The Impact of Information Acquisition on Farmers’ Drought Responses: Evidence from China

Abstract

Climate change presents major challenges to agriculture, especially in economically underdeveloped regions. In these areas, farmers often lack access to resources and timely information, which limits their ability to respond effectively to drought and threatens agricultural sustainability. This study uses survey data from farmers in underdeveloped regions of China to examine the association between their ability to acquire information and their drought response behaviors. The results indicate that better information acquisition ability is significantly correlated with more effective and scientifically informed decision-making in drought adaptation strategies. To explore the underlying mechanism, we introduce value perception—that is, farmers’ beliefs about the usefulness and benefits of drought adaptation strategies—as a mediating variable. A mechanism model is constructed to test how information acquisition ability relates to behavior indirectly through this perception. We apply a threshold regression model to identify potential nonlinear associations, finding that the relationship between information acquisition ability and drought response behaviors becomes stronger once a certain threshold is surpassed. Additionally, we employ the Item Response Theory (IRT) model to measure the intensity and quality of farmers’ adaptation behaviors more accurately. These findings provide theoretical insights and empirical evidence for enhancing agricultural resilience, while acknowledging that causality cannot be definitively established due to the cross-sectional nature of the data. The study also offers useful guidance for policymakers seeking to strengthen farmers’ access to information, improve value recognition of adaptive actions, and promote sustainable agricultural development in underdeveloped areas.

1. Introduction

Agriculture forms the backbone of national economies and serves as a critical sector in addressing climate change [1,2]. However, the escalating threats posed by climate change have increasingly endangered agricultural production [3]. The prevalence of extreme climate events jeopardizes global agricultural productivity and food supply chain stability, with these challenges being particularly acute in economically underdeveloped regions [4,5,6]. For example, the drought in Andalusia from 2005 to 2008 resulted in economic losses exceeding EUR 151 million, adversely affecting both farmers and consumers, thereby impeding the sustainable development of the agricultural economy [7].

Enhancing farmers’ capacities to implement scientifically informed adaptation measures is essential for effectively mitigating the impacts of extreme climate events [8]. Existing research indicates that information acquisition is a fundamental prerequisite for farmers to adopt adaptive behaviors. The ability to acquire information significantly improves farmers’ cognitive capacities, thereby optimizing their decision-making processes [9]. However, in economically underdeveloped regions, delays and inadequacies in information dissemination are common, limiting farmers’ timely adoption of effective measures and exacerbating agricultural production risks and poverty recurrence [10,11].

Previous studies on farmers’ adaptive behaviors to extreme climate events have predominantly focused on analyzing the effects of farmers’ characteristics (e.g., gender, education level, resource endowment, and labor availability) and external support conditions (e.g., information access, agricultural social services, and technology promotion) [12,13]. These studies have demonstrated that factors such as gender, landholding size, annual income, and household labor availability profoundly influence farmers’ adaptive responses to extreme climate events [14]. Furthermore, increased access to information channels and agricultural extension services has been shown to effectively promote positive adaptive behaviors among farmers [15,16].

Nonetheless, several research gaps remain. Existing studies primarily emphasize individual characteristics or household economic conditions as determinants of adaptive behaviors [17], while paying insufficient attention to the role of farmers’ information acquisition ability—a key cognitive factor—in shaping these behaviors. Moreover, current research often relies on simplified quantitative models to analyze the impact of climate service information, lacking a theoretical framework that reveals the internal mechanisms through which information acquisition ability influences behavioral decisions [18]. In addition, most existing studies focus on major agricultural production zones or impoverished regions globally [19,20], while neglecting farmers in economically underdeveloped areas of China, thus limiting the policy relevance of the findings in these local contexts.

To bridge these gaps, this study focuses on farmers in five southwestern provinces of China, representative of economically underdeveloped regions, using field survey data to examine the impact of information acquisition ability on drought response behaviors. By introducing value perception as a mediating variable, this research constructs a theoretical mechanism model to analyze the internal transmission paths through which information acquisition affects behavior. Furthermore, a threshold model is employed to explore the nonlinear characteristics and critical points of this influence. This integrated approach strengthens the linkage between existing theoretical gaps and the research objectives, aiming to reveal the cognitive pathways underlying farmers’ climate adaptation decisions and provide scientific support for formulating precise and resilient adaptation policies.

This study makes the following contributions: (1) it enriches the theoretical and empirical understanding of how information acquisition ability affects farmers’ adaptive behaviors in the context of extreme climate events; (2) it innovatively introduces value perception as a mediating variable to explore internal behavioral mechanisms; (3) it employs threshold effect analysis to identify the critical influence levels of information acquisition; and (4) it systematically investigates farmers in underdeveloped regions, addressing regional limitations in the existing literature.

The remainder of this paper is structured as follows: Section 2 reviews the theoretical framework and research hypotheses. Section 3 introduces the data sources, variable definitions, and methodology. Section 4 presents the empirical results and discussion. Section 5 concludes with policy implications and future research directions.

