How Does Financial Literacy Affect Digital Entrepreneurship Willingness and Behavior—Evidence from Chinese Villagers’ Participation in Entrepreneurship

Abstract

How financial literacy (FL) can contribute to both digital entrepreneurship (DE) willingness and behavior has become a widely concerned and controversial topic among scholars and practitioners, one that is particularly relevant for bottom of pyramid (BOP) entrepreneurs. However, existing literature has not established a systematic theoretical framework for this issue, and its internal mechanism has not been thoroughly revealed. Hence, based on behavioral finance theory and digital business-related literature, we construct a multi-layer linear and mediated-moderation model that relates latent variables of FL, financial inclusion (FI), digital literacy (DL), and DE. A total of 664 villagers carrying out DE are the sample from China, and the empirical results mainly reveal the following: FL significantly improves the DE willingness and behavior, and the sense of FI plays a partial mediating role between FL and DE willingness and behavior. DL is a crucial influencing factor that moderates a positive effect between FL and DE willingness and behavior. Moreover, DL plays a mediated-moderating role. By contrast, FL in poverty-stricken counties (PSC) has a more substantial influence on DE than that in non-poverty-stricken counties (NPSC). If the family background (social capital) is superior (excellent), and the resources are abundant, then the influence of FL on DE will be stronger. This research contributes to the theories and practices related to FL, DE, FI, and DL.

1. Introduction

Digital entrepreneurship (DE) is an omnipresent element in changes to the entire business world (Bruton et al., 2013) [1], one which has also prompted scholars and practitioners to conduct an in-depth investigation of it (Kraus et al., 2019) [2]. However, existing literature on DE predominantly focuses on establishing digital ventures or digital projects or platforms and other related issues. By contrast, the discussion on entrepreneurial activities based on existing digital platforms, defined as participation in DE (in reality most DE activities), is sparse, because the “threshold” to participate in DE is relatively low. For example, under some pioneers’ leadership, apps (e.g., Taobao Live broadcast, Tiktok, Kuaishou, and Weibo) are used for sales and precise production to rapidly make the product known. This method has not only increased (or stabilized) incomes but also opened the way for individuals to participate in DE. In recent years, similar inspirational stories have circulated widely in developing countries and in less developed regions, such as India and Pakistan (https://baijiahao.baidu.com/s?id=1671255945457330372andwfr=spiderandfor=pc accessed on 26 September 2022). However, although participation in DE opens a pathway out of poverty for people, specifically low-income and low-resource (double low) groups such as villagers, some critical antecedents exist that are essential to their success. How to successfully achieve DE a is very important question, particularly given a general lack of capital, capacity, and education, (Nambisan, 2017) [3]. The reason is that traditional labor methods and products can be “digitally” connected to the industrial supply chain only under the promotion of these factors (specifically capital). However, society seems to pay additional attention to the new DE mode. By contrast, the opportunities and challenges brought by important antecedents to participate in DE are seldom discussed (Kraus et al., 2019) [2]. Although the number of relevant studies is small, we can still find some studies that involve the influencing factors of digital entrepreneurial intention.

Recently, the literature has generally regarded digital literacy (DL) (Nambisan, 2017) [3] as the most crucial influencing factor of DE motivation. The reason is that people can stimulate the motivation and enthusiasm of DE only when they use internet devices well and have the essential ability to operate computers and mobile phones (Neumeyer, 2020) [4]. However, with the rapid development of digital technology in recent years, the use of digital (mobile) devices and applications (e.g., apps) is becoming increasingly intelligent and “dumb” DL’s influence on DE, specifically participation in DE, has also significantly changed, with some of the previous ideas having become somewhat outdated. For example, Sutter et al., (2019) [5] noted that capital, information, market, and management ability are still the most important factors influencing DE’s success. Therefore, the effect of DL is also different from the past. For most people (specifically the relatively poor), the importance of capital-related financial factors, including financial literacy (FL) and financial inclusion (FI), is prominent. Zhang and Xiong (2017) [6] assessed the FL of Hubei and Henan province. They found that the respondents’ FL was overall not high, specifically in terms of their understanding of financial knowledge, risks, and costs; something which is a significant obstacle to entrepreneurship and wealth. Su and Kong (2019) [7] also found that FL significantly affected the decision-making and behavior of entrepreneurship, and higher FL could promote entrepreneurial behavior. Then, how does the FL of low-income and low-resource groups affect their innovation behaviors, specifically their participation in DE?

