Risk preference enhances farmers’ access to formal lending by increasing their acceptance of risk-control clauses and willingness to seek financing, thereby effectively alleviating funding constraints and supporting the sustained development of their e-commerce operations. Specifically, risk-preferring farmers tend to prioritize long-term operational returns over short-term risks when making financing decisions, making them more willing to accept collateral, guarantees, and higher interest rates imposed by banks or other formal financial institutions (Hui and Yong, 2023). Such choices reflect rational behavior under a risk–return trade-off: they believe that potential high returns can offset financing uncertainties, and are therefore willing to provide asset collateral and sign strict contracts to overcome the risk-screening thresholds of formal lending and obtain larger loan amounts (Knight, 1921; Kihlstrom and Laffont, 1979). Moreover, when applying for large loans to fund infrastructure construction, brand promotion, warehousing and logistics, or the purchase of digital equipment for e-commerce operations, risk-preferring farmers exhibit stronger financing initiative. They are more willing to disclose detailed business plans and financial information to reduce information asymmetry with financial institutions, thereby improving credit evaluation efficiency and approval rates (Yu and Xiang, 2021). In this process, risk preference not only strengthens farmers’ ability to secure funds but also fosters longer-term cooperative relationships with financial institutions, forming a virtuous cycle for subsequent financing.

The “large-scale and stable” nature of formal lending aligns closely with the rigid investment demands of e-commerce operations (Zhao et al., 2024). E-commerce requires continuous investments in platform participation, brand development, supply chain integration, cold-chain logistics, and information system upgrades, with concentrated upfront capital needs and a relatively long payback period. Farmers obtaining formal financial support can make larger investments in brand building and technological upgrading, enhancing their bargaining power in platform competition and market expansion. At the same time, structured loan contracts provide greater flexibility in capital allocation, helping smooth cash flows during off-season periods and preventing operational disruptions caused by short-term liquidity shortages. More importantly, the risk management, credit record accumulation, and follow-up financial services provided by formal financial institutions improve farmers’ operational standardization and risk management capabilities, offering sustained credit backing for future business expansion and additional financing (Yu and Xiang, 2021).

Overall, risk preference not only directly drives farmers to overcome the entry barriers of formal lending, but also improves financing conditions and capital allocation efficiency, providing critical financial support for the initiation and expansion of their e-commerce operations and strengthening their long-term market viability. Based on the above analysis, this paper proposes the following research hypotheses:

Risk preference drives farmers to use social networks for informal borrowing by reducing the cost of personal connections and the sensitivity to risks of informal contracts, thereby alleviating financial constraints. On the one hand, risk-preferring farmers are less sensitive to the cost of favors in rural acquaintance societies and are more likely to view social networks as a low-cost financing channel (Zhang et al., 2020). While traditional rural lending is often accompanied by the stress of complex human interactions, risk preference individuals are more concerned with the immediate benefits of capital availability for entrepreneurial opportunities. For example, when facing a short-term financial gap, risk-preferring farmers will take the initiative to ask their friends and relatives for loans, and prioritize the need for entrepreneurial capital by weakening the psychological expectation of “pressure of repayment of favors.” On the other hand, risk-preferring farmers are more tolerant of the risks of informal contracts and are more willing to accept the flexibility of informal lending. In the case of insufficient formal financial credit, farmers usually lack the guarantee of standardized contracts and face higher risks of financial security when they raise funds through channels such as family and friends lending and mutual aid associations (Anh et al., 2022). However, based on their confidence in their own risk-taking ability, risk preference people will actively reduce the requirements for contract completeness and place more emphasis on the timeliness of the availability of funds. For example, in order to seize market opportunities, some farmers choose to accept verbal agreements or simple written covenants even though they are aware of the risk of information asymmetry that may exist in informal lending, in order to quickly obtain funds to invest in production.