2. Materials and Methods

2.1. Data Sources

The data utilized in this study were obtained from a survey conducted by the research team from March to October 2024 in five provinces/regions in Southwest China: Guizhou, Yunnan, Sichuan, Guangxi, and Chongqing. These regions were selected for the following reasons: This area represents a key battleground for consolidating the achievements of China’s poverty alleviation campaign. It is characterized by underdeveloped economic conditions, with rural livelihoods heavily dependent on agricultural production. Frequent extreme climate events have posed significant agricultural loss risks to farmers, potentially leading to large-scale poverty relapse [21,22]. Southwest China experiences frequent and severe droughts due to the combined effects of global warming and unique geographical conditions, presenting critical challenges to income stability among farmers [23]. Therefore, the analysis of farmers’ drought responses in this region provides a representative and valuable case for understanding the behavioral characteristics and mechanisms of farmers in economically underdeveloped areas.

The field survey employed a combination of probability-proportional-to-size sampling and multi-stage random sampling, conducted in three stages: In the first two stages, two townships were selected from each of the five sample regions: Qujing (Yunnan), Qiandongnan (Guizhou), Dazhou (Sichuan), Nanning (Guangxi), and Hechuan (Chongqing). Within each township, 6 to 10 villages were chosen. In the third stage, at least 25 households were randomly selected in each village for face-to-face questionnaire interviews. A total of 2475 households were surveyed. After eliminating invalid responses, 2306 valid questionnaires were retained, with an effective response rate of 93.17%.

2.2. Variable Selection

(1)

Dependent Variable: Farmers’ drought response behaviors

Drawing on the existing literature [24,25] and field-specific circumstances, farmers’ drought responses were categorized into three types:

Capital-oriented behaviors: These include increasing agricultural investment, applying for agricultural credit, purchasing agricultural insurance, or engaging in non-agricultural activities (e.g., secondary or tertiary industries or off-farm work).

Labor-oriented behaviors: These involve adjusting cropping schedules, enhancing irrigation facilities or frequency, intensifying the use of fertilizers and pesticides, or applying mulch and crop residue for soil conservation.

Technology-oriented behaviors: These include selecting drought-resistant crops, adopting water-saving irrigation technologies, implementing pest and disease control techniques, or using scientific fertilization methods.

Households were assigned a value of 0 if none of these responses were adopted and 3 if all three types of responses were implemented.

(2)

Independent Variable: Information acquisition ability

This variable reflects farmers’ information acquisition ability, which differs conceptually from mere information access. While information access typically refers to the objective availability or proximity of information sources (e.g., whether farmers have internet access or receive government bulletins), acquisition ability emphasizes the individual’s capacity to actively seek, understand, and utilize information from available channels. It reflects cognitive and behavioral competence in processing complex agricultural or climate-related content and making informed decisions.

To quantify this ability, we draw on previous studies [26,27] and use the Item Response Theory (IRT) model, which incorporates both difficulty and discrimination parameters of various information channels. Information sources were classified into four categories: online media, interpersonal communication, government departments, and agricultural training programs. Farmers’ responses to whether and how frequently they used these channels were used to estimate a latent trait score that represents their information acquisition ability, capturing both the diversity and depth of their engagement with drought-related information.

(3)

Mediating Variable: Value perception

Farmers’ perceived economic value of adopting drought responses influences their behavioral decisions, as rational economic agents aim to minimize economic losses [28]. The study quantified farmers’ value perceptions by assessing their agreement with the statement “drought responses can stabilize income,” using a five-point Likert scale.

(4)

Threshold Variable: Information acquisition ability

The impact of information acquisition ability on farmers’ decision-making behavior exhibits dynamic characteristics [29]. By selecting the level of farmers’ information acquisition ability as a threshold variable, this study aims to explore its phased effects on farmers’ drought response behaviors at different levels.

(5)

Control Variables: Individual and household characteristics

In this study, farmers’ individual characteristics and household characteristics are incorporated as control variables in the regression analysis. Individual characteristics include gender, age, education level, and health status, while household characteristics cover cultivated land area, internet connectivity, proportion of agricultural income, participation in land transfer, household labor force size, and frequency of natural disasters experienced in the past five years.

2.3. Descriptive Analysis

Descriptive statistics were conducted using SPSS 26.0 to summarize sample characteristics and variable distributions. As shown in

Table 1. Descriptive statistics of research variables.

Variable Category	Variable Name	Definition and Assignment	Mean	Standard Deviation
Dependent Variables	Capital-oriented behavior	Do you address drought by increasing agricultural investment, applying for agricultural credit, purchasing agricultural insurance, or engaging in non-agricultural activities (e.g., working in secondary/tertiary industries)? Yes = 1, No = 0	0.243	0.036
	Labor-oriented behavior	Do you address drought by adjusting crop planting schedules, increasing irrigation frequency/facilities, intensifying fertilizer/pesticide use, or using mulch/straw coverage? Yes = 1, No = 0	0.735	0.132
	Technology-oriented behavior	Do you address drought by choosing drought-resistant crops, adopting water-saving irrigation, pest control techniques, or scientific fertilization methods? Yes = 1, No = 0	0.307	0.461
Independent Variables	Information acquisition channels	Do you obtain drought response information through online media channels? Yes = 1, No = 0	0.624	0.398
		Do you obtain drought response information through interpersonal communication channels? Yes = 1, No = 0	0.631	0.854
		Do you obtain drought response information through agricultural training channels? Yes = 1, No = 0	0.186	0.073
		Do you obtain drought response information through government departments? Yes = 1, No = 0	0.435	0.448
Mediating Variable	Value perception	Do you agree that drought response can stabilize income and reduce economic risks? Strongly disagree ~ Strongly agree (1 ~ 5)	3.141	1.682
Control Variables	Individual characteristics	Gender: Male = 1, Female = 0	0.653	0.821
		Age: Actual age of the respondent	51.54	11.69
		Education level: Illiterate = 1, Primary school = 2, Junior high school = 3, High school = 4, College (or above) = 5	3.015	0.917
		Health status: Poor = 1, Average = 2, Good = 3	2.037	1.536
	Household characteristics	Internet connectivity: Yes = 1, No = 0	0.731	0.048
		Cultivated land area (ha)	0.250	2.224
		Proportion of agricultural income: 0~50% = 1, 50%~90% = 2, 90%~100% = 3	1.768	0.885
		Participation in land transfer: No transfer = 1, Land transfer in = 2, Land transfer out = 3	1.593	1.028
		Household labor force: 1–2 people = 1, 3–4 people = 2, 5 or more people = 3	1.976	1.005
		Number of natural disasters in the past five years: 1 or less = 1, 2–4 = 2, 5 or more = 3	2.368	1.558

