\mathrm{the}\mathrm{same}\mathrm{area}.\hfill \\ 0, \mathrm{Household}i\mathrm{and}j\mathrm{reside}\mathrm{in}\mathrm{the}\mathrm{different}\mathrm{area}.\hfill \end{array}(i\ne j)\right.$$

(5)

$W_{ij}$ represents the element in the spatial weight matrix $W$, indicating the neighboring relationship between farmer $i$ and farmer $j$. When defining the village as the neighboring area, the matrix value for farmers from the same village is 1; otherwise, it is 0. When defining the township as the neighboring area, the matrix value for farmers from the same township is 1; otherwise, it is 0.

4. Results and Discussion

4.1. Benchmark Regression Results

The 2SLS results in Table 4 show that digital economy participation improves farmers’ information skills. This shift from non-use to use significantly boosts their ability to get and assess price information. This demonstrates a digital access threshold effect. Using digital tools helps farmers overcome information barriers. They can access market data at lower cost. Comparing multiple sources also improves their judgment. Columns 3 and 4 provide further insight. Deeper digital engagement does not significantly improve information acquisition. This suggests diminishing returns after initial adoption. More importantly, deeper use harms information judgment. This may be caused by information overload. Farmers’ existing knowledge is often insufficient to handle complex market data. Too much unguided digital exposure leads to confusion and cognitive bias. This ultimately reduces judgment quality. These findings suggest that expanding rural digital access is not enough. Training in digital skills is also essential. Such support can help farmers overcome cognitive barriers.

4.2. Robustness Test

4.2.1. Changing the Measurement Method of the Dependent Variable

To test the robustness of the previous estimation results, this study adopts the approach of replacing the dependent variable. Specifically, the coefficient of variation method is employed to recalculate farmers’ price information acquisition and price information judgment, followed by a 2SLS regression. The results are presented in

Table 5. Regression results with the dependent variable replaced.

	(1)	(2)	(3)	(4)
	info_acquisition	info_judgment	info_acquisition	info_judgment
Participation in digital economy	0.253 ***	0.279 ***
	(0.076)	(0.083)
The degree of participation in the digital economy			−0.072	−0.172 **
			(0.072)	(0.074)
N	1220	1220	1220	1220

Notes: *** and ** indicate significance at the 1% and 5% levels, respectively. Standard errors are shown in parentheses.

. As shown, digital economy participation positively affects both information acquisition and information judgment at the 1% significance level. However, the extent of digital economy only negatively influences farmers’ information judgment. These findings confirm the robustness of the earlier estimation results.

4.2.2. LIML Method

To ensure the robustness of the instrumental variable estimation results, this study further employs the LIML method for regression analysis. The LIML method is a type of instrumental variable estimation. It provides robust testing for endogeneity in models. When weak instruments or model specification uncertainties exist, LIML typically outperforms 2SLS. It exhibits better finite sample properties. Specifically, it reduces estimation bias. In over-identified cases, LIML is less sensitive to the exogeneity assumption of instruments. Therefore, it is commonly used to verify the reliability of estimation results. According to Stock and Yogo [22], the LIML estimator exhibits superior finite-sample properties under weak instrument conditions, whereas 2SLS may yield biased estimates due to insufficient instrument strength. By comparing coefficient differences between the two methods, we can examine the sensitivity of core variable estimation results. As shown in

Table 6. LIML model results.

	(1)	(2)	(3)	(4)
	info_acquisition	info_judgment	info_acquisition	info_judgment
Participation in digital economy	0.296 ***	0.318 ***
	(0.075)	(0.075)
The degree of participation in the digital economy			−0.086	−0.194 ***
			(0.070)	(0.073)
N	1220	1220	1220	1220

Notes: *** indicate significance at the 1% levels, respectively. Standard errors are shown in parentheses.

, the coefficients for digital economy engagement and its intensity maintain consistent signs and statistical significance across both LIML and 2SLS estimations. This indicates the robustness of our benchmark results to alternative estimation methods.