The “convenience and flexibility” characteristics of informal lending make it an important tool for farmers to meet their short-term capital needs. First of all, when farmers are faced with unexpected financial needs such as e-commerce promotions and logistics emergency purchases, they can quickly raise funds through the network of friends and relatives to avoid missing market opportunities due to cumbersome processes. Especially in the peak season of the live broadcast with goods, farmers need to raise funds in a short period of time to expand inventory, informal lending “borrow and get” characteristics can quickly ease the liquidity constraints, to ensure the smooth implementation of promotional activities. Second, informal lending relies on the implicit guarantee mechanism of neighborhood reputation (Li Q. et al., 2023), which significantly reduces transaction costs through credit constraints. Risk-preferring farmers are more willing to take on potential social credit risks in exchange for the immediate availability of funds. Although the amount of this type of financing is usually small, the approval process is simple and the time for funds to be available is short, which can effectively respond to the high-frequency and small-amount capital needs in entrepreneurial operations and enhance operational flexibility. Although informal lending is difficult to scale up due to geographic constraints, its advantages in short-term financing and flexible risk sharing make it an important complementary mechanism to the formal financial system (Phan et al., 2024).

It is thus evident that informal lending plays an irreplaceable buffering role in meeting the flexible funding needs of e-commerce operations, providing critical support for risk-preferring farmers to seize short-term market opportunities. Based on the above analysis, this paper proposes the following research hypothesis:

The dependent variable in this paper is the e-commerce operations of farmers. Referring to the study of Qiu et al. (2025a), farmers’ sales activities through online platforms are regarded as having e-commerce operations and assigned the value of 1; conversely, the value is 0. This variable directly captures whether farmers have adopted e-commerce operations as a new business model, serving as the core decision and key behavioral outcome in this study.

The core independent variable of this paper is risk preference. Referring to the existing research (Aren and Nayman Hamamci, 2020), the risk preference is measured by the holdings of farmers’ financial products. The use of financial product holdings to measure risk preference is based on the following theoretical considerations. First, according to the theory of revealed preference, an individual’s true preference are reflected in their actual behavior. Farmers’ real investment decisions in the financial market provide a more objective and authentic representation of their inherent risk tolerance. Farmers who hold high-risk financial products have, in effect, demonstrated through their behavior that they possess a higher risk preference. Second, e-commerce operations share similar decision-making characteristics with financial investments, such as facing uncertainty and pursuing potentially high returns. Therefore, using risk preference in the financial domain to predict farmers’ adoption of e-commerce offers strong validity. Based on this logic, farmers are defined as having a high risk preference if they hold any of the specified financial products, which is recorded as a value of 1, and as having a low risk preference if they do not, which is recorded as a value of 0.

Based on the theoretical analysis, risk preference promotes farmers’ e-commerce operations by driving both formal and informal lending. Referring to the study of Phan et al. (2024), the questionnaire “whether farmers borrowed from formal banks” was selected to measure formal lending, and if they did, it was assigned a value of 1; on the contrary, the value is 0. Formal lending reflects a farmer’s proactive entry into the formal credit market and the use of external capital to expand business scale, demonstrating an “active financing capacity” based on risk tolerance. Meanwhile, the questionnaire “whether farmers borrowed from the private sector” was selected to measure informal lending, and if they did, it was assigned a value of 1; on the contrary, the value is 0. Informal lending relies more heavily on social networks and trust relationships, representing the farmer’s “mutual-assistance financing capacity.” Together, these two variables capture, from the perspectives of both formal and informal institutions, how risk preference shapes farmers’ resource-mobilization strategies, making them ideal mechanism variables to connect risk preference with e-commerce operations decisions.

In addition to the core independent variable, to accurately identify the net effect of risk preference on farmers’ e-commerce operations. Combining the existing data indicators and referring to the research of Rizwan et al. (2020), this paper selects the age, gender, education level and health status to describe the individual characteristics of farmers; household economic status, household size, household cultivated land area and household non-farm employment status are selected to reflect family characteristics; Select distance of the village from the county government, economic status of village and location to reflect the regional characteristics. The descriptive statistics of all variables are shown in Table 1.