, labor-oriented behaviors to drought are the most commonly adopted by farmers (73.5%), followed by technology-oriented behaviors (30.7%), while capital-oriented behaviors are the least frequent (24.3%). Farmers primarily acquire information through interpersonal communication channels (63.1%), online media channels (62.4%), and government departments (43.5%), with fewer obtaining information through agricultural technical training channels (18.6%). Farmers exhibit relatively high value perceptions of drought response behaviors, scoring an average of 3.141 out of 5. At the individual level, the surveyed sample comprises a higher proportion of male farmers (65.3%). The average age of the respondents is 51.54 years, with an average education level equivalent to junior high school. Most farmers rated their health status as average. Regarding household characteristics, the average cultivated land area is 0.25 ha, and the majority of households derive more than half of their income from agriculture. Participation in land transfer is relatively low, with an average household labor force size of at least two members. On average, households have experienced at least one natural disaster over the past five years.

2.4. Research Methods

(1)

Information Acquisition Ability Assessment Model

Following the approach of Abdul-Salam and Phimister [30], this study evaluates farmers’ information acquisition ability based on their use of various combinations of information channels. An Item Response Theory (IRT) model is employed, represented as follows:

$$P_{\mathrm{ij}}={\displaystyle \frac{\exp [r_j(a_i-b_j)]}{1+\exp [r_j(a_i-b_j)]}}$$

(1)

where P_ij denotes the probability that farmer i acquires drought response information through channel j. The primary information acquisition channels include online media, interpersonal communication, agricultural training, and governmental departments. ai represents the information acquisition ability of the farmer. r_j denotes the discrimination parameter of the selected information channel, and b_j reflects the difficulty parameter of the selected information channel.

The selected items (i.e., information channels) include online media platforms (e.g., WeChat, agricultural apps), interpersonal communication (e.g., neighbors, relatives), agricultural training sessions (e.g., from cooperatives or extension agents), and official information from governmental departments (e.g., village cadres, public broadcasts). A binary indicator is assigned to each channel based on whether it is used by the farmer.

The IRT model estimates are obtained using marginal maximum likelihood estimation (MMLE), allowing for the recovery of both item parameters and the individual-level latent trait. Compared to simple additive scoring methods, the IRT approach provides a theoretically grounded, probabilistic framework that accounts for the heterogeneity of information sources in terms of accessibility and discriminatory power, thereby yielding a more accurate and individualized measure of information acquisition ability.

(2)

Regression Model

To examine the impact of information acquisition ability on farmers’ drought response behaviors, a fixed-effects regression model is constructed based on the variable settings, referencing prior studies [31]:

$$R_{ln}=\alpha +\beta I_{ln}+\gamma X_{ln}+{\sigma }_n+{\epsilon }_{ln}$$

(2)

where R_ln represents farmers’ drought response behaviors, categorized into capital-oriented, labor-oriented, technology-oriented, capital + labor-oriented, capital + technology-oriented, labor + technology-oriented, and capital + labor + technology-oriented behaviors. I_ln represents the farmers’ information acquisition ability. X_ln represents control variables including individual and household characteristics. a represents the constant term. β and γ represent coefficients to be estimated. ε_ln represents the random error term. l and n represent farmers and provinces, respectively. To mitigate estimation bias caused by potential endogeneity stemming from regional-level factors influencing both drought response behaviors and information acquisition ability, province fixed effects (σ_n) are included in the model.

(3)

Threshold Model

Drawing on previous research [32], this study adopts a threshold regression model to examine the segmented effects of farmers’ information acquisition ability on their drought response behaviors. The model is expressed as follows:

$$R_{ln}={\alpha }_0+{\beta }_1{I}_{ln}\left(I_{ln}\le \gamma \right)+{\beta }_2{I}_{ln}\left(I_{ln}>\gamma \right)+\delta Z_i+u_{ln}+{\sigma }_n$$

(3)

where α₀ represents the constant term. I_ln represents the farmers’ information acquisition ability. γ represents the threshold variable. Z_i represents other control variables influencing farmers’ drought response behaviors. u_ln represents the random disturbance term.