Previous studies have shown that, compared with other factors, FL plays an increasingly significant role in promoting DE, specifically in participation in DE. Furthermore, in this effect, DL and FI play different roles. However, existing studies have not established a systematic theoretical framework for this issue. Moreover, the internal mechanism has not been thoroughly explored, which also provides a new research direction and perspective for this study: under the constant digitalization penetration, will people’s FL affect their willingness and behavior to participate in DE? If this effect exists, what is the underlying mechanism? Moreover, under different social and economic macro environments (primarily regional circles) and the surrounding conditions of individual family backgrounds, will the functional relationship between variables be different, and what are the possible causes? To our best knowledge, few studies have addressed these issues specifically.

Because participation in DE is most widely adopted by those in the bottom of the pyramid (BOP)—namely, low-income and low-resource groups—the current article seeks to address the shortcomings of previous literature that converged DE within the framework of willingness and behavior to participate in DE and focuses on the influence of FL on willingness and behavior to participate in DE. Furthermore, some studies have proposed or verified a role for FI and DL as similar influences. Thus, this study also introduces hypotheses that state that FI and DL’s act as mediators and moderators, respectively.

2. Literature Review and Hypothesis

2.1. DE Theory and Participating in DE

DE is a new research field in entrepreneurship theory that has emerged as some widely respected successful companies, such as Facebook, Bytedance, and Tencent, have set examples of DE. Different from traditional entrepreneurship theories, DE is highly subversive (for the economy), mobile (for the society), and driving (for the industry) (Yu et al., 2017; Nambisan et al., 2017) [3,8]. Zhu et al., (2020) [9] provided a complete definition of DE, that is where, under the vigorous development of the digital economy, entrepreneurs and teams create digital products and digital services through the massive use of digital technologies, social media, and other emerging information and communication technologies. Related research implies two different types of DE and entrepreneurial groups.

The first is the type of digital enterprise (organization) that develops and operates digital products and services. Digital entrepreneurs build digital platforms and create digital running institutions (including digital platforms, such as Taobao and TikTok). Their success depends on their high DL or digital ability (Scuotto and Morellato, 2013) [10]. The second is participating in DE, that is, starting a business around a digital enterprises’ products and services. These participants conduct entrepreneurial activities around digital operating institutions (or platforms) (e.g., Taobao sellers and Douyin creators), and their success does not require excessive DL (He and Li, 2019) [11].

Previous studies on DE have mainly focused on digital entrepreneurs and discussed the antecedents and consequences of their entrepreneurial intention and behavior (Hafezieh et al., 2011; Hamid and Khalid, 2016; Dy et al., 2018) [12,13,14]. Moreover, research on DE participants and their DE activities is limited. In comparison, the number of participants in DE is vast, and its influence cannot be ignored. He and Li (2019) [11] have stated that most participants of DE can still succeed in starting a business even if they are low in DL, which attributes to the increasingly vital convenience, security, and reliability of DE platforms. More importantly, participation in DE can provide an easy way to start a business for the less educated and less well-qualified, including the urban poor (perhaps from slums in India, South-east Asia, and Africa) and ordinary rural workers (e.g., villagers).

The recent literature has focused on DE among villagers, small towns, and peri-urban groups because they comply with the sample characteristics of the “double low” group. Relevant studies mainly include motivation, willingness, behavior, and results of DE (Jiang and Guo, 2012; Liu et al., 2019; Wu et al., 2020) [15,16,17]. For example, Amit and Muller (1995) [18] proposed that the pursuit of a happy life is first manifested in the motivation and awareness of starting a business and then in entrepreneurial behavior. However, Zeng and Wang (2009) [19] point out that entrepreneurial motivation or consciousness will continuously stimulate, maintain, and moderate their entrepreneurial activities, and entrepreneurial intention and behavior are often intertwined. Therefore, the two are discussed as a whole in most cases. In addition, Nambisan et al., (2017) [3] emphasized that the advantages of DE are low diffusion cost, fast speed, and small spatial-temporal constraints, which are very suitable for groups with insufficient venture capital and relatively free and abundant time. In recent years, some domestic digital enterprises have set up unique platforms to attract BOP to participate in DE, such as JingDong Fresh and Rural Taobao. This process further promotes villagers’ willingness and behavior to participate in DE.