As the e-commerce operations of farmers is a binary variable, this paper uses the binary logit model to empirically test the impact of risk preference on the e-commerce operations of farmers, and constructs the regression model as follows:

In Equation (1), Y i represents farmers’ e-commerce operations; R P i is the risk preference; X i is a series of control variables; α is a constant term; βi , δi are the coefficient of the model to be estimated; εi is the error term.

On the basis of exploring the impact of risk preference on farmers’ e-commerce operations, this paper further examines its mechanism of action. With reference to Baron and Kenny (1986), constructed the following mechanism effect model. The model Equations (2–4) are as follows.

In equation: Y i , R P i , M ni are for the e-commerce operations, risk preference and mechanism variables, respectively, n is taken as 1 and 2, M n 1 is formal lending and M n 2 is informal lending; X i denote the control variables; γ 1 , γ 2 , γ 3 are the constant term; ϖ i , ψ i , ϕ i , λ i , a i , b i , c i are the coefficients of the model to be estimated; ω i , μ i , σ i are the error term.

In order to investigate whether the previous regression results are robust, this paper uses the amount of financial products purchased to measure risk preference and conduct regression. In order to avoid the interference of outliers, a 1% bilateral tailing treatment was applied to the risk preference index. The regression results are shown in column (1) of Table 3. The results show that risk preference still has a significant positive impact on Farmers’ e-commerce operations, and promotes farmers’ e-commerce operations. The above regression results are robust.

Considering that farmers’ e-commerce operations is a binary variable, this paper adopts Probit model instead of logit model for re-estimation, and the results, as shown in column (2) of Table 3, the risk preference is still significantly positive at 1% statistical level, indicating the robustness of the above regression results.

Considering the special characteristics of old farmers in physical ability and cognition, in order to avoid the interference of special samples, this paper deletes the sample of farmers over 70 years old for re-regression, and the regression results, as shown in column (3) of Table 3, show that the risk preference still has a significant positive impact on the e-commerce operations of farmers, which confirms that the previous results are robust.

Referring to existing research (Qiu et al., 2025b), this paper uses the number of online stores to measure the e-commerce operations of farmers and conducts a re-regression. The regression results are shown in column (4) of Table 3. The impact of risk preference on farmers’ e-commerce operations is significantly positive at the 1% statistical level, further verifying the robustness of the benchmark regression results.

In order to determine whether the unobserved randomization factor affects the regression results, this paper sets up an experimental group for the placebo test with reference to the study of Niu et al. (2022), and the results are shown in Figure 3. After repeating this randomization process 1,000 times, the distribution of the regression coefficients of risk preference is concentrated around zero, which is approximately normally distributed, in line with the expectation of “no true effect.” At the same time, the coefficients of the benchmark regression are much higher than the distribution of the coefficients generated by random replacement, indicating that the significant positive impact of risk preference on farmers’ e-commerce operations is not caused by random noise, and the placebo test passes, eliminating the potential impact of unobserved factors, indirectly proving that risk preference has a significant positive impact on farmers’ e-commerce operations, and verifying the reliability of the results of the previous paper.

Possible endogeneity of reverse causality between risk preference and farmers’ e-commerce operations. On the one hand, farmers with a high risk preference are more inclined to choose e-commerce operations in pursuit of high returns from entrepreneurship; on the other hand, fluctuations in returns from e-commerce operations may in turn reverse the shaping of farmers’ risk preference. Therefore, in this paper, we refer to the study of and use the instrumental variables approach to address the potential endogeneity problem (Qiu et al., 2024b).