4.3. Heterogeneity Analysis

4.3.1. Education

Education level is a key factor influencing farmers’ capabilities, significantly shaping how the digital economy affects their acquisition and judgment of price information. Farmers with different education levels exhibit inherent disparities in traditional information channels and digital literacy: low-education farmers often rely more on the intuitive interfaces of digital tools to compensate for deficiencies in traditional channels, whereas high-education farmers, with their stronger information integration skills, may experience diminishing marginal benefits from digital technologies, or even face filtering burdens due to information overload [23]. Moreover, translating information into effective market decisions requires advanced skills heavily dependent on the cognitive abilities associated with education levels. Thus, our group analysis based on household heads’ education levels aims to clarify the extent and limitations of digital technology’s benefits for different populations, particularly whether it preferentially supports disadvantaged low-education groups and whether skill gaps exacerbate inequities in the distribution of digital dividends.

We constructed an interaction term between education and the digital economy as well as the degree of participation in the digital economy, and conducted a 2SLS regression. Based on the 2SLS regression results in

Table 7. Results of educational heterogeneity analysis.

	(1)	(2)	(3)	(4)
	info_acquisition	info_judgment	info_acquisition	info_judgment
Participation in digital economy	0.2655 ***	0.3094 ***
	(0.0777)	(0.0782)
Education	0.0359 ***	0.0074	0.0444	0.0214
	(0.0093)	(0.0094)	(0.0101)	(0.0106)
Participation in digital economy ∗ Education	−0.0556 ***	−0.0156
	(0.0212)	(0.0213)
Degree of participation in digital economy			−0.0848	−0.1947 ***
			(0.0701)	(0.0731)
Degree of participation in digital economy ∗ Education			−0.0200	0.0061
			(0.0216)	(0.0226)
Control variable	YES	YES	YES	YES
N	1220

Notes: *** indicate significance at the 1% levels, respectively. Standard errors are shown in parentheses.

, the interaction term between digital economy participation and education significantly negatively affects farmers’ information acquisition ability. However, it has no significant impact on information judgment ability. The interaction term between the extent of digital economy participation and education shows no effect on either ability. This may occur because highly educated farmers have already established stable information acquisition methods through traditional channels like school training or social networks. Consequently, this reduces the marginal benefits of digital economy participation. It might even cause negative effects due to information overload or selective neglect. Meanwhile, information judgment ability depends more on intrinsic cognitive factors such as experience or critical thinking. It is less influenced by external interactions. Thus, education cannot significantly improve this ability through digital economy participation. The absence of effect from the participation extent interaction term suggests education has a limited moderating role. This is because the participation extent itself captures key behavioral differences. Education does not further change its impact pathway.

4.3.2. Per Capita Household Income

The level of household income is crucial for farmers’ ability to effectively use digital technologies to obtain and process price information. It directly affects farmers’ basic capabilities, such as purchasing smartphones and paying for Internet access [24]. More importantly, the level of income also determines the types of information acquisition channels they can access, their cognitive ability to understand information, and the quality of their social relationships. High-income farmers usually have the financial means to use various digital channels, such as e-commerce platforms and social media, to obtain market information. In contrast, low-income farmers are restricted by their limited funds. They often have to rely on traditional methods to acquire information and may lack the ability to screen and process information. This “digital divide” caused by income disparities may lead to the benefits of the digital economy mainly flowing to high-income groups, exacerbating inequality. Therefore, we need to analyze the situations of farmers with different income levels to understand who the digital economy really helps and what the limiting conditions for its effectiveness are.

We established interaction terms between income and the digital economy as well as the degree of participation in the digital economy, and conducted 2SLS regression respectively. Based on the 2SLS regression results from

Table 8. Results of income heterogeneity analysis.