2.2. Impact of FL on DE Willingness and Behavior

Traditional entrepreneurship theories pay much attention to discussing the influence of venture capital on entrepreneurs’ willingness and behavior. Nonetheless, whether entrepreneurs have sufficient capital is closely related to their FL (Ding and Zhang, 2019) [20]. Hazlett (2015) [21] first proposed the concept of FL and then broadly summarized FL as the use, management, and decision-making ability of funds. Huston (2010) [22] pointed out that literacy should focus more on micro individuals rather than excessively on organizations and groups. Therefore, he believed that FL is a measure of an individual’s ability to understand, process, and use financial-related information, and was supported by some scholars (Albaity, 2019; Wang et al., 2020) [23,24].

Based on years of Chinese household financial situation assessment, the CHFS survey database funded by Professor Gan Li’s team continues to examine the financial conditions of Chinese households (including wealth accumulation, debt, financial knowledge, and others), individual employment, entrepreneurial behavior, and other indicators from one aspect and regularly releases research reports. Some of these studies have explored FL and its consequences. According to the China Rural Family Financial Development Report released by CHFS (http://www.springer.com/gb/book/9789811004087 accessed on 26 September 2022), rural families’ FL is relatively low, something which has also been verified by some experience. Moreover, some empirical research results have shown that, compared with the urban population, the FL of the BOP is generally weak (Wu et al., 2016; He and Li, 2019) [11,25]. In recent years, the FL of this group (BOP) has attracted scholars’ attention, and some studies have been accumulated. Existing studies generally believe that FL has a significant effect on entrepreneurial behavior (Li et al., 2014; Yin et al., 2015) [26,27]. However, as for different backgrounds and groups, the study of FL on entrepreneurial intention and behavior is still insufficient.

Moreover, scholars generally believe that the lack of FL leads to a shortage of venture capital, which significantly affects the willingness and motivation to start a business (Bernheim, 2001; Lusardi, 2014) [28,29]. Previous studies have shown that insufficient funds are the biggest obstacle to innovation and entrepreneurship for the BOP. This effect is higher than education and entrepreneurial ability (Banerjee and Newman, 1993; Zhang and Zhang, 2013) [30,31] gender, age, per capita income level, migrant workers (He and Li, 2019) [11], broadband and traffic development level (Alderete, 2015) [32] and others. In recent years, with the increasing popularity of digitization, the cost of starting a business using digital platforms is far lower than the traditional way of starting a business (Nambisan et al., 2017; Yu et al., 2017) [3,8].

The study found that those with the characteristics of “double low” could not break away from their original and familiar working methods (e.g., part-time jobs and engaging in DE with leisure time) when participating in DE, but could better solve the problems of products and services selling and precise production. In reality, the villages and small handcraft producers in some rural areas have quickly joined network broadcast and internet sales ranks. They have gradually established a strong relationship with end consumers.

Hence, in-depth research on the DE willingness and behavior of BOP groups is necessary. In recent years, we have found that some studies have begun to discuss the effect of FL on their entrepreneurial intention and behavior (Zhang and Xiong, 2017) [6]. However, existing studies have not fully revealed its internal mechanism of action, specifically the lack of reliable empirical research conclusions. Based on this, we propose the hypothesis H1.

Hypothesis 1 (H1):

FL promotes the willingness and behavior of DE.