Imprinting Theory suggests that childhood experiences influence an individual’s character formation, cognitive attitudes, and behavioral patterns, among other things (Daines et al., 2021). In terms of physiological and psychological theories, the brain nerves are extremely sensitive to personal experiences, and the feelings generated by the events people experience create memory anchors and are internalized into a belief (Wing et al., 2018), which serves as information that influences preference and decision-making judgments. The Great Famine of 1959–1961 in China was a major exogenous event that caused a large number of population deaths (Cheng et al., 2021). Individuals’ early fear of famine trauma carries over into adulthood, creating a negative impact that stays with them throughout their lives (Arage et al., 2021; Zhou et al., 2022; Liang et al., 2023; Guo et al., 2024). It has been found that early famine experiences profoundly affect farmers’ perceptions and risk preference (Chen et al., 2023a; Wang et al., 2024), leading to higher savings rates (Ding et al., 2022; Guo and Chen, 2025), and that farmers who have experienced a famine are more likely to seek stable returns that manifests itself in risk-averse tendencies. It’s clear that famine experiences meet both the exogeneity and correlation requirements by shaping risk preference. Therefore, this paper chooses famine experience as an instrumental variable and measures it according to the age of the survey respondents. If the respondent was born in 1961 and before, he/she is considered to have experienced the Great Famine and is assigned a value of 1. If he/she was born in 1962 and after, he/she is considered not to have experienced the Great Famine and is assigned a value of 0.

The results of the endogeneity test of the effect of risk preference on farmers’ e-commerce operations are shown in Table 4. From the Wald and Corr values, the original hypothesis that the variables are exogenous is rejected, i.e., there is an endogeneity problem. The results of the first-stage regressions of the IV-Probit model and the ERM model show that the effect of famine experience on the risk preference of the farmers is significantly negative at the 1% statistical level and the F-value of the joint test of significance is greater than 10, which suggests that there is no problem of weak instrumental variables. Meanwhile, the second-stage regression results of both the IV-Probit model and the ERM model show that risk preference still has a significant positive impact on farmers’ e-commerce operations after the introduction of instrumental variables, which further validates the reliability of the findings of this paper.

Risk preference, as a subjective psychological characteristic of farmers, may be influenced by certain unobserved factors, leading to a non-random selection of the sample and creating endogeneity problems. In order to effectively solve the endogeneity problem caused by self-selection bias, this paper adopts the propensity score matching (PSM) method to compensate for the limitations of traditional regression analysis methods. Referring to the study of Qiu et al. (2024a), eight different matching techniques, namely k-nearest-neighbor matching, radius matching, kernel matching, one-to-one matching, nearest-neighbor matching, mahalanobis matching, local linear matching, and caliper matching, were used in this study to re-run the regression analysis, and the results are shown in Table 5. The results of the analysis show that risk preference has a significant positive impact on farmers’ e-commerce operations, which further validates the robustness of the study’s findings.

PSM model specification needs to meet the two preconditions of overlapping assumption and balance characteristics (Caliendo and Kopeinig, 2008). For the overlapping hypothesis that the public support domain needs to be tested, Figure 4 shows the distribution of propensity scores for testing the public support domain. After matching, the high-risk preference sample group and the low-risk preference sample group almost overlap in the propensity score, and there is a large common support interval, which indicates that the matching is more reasonable.

In order to check whether the above propensity score matching estimation results balance the data well, it is necessary to conduct a balance test. The balance test results are shown in Table 6. In this paper, the one-to-one matching method is taken as an example. Compared with before matching, the standardized deviation rate of control variables is reduced after matching, and the standard deviation rate of most control variables is less than 10%. At the same time, the t-test results of most control variables do not reject the original hypothesis that there is no systematic difference between the treated group and the control group, and effectively balance the differences in the distribution of covariates between the treated group and the control group. Therefore, the propensity score matching results passed the balance test.