(4)

Mediation Effect Model

Referring to Zhang’s [33] methodology, this study employs a stepwise regression approach to explore the mediating role of value perception in the relationship between information acquisition ability and farmers’ drought response behaviors. The models are specified as follows:

$$R_{ln}=c{I}_{ln}+e_1+{\sigma }_n$$

(4)

$$M=a{I}_{ln}+e_2+{\sigma }_n$$

(5)

$$R_{ln}=c^{\text{′}}{I}_{ln}+bM+e_3+{\sigma }_n$$

(6)

where M represents the mediator variable. a, b, c, and c’ represent the parameters to be estimated. e₁, e₂, and e₃ represent the error terms.

3. Results

3.1. Measurement of Farmers’ Information Acquisition Ability

As shown in

Table 2. Parameter estimation results for information acquisition channels.

Information Acquisition Channels	Discrimination Parameter		Difficulty Parameter
	Estimated Value	Standard Error	Estimated Value	Standard Error
Online Media Channel	1.575 **	0.642	1.231 **	0.586
Interpersonal Communication Channel	0.448 ***	0.101	0.536 ***	0.164
Agricultural Training Channel	2.271 ***	0.848	4.432 ***	0.896
Government Department Channel	1.003 ***	0.258	1.392 ***	0.491

Notes: *** and ** indicate significance at the 1% and 5% levels, respectively.

, the discrimination parameters and difficulty parameters of the four channels for acquiring drought response information have passed the significance test at the 1% level. This indicates that these four channels are significantly associated with farmers’ information acquisition ability. The parameter estimation results suggest that online media, agricultural training, and government departments are effective information acquisition channels, while interpersonal communication shows relatively lower discrimination and difficulty parameters.

The Bayesian expected a posteriori estimation method was applied to assess farmers’ information acquisition ability. As presented in

Table 3. Parameter estimation results for information acquisition ability.

Combination Type of Channels	Information Acquisition Ability Parameter	Proportion (%)
No Channels	−0.917	10.67
Online Media	0.116	19.03
Interpersonal Communication	−0.335
Agricultural Training	−0.468
Government Department	0.097
Online + Interpersonal	0.106	28.78
Online + Training	0.394
Online + Government	0.209
Interpersonal + Training	0.137
Interpersonal + Government	0.153
Training + Government	0.228
Online + Interpersonal + Training	0.878	33.51
Online + Interpersonal + Government	0.774
Interpersonal + Training + Government	0.635
Online + Interpersonal + Training + Government	1.839	8.01

, the parameters for farmers’ information acquisition ability range from [−0.917, 1.839]. This range lies within the [−3, 3] interval, indicating negligible bias in the model’s distribution, which conforms to a standard normal distribution. The results show that 10.67% of farmers reported having no channels to acquire drought response information. Meanwhile, 19.03%, 28.78%, 33.51%, and 8.01% of farmers could access information through one, two, three, and four channels, respectively. Farmers with access to more channels or higher-quality combinations of channels demonstrated stronger information acquisition abilities.

These findings confirm that the breadth and quality of information access play a critical role in improving farmers’ information acquisition abilities. As information acquisition ability increases, farmers become more equipped to respond to challenges like droughts, aligning with the notion that diversified and reliable information channels are crucial for enhancing adaptive behaviors.

3.2. The Impact of Information Acquisition Ability on Farmers’ Drought Response Behaviors

As shown in

Table 4. Results of baseline model analysis.

Variable	COB	LOB	TOB	CLOB	CTOB	LTOB	CLTOB
Information Acquisition Ability	0.114 *** (0.015)	0.043 *** (0.002)	0.235 *** (0.046)	0.164 *** (0.011)	0.243 *** (0.003)	0.255 *** (0.046)	0.418 *** (0.065)
Value Perception	0.115 *** (0.031)	0.142 *** (0.067)	0.104 *** (0.018)	0.236 *** (0.031)	0.157 *** (0.007)	0.302 *** (0.033)	0.426 *** (0.004)
Gender	0.054 (0.066)	0.047 (0.715)	0.008 (0.073)	0.023 (0.015)	0.047 (0.026)	0.005 (0.142)	0.069 (0.645)
Age	−0.033 (0.029)	−0.021 (0.067)	−0.019 (0.021)	−0.033 (0.029)	−0.011 (0.067)	−0.063 (0.021)	−0.010 (0.009)
Education Level	0.031 *** (0.001)	0.088 *** (0.067)	0.141 *** (0.002)	0.091 *** (0.030)	0.169 *** (0.027)	0.128 *** (0.013)	0.339 *** (0.054)
Health Status	0.054 *** (0.007)	0.250 *** (0.039)	0.139 *** (0.027)	0.108 *** (0.002)	0.094 *** (0.021)	0.122 *** (0.017)	0.284 *** (0.021)
Household Cultivated Land Area	0.043 (0.032)	0.078 (0.030)	0.062 (0.054)	0.075 (0.023)	0.084 (0.043)	0.082 (0.064)	0.033 (0.017)
Proportion of Agricultural Income	0.001 (0.001)	0.011 (0.013)	0.021 (0.019)	0.043 (0.041)	0.082 (0.066)	0.095 (0.084)	0.007 (0.008)
Land Transfer Participation	0.312 *** (0.039)	0.357 *** (0.025)	0.074 *** (0.004)	0.121 *** (0.020)	0.232 *** (0.032)	0.097 *** (0.004)	0.208 *** (0.095)
Household Labor Force	0.075 (0.062)	0.171 (0.157)	0.064 (0.065)	0.063 (0.060)	0.112 (0.133)	0.067 (0.055)	0.059 (0.037)
Frequency of Natural Disasters in Past Five Years	−0.151 *** (0.055)	−0.427 *** (0.068)	−0.122 *** (0.047)	−0.136 *** (0.021)	−0.327 *** (0.051)	−0.421 *** (0.057)	−0.319 *** (0.058)
Constant	0.493 **	0.554 ***	0.252 ***	0.772 **	0.769 ***	0.773 ***	0.868 ***
Provinces	Yes	Yes	Yes	Yes	Yes	Yes	Yes
p-Value	0.000	0.000	0.000	0.000	0.000	0.000	0.000
R2	0.224	0.276	0.217	0.109	0.334	0.258	0.328