2.3. Mediating Role of FI

FI is generally considered the antecedent or predictive variable of financial behaviors (e.g., loan, deposit, investment, and financial management). FI is also widely used to explain the influence on people’s economic activities, such as entrepreneurship and consumption (Zhan and Xu, 2017) [33]. FI originated from the concept of “inclusive finance” proposed by the United Nations in 2005, which emphasizes the accessibility, availability, and permeability of financial services in economies (Sarma and Pais, 2011; He and Miao, 2015; Zhang et al., 2019) [34,35,36]. Among these, the most important thing is to highlight the friendliness of finance. Most previous studies have proposed that FI is generally an external environmental factor. However, with the continuous improvement of financial inclusiveness, such changes can also be transmitted to people’s minds, forming a change of people’s “feeling” in FI and influencing people’s new judgment of FI (or friendliness). According to the description of Kpodar (2011) [37] and Wang and Chen (2016) [38], this study summarizes the sense of FI as the process and status of obtaining a financial product or service in a timely and effective manner, within an affordable cost, and in a relatively fair way.

Considering that the sense of FI mainly comes from FI’s external stimulus, a positive influence relationship with FI should be formed. Although the discussion on FI’s antecedent effect in current studies is limited, many accumulations of antecedent studies on FI as a variable still exist. Based on the existing literature, relevant studies have mainly discussed the influence of external factors, such as residents’ income, population structure, and technological progress on FI. For example, Li and Fan (2006) [39] emphasized the importance of residents’ income and demographic composition to FI. Tian (2012) [40] and Geach (2007) [41] believed that communication technology had improved the inclusiveness of finance. In a minimal number of papers, we also find some studies discussing the influence and role of FI. For example, Zhou et al., (2018) [42] verified the positive effect of individual FL on FI by using the sectional data of 1694 counties in China in 2010, which was also verified by Wagner et al., (2018) [43].

In addition, the sense of FI strongly influences people’s activities (consequences), specifically entrepreneurial behavior. For example, Banerjee and Newman (1993) [30] and Yin et al., (2015) [27] proposed that people’s “good perception” of financial services (timely + effective) can greatly promote financial behaviors, thus easing financing constraints, dispersing operational risks, and improving the success rate of entrepreneurship. Zhang et al., (2008) [44] verified the role of accessibility in FI from the perspective of financial institutions. They pointed out that FI could effectively reduce the restriction of family-owned property level on entrepreneurship, significantly alleviate financial pressure and improve entrepreneurial motivation. Moreover, Cnaan et al., (2012) [45] and Lu et al., (2014) [46] verified that the level of financial development (service quality + permeability) is an essential factor affecting entrepreneurial intention. Zhan and Xu (2017) [33] investigated the influence of inclusive financial development level on entrepreneurial ability. They pointed out that if their FI level is high, then people have great entrepreneurial ability.

The literature review finds that FI is a mediator in FL’s influence on entrepreneurial intention and behavior. Based on this consideration, we propose the following hypotheses:

Hypothesis 2 (H2):

FL promotes a sense of FI.

Hypothesis 3 (H3):

The sense of FI promotes the willingness and behavior of DE.

Hypothesis 4 (H4):

The sense of FI mediates FL’s promotion effect on DE willingness and behavior.

2.4. Moderating Effect and Moderated-Mediating Effect of DL

Management scholars first regarded DL as an ability to explain people’s digital behavior and activities. Therefore, Eshet (2012) [47] defined DL as a set of various digital capabilities. The reason is that Alkalai first proposed the concept of DL and summarized it as “the ability to understand and use various digital resources and information displayed on the computer.” Many scholars have recognized this view (Gilster, 1997) [48].

However, the original viewpoint has been greatly expanded. For example, Eshetalkalai (2004) [49] proposed that DL is the digital age product, which is the embodiment of an individual’s ability and contains the digital temperament, vision, and attitude that an individual should have toward numbers. Currently, many Chinese scholars have expanded the concept of DL from the perspective of culture and literacy, such as Chen and Shi (2019) [50] due to the rapid development of digital technology and business in China. According to the latest research, DL has been constructed as a more general concept that includes the digital ability and related connotations of digital psychology, consciousness, and ethics.