There are differences in the bargaining power and capital constraints of different e-commerce scales, which may lead to differences in the e-commerce operation effects of risk preference. Therefore, this paper takes the number of online stores as a measure of e-commerce business scale, and regards the situation that the number of online stores is greater than 1 as a scale e-commerce, on the contrary, it is a non-scale e-commerce, and, respectively, discusses the impact of risk preference on different e-commerce operation scales. The results are shown in columns (1) and (2) of Table 9. The results show that risk preference can better promote farmers’ scale e-commerce operation. Under the non-scale e-commerce operation mode, the resource allocation is concentrated on basic operation, the capital and technology threshold is low, and the risk is controllable, but the market expansion has structural bottlenecks, which is difficult to fit the decision-making characteristics of risk-preferring farmer. In contrast, scale e-commerce operations involve higher sunk costs and longer return cycles, presenting a typical high-risk, high-return profile. This model is better aligned with the traits of risk-preferring farmers, namely their higher tolerance for uncertainty and long-term income expectations. Moreover, by leveraging economies of scale, farmers can reduce unit operating costs, thereby achieving synchronous improvement in both operational scale and economic benefits. Therefore, risk-preferring farmer are more inclined to operate scale e-commerce.

At this stage, there are two main types of e-commerce operation models: social e-commerce and platform e-commerce (Liu et al., 2021), and there are differences in the adaptation rules and business strategies of different e-commerce models. This paper further explores the differences in the impact of risk preference on different e-commerce models, and the results are shown in columns (3) and (4) of Table 9. The results show that risk preference are more likely to promote farmers to operate social e-commerce model. The possible reasons for this are that, on the one hand, platform e-commerce has a high degree of operational threshold and standardization, which usually requires high initial investment, such as store decoration, deposit, operating costs and continuous professional management ability, and the fixed costs and complexity of platform e-commerce also offset the incentive effect of risk preference to a certain extent; on the other hand, it lies in the fact that social e-commerce relies mainly on personal social networks such as the promotion of WeChat circle of friends. The start-up costs are low, the operation mode is more flexible, and the form of return is more direct. Therefore, risk-preferring farmer are more inclined to choose the social e-commerce model.

As a new technological approach, live-streaming is gradually being used in e-commerce operation activities (Zhang et al., 2024). This paper further examines the differences in the impact of risk preference on different e-commerce technology utilization methods, and the results are shown in columns (5) and (6) of Table 9. The results show that risk preference significantly contributes to live-streaming e-commerce operations. The possible reason is that live broadcasting, as an emerging business mode, on the one hand, breaks the geographical and time limitations of the traditional sales model, can directly face the vast consumer groups, and the sales of the products may be explosive growth, which will bring lucrative profits; on the other hand, live-streaming e-commerce has a high degree of uncertainty and risk, and the live-streaming effect is subject to multiple factors such as the anchor’s personal ability, market trends, audience preferences, etc. The effect of the live broadcast is affected by multiple factors such as the anchor’s personal ability, market trends, audience preferences, etc., and may face risks such as no one watching the live-streaming and poor sales. Therefore, risk-preferring farmers are more inclined to choose live-streaming e-commerce.

Products with different attributes are differentiated in terms of market strategies and business objectives due to differences in their type characteristics. Therefore, this paper further examines the differences in the impact of risk preference on different types of products operated and the results are shown in columns (7) and (8) of Table 9. The results show that risk preference significantly contributes to farmers’ operation of primary processed products. The possible reason is that, on the one hand, the production of agricultural products is greatly affected by natural factors, the yield and price fluctuations are relatively large, the operational risk is difficult to control by themselves, and the return is relatively limited, the farmers’ willingness to operate agricultural products e-commerce operation is lower; on the other hand, the primary processed products are simply processed, with high added value and profit margins, and the market demand is more stable, and the shelf life is longer, the storage and transportation conditions are more flexible, which is more suitable for the pursuit of high returns by risk-preferring farmers. Therefore, risk-preferring farmers are more inclined to choose to operate primary processed products e-commerce.