	(1)	(2)	(3)	(4)
	info_acquisition	info_judgment	info_acquisition	info_judgment
Participation in digital economy	0.3108 ***	0.3249 ***
	(0.0780)	(0.0783)
Per capita household income	0.2081 ***	0.2316 ***	0.2389 ***	0.3165 ***
	(0.0432)	(0.0434)	(0.0489)	(0.0520)
Participation in digital economy ∗ Per capita household income	0.0962	0.0464
	(0.0979)	(0.0984)
Degree of participation in digital economy			−0.0847	−0.2104 ***
			(0.0723)	(0.0770)
Degree of participation in digital economy ∗ Per capita household income			−0.0190	0.2450 *
			(0.1209)	(0.1286)
Control variable	YES	YES	YES	YES
N	1220

Notes: *** and * indicate significance at the 1% and 10% levels, respectively. Standard errors are shown in parentheses.

, the interaction term between the digital economy and per capita income shows no significant effect on farmers’ information acquisition or information judgment. However, the interaction term between the extent of digital economy participation and income has a significant positive impact on information judgment, but not on information acquisition. This may be because higher-income farmers have already reached saturation in information acquisition through basic digital access. Thus, the moderating role of income is limited. In contrast, greater digital economy participation, such as frequent use of digital tools, combined with income advantages, can enhance information screening and evaluation capabilities. Income improves farmers’ access to high-quality content, training opportunities, and cognitive resources. This ultimately strengthens information judgment. Meanwhile, information acquisition depends more on technological accessibility rather than deep participation driven by income.

4.3.3. Social Network

Studies show that social networks serve as key channels for rural information flow, facilitating farmers’ information access by lowering search costs, accelerating technology diffusion, and strengthening information credibility [25]. In regions with scarce formal information sources, long-term reciprocal exchange within kinship and local networks helps compensate for gaps in material and human capital. However, the structure of social networks may shape their role in digital empowerment. Farmers with broad social ties often obtain market intelligence through dense interpersonal channels, which may reduce the added value of digital tools. By contrast, those with narrower networks may rely more heavily on digital means to overcome information barriers. Moreover, the composition of mobile contact lists (a common proxy for social networks) can influence information redundancy and filtering costs, thereby shaping how effectively digital tools improve information acquisition. Examining these variations helps clarify the boundary conditions of digital empowerment and supports more targeted policy interventions.

To analyze how social networks shape digital outcomes, we introduced interaction terms between social networks (“number of mobile contacts”) and both the fact and the intensity of digital engagement, estimating the model via two-stage least squares (2SLS).

Table 9. Results of heterogeneity analysis of social networks.

	(1)	(2)	(3)	(4)
	info_acquisition	info_judgment	info_acquisition	info_judgment
Participation in digital economy	0.2871 ***	0.3539 ***
	0.0978)	(0.0982)
Mobile phone contact	0.0004	0.0002	0.0013 ***	0.0017 ***
	(0.0002)	(0.0002)	(0.0005)	(0.0005)
Participation in digital economy ∗ Mobile phone contact	−0.0001	0.0005
	(0.0008)	0.0008
Degree of participation in digital economy			−0.1066	−0.2231 ***
			(0.0745)	(0.0788)
Degree of participation in digital economy ∗ Mobile phone contact			−0.0008 *	−0.0011 **
			(0.0004)	(0.0005)
Control variable	YES	YES	YES	YES
N	1220

Notes: ***, **, and * indicate significance at the 1%, 5%, and 10% levels, respectively. Standard errors are shown in parentheses.

summarizes the results.

The interaction between basic digital engagement and social networks showed no significant effect on either information acquisition or judgment. In contrast, the interaction involving digital engagement intensity negatively affected both information capacities. The null result for basic engagement implies that social networks do not exert additional influence at general participation levels. The negative interaction with engagement intensity indicates a substitution effect. At high digital usage levels, traditional social ties may be crowded out or conflict with digital information sources, weakening information abilities. This aligns with the “optimal network hypothesis”: beyond a threshold, rising filtering costs due to excessive contacts can offset digital benefits, consistent with earlier evidence on the inverted U-shaped link between network size and information absorption [26].

4.4. Spatial Effects Analysis

Before estimating spatial econometric models, we tested key variables for spatial dependence. These tests determine whether variables show correlated patterns across geographical units, which would necessitate spatial modeling approaches. We employed Moran’s I statistic, with

Table 10. Spatial Correlation Test.