Notes: *** and ** indicate significance at the 1% and 5% levels, respectively. COB: capital-oriented behavior; LOB: labor-oriented behavior; TOB: technology-oriented behavior; CLOB: capital + labor-oriented behavior; CTOB: capital + technology-oriented behavior; LTOB: labor + technology-oriented behavior; CLTOB: capital + labor + technology-oriented behavior.

, the estimated coefficients of information acquisition ability on farmers’ drought response behaviors (capital-oriented, labor-oriented, technology-oriented, capital + labor-oriented, capital + technology-oriented, labor + technology-oriented, and capital + labor + technology-oriented behaviors) are positive and significant at the 1% level. This indicates that farmers with higher information acquisition ability are more likely to adopt measures to address drought. Notably, the effect of information acquisition ability is particularly significant for technology-oriented behaviors, capital + labor-oriented behaviors, capital + technology-oriented behaviors, and comprehensive behaviors (capital + labor + technology). This suggests that enhancing information acquisition ability can help farmers adopt more diversified strategies when facing extreme climatic conditions. For example, better-informed farmers are more likely to adopt drought-resistant seed varieties, drip irrigation systems, soil moisture monitoring devices, and mobile agricultural apps that provide weather forecasts and water-saving guidance. In contrast, less-informed farmers may continue to rely on traditional practices such as increasing irrigation frequency or hiring additional labor without technology support.

Table 4 also shows that the estimated coefficients of value perception on farmers’ drought response behaviors are positive and significant at the 1% level. This indicates that farmers with higher value perception of drought response behaviors are more likely to adopt decision-making strategies to cope with drought. Importantly, value perception has the strongest influence on labor + technology-oriented behaviors and comprehensive behaviors (capital + labor + technology), indicating that stronger value perception drives farmers to adopt more integrated strategies to address climatic challenges. This finding highlights the role of farmers’ awareness of their own interests and environmental values in promoting diversified adaptation measures.

Table 4 shows that several control variables are significantly associated with farmers’ drought response behaviors. Individual characteristics: Farmers with higher education levels and better self-rated health tend to exhibit stronger information acquisition ability, which is correlated with a higher likelihood of adopting technology-oriented and integrated behaviors. Household characteristics: Participation in land transfer is positively associated with farmers’ drought response behaviors, potentially due to improved sustainability of land production and reduced drought losses. Conversely, the frequency of natural disasters experienced by households is negatively correlated with drought response behaviors, especially for labor-oriented and capital + technology-oriented behaviors, where these negative associations are more pronounced. This pattern may indicate that frequent natural disasters are linked to diminished adaptive capacity, making proactive measures more challenging.

3.3. Analysis of the Threshold Effect of Information Acquisition Ability

Bootstrap testing reveals a significant threshold effect of information acquisition ability on farmers’ drought response behaviors (p < 0.1). Different response behaviors correspond to different threshold values: capital-oriented (0.262), labor-oriented (0.014), technology-oriented (0.374), capital + labor-oriented (0.421), capital + technology-oriented (0.535), labor + technology-oriented (0.406), and capital + labor + technology-oriented behaviors (1.109). These thresholds represent the minimum level of information acquisition ability required to significantly trigger each type of drought response. For instance, a relatively low threshold (0.014) suggests that even limited information can promote labor-oriented responses, while a higher threshold (1.109) indicates that comprehensive adaptive behaviors require substantially more information access. Regression estimates were conducted separately for cases where information acquisition ability fell below or exceeded these critical values.

When farmers’ information acquisition ability falls below the threshold, the regression coefficients for all response behaviors are positive. This suggests that, even with lower information acquisition ability, farmers are inclined to adopt drought response measures. However, the intensity of their responses is significantly weaker compared to those with higher information acquisition ability. Notably, even with limited information acquisition ability, farmers may rely on varying resource combinations to respond, albeit with limited overall intensity. For instance, the regression coefficient for labor-oriented behavior is 0.353, indicating that farmers with lower information acquisition ability still rely heavily on labor inputs to address drought challenges. This implies that, in the absence of sufficient informational resources, farmers tend to adopt traditional labor-oriented behaviors (

Table 5. Regression estimation results with farmers’ information acquisition ability as the threshold variable (information acquisition ability ≤ threshold value).