Matrix Type	Variable	Moran’s I	Z-Value	p-Value
Township proximity matrix W1	info_acquisition	0.012	2.708	0.007
	info_judgment	0.001	0.453	0.651
	Participation in digital economy	0.003	0.706	0.480
	Degree of participation in digital economy	−0.005	−0.912	0.362
Village proximity matrix W2	info_acquisition	0.021	2.494	0.013
	info_judgment	0.018	2.160	0.031
	Participation in digital economy	−0.004	−0.349	0.727
	Degree of participation in digital economy	−0.004	−0.407	0.684

displaying the results. When using a town-level adjacency matrix, information acquisition exhibited spatial correlation. With village-level weights, both information acquisition and judgment showed spatial dependence. This confirms spatial spillover effects for both variables, justifying the inclusion of spatial effects in our specification.

We subsequently conducted spatial model specification tests (

Table 11. LM test.

	Township Proximity Matrix W1				Village Proximity Matrix W2
	info_acquisition		info_judgment		info_acquisition		info_judgment
	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)
Participation in digital economy	YES		YES		YES		YES
Degree of participation in digital economy		YES		YES		YES		YES
Control variable	YES	YES	YES	YES	YES	YES	YES	YES
LM-lag	0.048	0.031	0.494	0.542	0.615	0.421	0.500	0.420
LM-error	0.100	0.064	0.443	0.466	0.815	0.476	0.566	0.492

Note: The p-values of the LM test are reported in the table.

). For both Model (1) and Model (2), LM-lag tests were significant under both weight matrices. Although the LM-error test was significant for Model (2), the robust LM-error version was not. Accordingly, we selected the spatial lag model (SLM) for both specifications. The SLM results (

Table 12. SLM results.

	Info_Acquisition
	Township as the Area
Participation in digital economy	0.3481
	(0.0707)
Degree of participation in digital economy		0.0949
		(0.0268)
Control variable	YES	YES
$\rho$	0.0058 ** (0.0025)	0.0064 ** (0.0025)
Control variable	YES	YES
N	1220

Notes: ** indicate significance at the 5% levels, respectively. Standard errors are shown in parentheses.

) reveal a significant positive spillover effect in farmers’ information acquisition at the township level. This indicates that farmers within the same township positively influence each other’s information gathering, likely through frequent exchanges, technology demonstrations, and collective learning activities.

Notably, neither digital economy engagement nor information judgment displayed spatial effects. This absence may stem from fragmented social networks or uneven digital access in rural areas, limiting cross-farmers skill imitation. Judgment ability appears more tied to individual factors like education and experience, which do not diffuse easily through spatial mechanisms.

4.5. Further Analysis: Farmers’ Price Information Responsiveness and Agricultural Income

After examining how the digital economy influences farmers’ price information capabilities, we explore whether these capabilities translate into higher agricultural income. Specifically, we assess the income effects of both information acquisition and judgment abilities. Using survey data from kiwi growers, we regress agricultural income on both capabilities while controlling for individual and household characteristics (excluding per capita income).

Table 13. Farmers’ price information responsiveness and Agricultural Income.

	Agricultural Income
	Coefficient	Std. Dev.
info_acquisition	1095.02	3146.771
info_judgment	6802.928 **	3098.17
Control variable	YES	YES

Notes: ** indicate significance at the 5% levels, respectively.

presents the results.

The findings show that information acquisition alone does not significantly increase agricultural income. In contrast, information judgment capability has a strong positive effect—farmers skilled in interpreting and applying price information earn substantially more. This indicates that while information access provides a foundation, the ability to process and use information drives real economic gains. However, since we previously found that deep digital engagement impairs judgment ability, a key challenge emerges.

These results imply that although digital tools create potential for income growth, realizing this potential requires enhancing farmers’ information analysis and decision-making skills. This underscores the need to complement digital infrastructure with targeted training in information literacy and practical application.

5. Conclusions

Our study investigated 1220 kiwifruit growers in Shaanxi and Sichuan provinces, China. The results indicated that participation in the digital economy significantly enhanced farmers’ information acquisition capacity, but deeper engagement impaired their information judgment ability. This finding reveals a critical issue: a “capability trap” exists in digital economy development. While initial digital participation improved farmers’ information capacity, excessive use damaged information judgment, primarily due to information overload. This pattern remained consistent across both regions despite their differing development levels, suggesting the phenomenon may extend beyond local contexts.

Rural China has achieved rapid digital infrastructure coverage; however, significant quality disparities exist across digital platforms. Although broadband and mobile networks are now widely available, specialized platforms for agricultural production face multiple challenges—such as outdated information, homogeneous services, and poorly designed user interfaces. This mismatch between infrastructure and platform quality limited the effectiveness of digital tools and negatively affected farmers’ user experience. Particularly in specialty crop growing areas, existing platforms often failed to deliver timely market information and lacked professional decision-support functions. Consequently, hardware investments in digital infrastructure did not fully improve farmers’ information capabilities. This reality highlights an urgent need in rural digitalization: while network coverage remains important, equal attention must be given to platform service quality as critical soft power.

The study also found that traditional advantage factors—education, income, and social networks—functioned differently in digital contexts. Higher education weakened the positive effect of digital participation on information acquisition. Extensive social networks intensified the negative impact of digital use on information judgment. Only income demonstrated a protective effect, mitigating the damage of digital engagement to judgment ability.

Spatial effect analysis showed no significant spatial spillover effects from the digital economy. Information acquisition capacity exhibited positive spatial dependence among farmers, whereas information judgment ability showed no such spatial correlation. Further analysis confirmed that information acquisition alone did not increase agricultural income, while information judgment significantly improved it.

This study makes several key contributions. First, it identifies a capability trap where digital participation initially improves information acquisition but later impairs judgment due to information overload. Second, it shows that traditional advantages function paradoxically in digital contexts—education weakens benefits, social networks amplify negatives, while only income provides protection. Third, it demonstrates the absence of digital economy spillovers and confirms that judgment ability, not mere acquisition, drives income growth.

The study acknowledges several limitations and proposes future research directions. Constrained by data availability, the analysis relied solely on cross-sectional data. Future studies should employ longitudinal data to examine the long-term effects of digital economy participation on smallholders’ price information responsiveness. In addition, while this study focused on Chinese kiwifruit growers, the generalizability of findings to other crops remains uncertain. Future research should expand to diverse agricultural products. Kiwifruit represents a typical high-value economic crop—nutritious, high-priced, and requiring trellis cultivation with large initial investment and a long growth cycle. It also has strong storage and transport properties, supporting high commercialization. Thus, these findings may be more applicable to high-value cash crops similar to kiwifruit. Nevertheless, they offer valuable insights for China and similar contexts to safeguard smallholder interests and promote equitable digital development.

Policymakers should develop targeted responses based on these specific findings. Since initial digital participation improves information skills but deeper engagement impairs judgment, specialized information filtering tools should be created for high-engagement farmers, and decision support systems should be integrated into village-level digital platforms. Given that higher education weakens information acquisition effects, differentiated training programs emphasizing information verification and field application should be designed for this group. Considering that income demonstrates a protective effect on information judgment, low-income farmers should receive subsidies for digital device purchases, and a dedicated risk mitigation fund should be established. To address the finding that social networks amplify negative impacts, information quality control measures should be implemented within villager social groups, such as regularly publishing verified market analysis reports to help curb misinformation spread. Given the lack of spatial spillover effects in the digital economy, information sharing groups should be organized by natural village, with farmers who have strong information analysis capabilities serving as coordinators to systematically share decision methods and experience. Finally, since information judgment capability—rather than information access quantity—directly drives agricultural income, training should shift its focus from simple search techniques to real-scenario-based judgment training. Simulated decision exercises can help farmers effectively transform information resources into operational profits.