Variable	COB	LOB	TOB	CLOB	CTOB	LTOB	CLTOB
Information Acquisition Ability	0.042 *** (0.007)	0.353 *** (0.026)	0.066 *** (0.031)	0.107 *** (0.009)	0.162 *** (0.068)	0.214 *** (0.053)	0.275 *** (0.072)
Control Variables	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Provinces	Yes	Yes	Yes	Yes	Yes	Yes	Yes
R2	0.144	0.253	0.169	0.197	0.285	0.226	0.279
Threshold Value	0.262	0.014	0.374	0.421	0.535	0.406	1.109
LM Test Value	32.284	28.469	34.269	35.554	39.857	34.273	50.263
Bootstrap p Value	0.004	0.006	0.004	0.003	0.036	0.011	0.026
95% Confidence Interval	[−0.917, 1.839]
Heteroskedasticity Test p Value	0.013	0.010	0.21	0.024	0.032	0.028	0.036

Note: *** denotes significance levels at 1%. COB: capital-oriented behavior; LOB: labor-oriented behavior; TOB: technology-oriented behavior; CLOB: capital + labor-oriented behavior; CTOB: capital + technology-oriented behavior; LTOB: labor + technology-oriented behavior; CLTOB: capital + labor + technology-oriented behavior.

).

Conversely, when farmers’ information acquisition ability exceeds the threshold, the regression coefficients for all behavior types increase significantly. This indicates that farmers with higher information acquisition ability are more likely to adopt diversified and resource-intensive response strategies. Notably, the regression coefficient for capital + labor + technology-oriented behavior is 0.447, markedly higher than that of the low-ability group. This demonstrates that enhanced information acquisition capacity enables farmers to better integrate various resources and adopt more comprehensive response measures. Additionally, the regression coefficient for technology-oriented behavior is 0.266, significantly higher than that of the low-ability group, suggesting that higher information acquisition capacity greatly enhances farmers’ propensity to utilize technology. Higher information acquisition ability allows farmers to access and apply new agricultural technologies more rapidly, thereby improving their ability to address droughts effectively (

Table 6. Regression estimation results with farmers’ information acquisition ability as the threshold variable (information acquisition ability ≥ threshold value).

Variable	COB	LOB	TOB	CLOB	CTOB	LTOB	CLTOB
Information Acquisition Ability	0.133 *** (0.001)	0.356 *** (0.026)	0.266 *** (0.028)	0.253 *** (0.015)	0.285 *** (0.036)	0.236 *** (0.024)	0.447 *** (0.075)
Control Variables	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Provinces	Yes	Yes	Yes	Yes	Yes	Yes	Yes
R2	0.106	0.224	0.169	0.197	0.285	0.226	0.279
Threshold Value	0.262	0.014	0.374	0.421	0.535	0.406	1.109
LM Test Value	32.284	28.469	34.269	35.554	39.857	34.273	50.263
Bootstrap p Value	0.004	0.006	0.004	0.003	0.036	0.011	0.026
95% Confidence Interval	[−0.917, 1.839]
Heteroskedasticity Test p Value	0.013	0.010	0.21	0.024	0.032	0.028	0.036

Note: *** denotes significance levels at 1%. COB: capital-oriented behavior; LOB: labor-oriented behavior; TOB: technology-oriented behavior; CLOB: capital + labor-oriented behavior; CTOB: capital + technology-oriented behavior; LTOB: labor + technology-oriented behavior; CLTOB: capital + labor + technology-oriented behavior.

).

The results of the threshold effect analysis reveal that information acquisition ability exerts varying impacts on farmers’ response behaviors at different levels. When information acquisition ability is low, farmers primarily rely on traditional resources such as labor and capital, adopting relatively simple response behaviors. As information acquisition ability increases, farmers tend to utilize technology and integrate capital and labor resources to implement more diversified response strategies. These findings underscore the importance of enhancing farmers’ information acquisition ability. Specifically, improving information accessibility can effectively encourage the adoption of more complex and comprehensive strategies to cope with extreme climatic events.

3.4. Heterogeneity Analysis

Although the preceding analysis has demonstrated the positive impact of information acquisition ability on farmers’ drought response behaviors, it does not account for potential heterogeneity across different farmer types. To address this, the study further categorizes farmers into connected households (with internet access) and unconnected households (without internet access), as well as into younger farmers (aged below 50) and older farmers (aged 50 and above). From the perspective of household connectivity, the results indicate that information acquisition ability has a significant positive effect on the drought response behaviors of connected households, whereas the effect is not significant for unconnected households. This may be attributed to the ability of connected households to access diversified information channels, thereby enhancing their ability to respond to droughts. From the perspective of farmers’ age characteristics, younger farmers exhibit a significant positive response to information acquisition ability. In contrast, older farmers, while showing a positive effect, do not display a statistically significant response. This discrepancy could stem from the relatively weaker information acquisition ability of older farmers, resulting in insufficient knowledge accumulation and behavioral responses to drought compared to their younger counterparts (

Table 7. Heterogeneity analysis.

Variable	Unconnected Households	Connected Households	Older Farmers	Younger Farmers
Information Acquisition Ability	0.069 (0.057)	0.226 *** (0.029)	0.104 (0.098)	0.342 *** (0.085)
Constant Term	0.332 *	0.025 ***	0.173 **	0.054 ***
Control Variables	Yes	Yes	Yes	Yes
Provinces	Yes	Yes	Yes	Yes
p-value	0.000	0.000	0.000	0.000
R2	0.021	0.158	0.044	0.236

Note: ***, **, and * denote significance levels at 1%, 5%, and 10%, respectively.

).

3.5. Analysis of the Mediating Effect of Value Perception

As shown in

Table 8. Testing the mechanism of information acquisition ability’s effect on farmers’ drought response behaviors.

Variable	Value Perception	COB	LOB	TOB	CLOB	CTOB	LTOB	CLTOB
Information Acquisition Ability	0.324 *** (0.022)	0.076 *** (0.025)	0.087 *** (0.008)	0.081 *** (0.013)	0.076 *** (0.0250)	0.087 *** (0.008)	0.078 *** (0.009)	0.084 *** (0.007)
Value Perception		0.064 *** (0.003)	0.079 *** (0.007)	0.078 *** (0.021)	0.064 *** (0.003)	0.079 *** (0.007)	0.068 *** (0.008)	0.097 *** (0.006)
Constant	1.281 ***	0.233 ***	0.524 ***	0.337 ***	0.233 ***	0.524 ***	0.267 ***	0.374 ***
Control Variables	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Province	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
p-value	0.000	0.000	0.000	0.000	0.000	0.000	0.000	0.000
R2	0.145	0.061	0.079	0.068	0.061	0.079	0.058	0.049
Mediating Effect Ratio		0.276	0.263	0.256	0.284	0.363	0.308	0.433

Note: *** denotes significance levels at 1%. COB: capital-oriented behavior; LOB: labor-oriented behavior; TOB: technology-oriented behavior; CLOB: capital + labor-oriented behavior; CTOB: capital + technology-oriented behavior; LTOB: labor + technology-oriented behavior; CLTOB: capital + labor + technology-oriented behavior.

, the total effect of information acquisition ability is positively associated with farmers’ drought response behaviors—including capital-oriented, labor-oriented, technology-oriented, capital + labor-oriented, capital + technology-oriented, labor + technology-oriented, and capital + labor + technology-oriented behaviors—with statistical significance at the 1% level. This suggests a strong relationship between information acquisition ability and drought response behaviors, potentially operating through enhanced value perception. Further analysis indicates that value perception partially mediates the association between information acquisition ability and farmers’ drought response behaviors. Notably, the mediating effect of value perception accounts for 43.3% of the total association in the case of capital + labor + technology-oriented behaviors, highlighting value perception as an important pathway linking information acquisition ability to the adoption of comprehensive response measures.

4. Discussion

4.1. Policy Implications

This study provides both short-term and long-term policy insights to improve farmers’ drought response behaviors through enhanced information acquisition ability and value perception, particularly in economically underdeveloped regions.

Firstly, breaking down information barriers is essential. To quickly address the issue of limited information dissemination in underdeveloped rural areas, governments and related organizations should prioritize expanding rural network infrastructure and ensuring the accessibility of agricultural information platforms. Low-cost, user-friendly mobile apps, SMS services, and village broadcast systems can serve as immediate solutions to improve the breadth and timeliness of information delivery. Establishing local information-sharing stations can further increase coverage and immediacy.

Secondly, enhance agricultural training services with demonstration effects. In the short term, local governments should organize seasonal and drought-specific training programs, emphasizing hands-on learning. Setting up field demonstration bases where farmers can observe the outcomes of technologies such as drought-resistant seed varieties, drip irrigation, and mulching techniques can boost technology adoption rates.

Thirdly, improving farmers’ value recognition is crucial for sustained behavioral change. Over time, farmers are more likely to invest in adaptive behaviors when they understand their long-term benefits. Thus, policy efforts should focus on integrating economic literacy and environmental education into extension services. Developing visual tools such as cost–benefit charts and multi-year case studies can help shift farmers’ mindset from short-term cost avoidance to long-term risk management.

Fourthly, advancing digital agricultural services is a strategic long-term investment. As younger generations of farmers are more tech-savvy, the development of AI-powered information assistants, predictive drought analytics, and blockchain-based data verification systems will be essential to ensure both accessibility and trust in agricultural information services. Such technologies can eventually be integrated into a regional smart farming ecosystem.

This study also identifies a threshold effect in the relationship between information acquisition ability and drought response behavior. For policymakers, this means that the effect of information acquisition is not linear—it becomes significantly more impactful only after farmers reach a certain level of access and understanding. Therefore, efforts should not only increase the number of information sources but also ensure the quality, usability, and interpretation support of the information provided. Policy tools such as tiered training programs and personalized digital services can help more farmers cross this threshold, thereby unlocking the full benefits of adaptive decision-making.

4.2. Limitations

Despite the valuable theoretical insights and empirical evidence provided by this study, several limitations warrant further reflection, particularly regarding the generalizability of the findings.

Firstly, the data were primarily collected through field surveys in five provinces of Southwest China. While these areas are representative of economically underdeveloped regions with frequent drought occurrences, the socio-economic, cultural, and institutional contexts in other underdeveloped regions—such as those in Northwest or Central China—may differ significantly. For instance, variations in ethnic composition, land tenure systems, and infrastructure development could affect farmers’ access to information and their behavioral responses. Therefore, caution should be exercised when generalizing the findings beyond the sampled areas.

Secondly, this study focuses on information acquisition ability and value perception as key explanatory variables. While these are critical factors, other potentially influential dimensions—such as social capital (e.g., trust networks and community norms), psychological traits (e.g., risk aversion, self-efficacy), and policy incentives (e.g., subsidies or insurance schemes)—were not explicitly modeled. The exclusion of these variables may limit the explanatory power of the model and obscure the interplay between informational, cognitive, and institutional drivers of behavior. Moreover, the value perception variable was operationalized using a single Likert item (“drought responses can stabilize income”), which may not fully capture the multidimensional nature of farmers’ value recognition. This narrow construct mainly reflects economic considerations while overlooking potential social or psychological dimensions, such as perceived social responsibility, intergenerational benefits, or peace of mind. Future research could enhance construct validity by developing multi-item scales that measure economic, social, and psychological aspects of value perception more comprehensively.

Thirdly, the study relies on cross-sectional data, which constrains its ability to capture dynamic behavioral adaptations over time. Farmers’ information acquisition ability and adaptation strategies may evolve with repeated exposure to extreme climate events, changing policy environments, or the gradual diffusion of digital agricultural technologies—none of which can be fully observed through a single time point. This limitation not only results in conclusions that reflect a static snapshot of behavior, but also restricts the ability to infer causal relationships and long-term adaptation trends. Consequently, the findings may underestimate or misrepresent the cumulative effects of technology promotion or policy interventions, thus limiting their applicability for designing sustainable and forward-looking adaptation strategies.

Lastly, while the Item Response Theory (IRT) and threshold models offer valuable analytical frameworks, they are based on certain assumptions (e.g., unidimensionality of information acquisition ability) that may oversimplify the complex, multi-faceted nature of information use in rural settings. Future studies could benefit from integrating mixed-methods approaches, such as combining quantitative panel data with qualitative case studies, to provide a more nuanced and context-sensitive understanding.

4.3. Future Research Directions

Future research could investigate the associations between information acquisition ability and farmers’ adaptive behavior by integrating psychological capital (e.g., self-efficacy, optimism), risk preferences, and social capital (e.g., trust networks, organizational participation) as potential mediating or moderating variables. Such studies would help clarify not only whether information acquisition ability is related to decision-making processes under climate stress, but also the mechanisms through which these relationships operate.

Further, the study of value recognition can be deepened by distinguishing between economic, ecological, and social value dimensions, and by examining their relative weights in farmers’ decision-making. For instance, experimental vignette studies could simulate real-world decision contexts to understand how different types of value information alter farmers’ choices in the short and long term.

From a regional perspective, comparative studies across different ecological zones (e.g., arid vs. humid) and socio-economic contexts (e.g., remote vs. peri-urban areas) would allow for more generalizable insights and targeted policy interventions. These studies could also help identify region-specific bottlenecks in information dissemination and drought response capacity.

To build a more comprehensive analytical framework, future studies are encouraged to adopt interdisciplinary approaches by integrating perspectives and methods from agricultural economics, climate science, rural sociology, behavioral psychology, and data science. For example, coupling agent-based models (from complexity science) with household survey data could simulate behavioral shifts under different policy or climate scenarios.

In terms of methodology, longitudinal tracking using panel data is essential to capture the evolution of information access, behavioral responses, and resilience over time. Additionally, quasi-experimental methods—such as difference-in-differences (DID) based on policy rollout timing or randomized controlled trials (RCTs) of information interventions—can provide stronger causal evidence.

From a policy evaluation standpoint, future research should conduct impact assessments of current agricultural information platforms and digital technology extension programs, examining their inclusiveness, accuracy, and user satisfaction. This could be enhanced through AI-powered analytics (e.g., machine learning to identify information bottlenecks) and big data (e.g., mobile usage, weather alerts, and digital trace data) to design and refine precision-targeted information services.

Lastly, collaborative field-based pilot projects, involving researchers, local governments, and farmer organizations, can serve as “living laboratories” to test the real-world applicability of proposed interventions, generating actionable evidence for scaling up.

5. Conclusions

This study investigates the association between information acquisition ability and farmers’ drought response behaviors in economically underdeveloped regions, using first-hand field survey data. The findings reveal that information acquisition ability is positively correlated with farmers’ adaptive behaviors and also relates indirectly to behavioral changes through strengthened value recognition. Farmers with stronger information acquisition capacities tend to access early-warning data, technical guidance, and relevant policy incentives more effectively, which is associated with more informed and scientific decision-making in the face of climatic stress.

A key contribution of this study is the quantification of the nonlinear and mediated associations of information acquisition: specifically, the identification of a threshold effect, whereby adaptive behaviors appear to increase substantially only once a critical level of information acquisition is surpassed. This suggests that fragmented or superficial improvements in information services may have limited correlation with behavioral change; instead, a comprehensive and sustained approach to enhancing information access seems necessary to observe meaningful shifts in behavior.

Furthermore, the mediating role of value recognition highlights the importance of not just providing information, but also shaping farmers’ understanding of the long-term economic and environmental benefits of drought response measures. This dual-pathway framework—linking information capacity with psychological drivers—offers a more nuanced perspective for understanding farmer adaptation.

From a practical standpoint, the results offer insights for policymakers. Investments in rural digital infrastructure, development of multi-source information-sharing platforms, and context-specific training programs could help farmers cross the identified information thresholds and adopt more diversified, proactive adaptation strategies.

This study contributes novel empirical evidence to the climate adaptation literature and opens avenues for future interdisciplinary and longitudinal research. For example, future studies could explore how digital information services interact with social capital, psychological traits, and environmental exposure to shape adaptive decision-making over time. In sum, this research demonstrates that empowering farmers with adequate and relevant information is strongly associated with enhanced resilience to climate shocks, while recognizing that causal inferences require further investigation through longitudinal or experimental designs.