However, in a narrow sense, DL can still come down to people’s numerical ability. Throughout the previous studies, DL has been carried out mainly from the narrow definition of digital ability (Alamutka, 2011) [51]. Previous studies have focused on the effect of DL. For example, based on the unified theory of acceptance and use of technology model, Mohammadyari and Singh (2015) [52] concluded that DL positively affected people’s behavioral expectations. Oggero (2020) [53] found that if the DL level is high, then entrepreneurs’ innovation spirit and entrepreneurial behavior can be easily stimulated. Islami et al., (2019) [54] found that DL is positively correlated with college students’ entrepreneurial orientation and behavior. Furthermore, some recent literature has explored the interaction between DL and other variables. For example, Zhu et al., (2020) [9] believed that DE is not only a “net effect” of a single factor of DL but a relatively complex process, which will be comprehensively affected by the interaction and integration of different factors. Nambisan (2017) [3], in particular, believes that the expansion of digital technology and DL interaction have a positive effect on DE. Zhu et al., (2020) [9] even considered that the interaction between technology and DL is the beginning of digital behavior (including DE). Yu et al., (2017) [8] believe that the interaction of DL and willingness to adopt information technology (ICT) can influence the adoption behavior of information communication technology. The above findings suggest that, with further research, DL’s moderating effect my come to be valued. Therefore, this study proposes that the interaction mechanism between DL and FL is likely to affect people’s willingness and behavior of DE. Based on this, we propose hypothesis H5.

Hypothesis 5 (H5):

DL moderates the positive effect of FL on DE willingness and behavior.

Existing research shows that DE needs to have certain DL, including equipment operation, information processing, communication, and collaboration capabilities. Scholars have found that DL can effectively promote farmers’ willingness to participate in DE. The fundamental reason is that DL can significantly improve the perception of financial friendliness of farmers with good FL at the micro level. In other words, individuals with higher DL have a more subjective initiative for information search and a more objective ability for information acquisition, which are helpful to overcome financing constraints, enhance the perception of financial friendliness, and then germinate the willingness to participate in DE (Su et al., 2019) [7]. To further explore the influence mechanism of DL, we construct a mediated-moderating model, aiming to investigate whether the moderating effects of DL can be generated through the mediating factors of FI, thus affecting the motivation and behavior of DE. Specifically, we believe that the indirect effect of FL on DE through FI depends on the strength of DL. The stronger the DL, the greater the positive impact of FL on FI, which in turn leads to a stronger mediating role of FI between FL and DE. Based on the above analysis, we propose hypothesis H6.

Hypothesis 6 (H6):

The moderating effect of DL plays a moderated-mediating role through the mediating effect of FI.

Few pieces of literature categorize and discuss the effect of FL and digital entrepreneurial intention and behavior by including macro background and unique micro family background of different social development levels into the research framework. However, the comparative analysis of individual consciousness and behavior based on differences in the social environment and family background is more conducive to discovering the details of the functional relationship between variables in the theoretical framework and producing additional valuable findings. Therefore, this study will also conduct an in-depth analysis of the theoretical framework constructed in these two dimensions to establish a strong theoretical basis. In this regard, Figure 1 shows a research model summarizing all the hypotheses.

3. Research Methods and Variable Measurement

3.1. Data Collection and Sample

Sutter et al., (2019) [5] noted that the threshold of start-up capital and resources needed for DE is not high at the beginning, something which is relatively in line with the entrepreneurship conditions for the BOP group. Moreover, participating in DE is different from setting up a digital company or project. Generally speaking, the initial scale is relatively small, with lower requirements for initial investment (e.g., setting up an online store on an e-commerce platform). Many participating digital start-ups begin by using the scattered time in the main work, but the requirements for collection time are higher. Compared with urban residents, the form of villagers’ labor (work) is relatively unitary, mainly engaged in agricultural production (mainly planting, followed by small-scale farming type). Farmers’ work also shows strong seasonal characteristics due to the significant influence of the agricultural planting season. They are less abundant in work than workers in peri-urban and small towns but more abundant in leisure and spare time. In terms of China’s actual situation, villagers rarely establish digital companies and platforms. In most cases, they start businesses based on existing digital platforms (participating in DE) because most villagers lack the capital and digital ability to establish digital enterprises. At least in China’s context, villagers’ DE is the most common and typical way to participate in DE. They mainly use their spare time and use digital platforms (mostly commercial platforms) to better display and sell their agricultural products or small handmade products.

Moreover, farmers’ participation in DE is also a form of entrepreneurship strongly advocated by the state. Under the initiative of the government, almost all of China’s more than 50,000 rural towns have established e-commerce service outlets (stores), and some online stores even reach villages and groups (Wang et al., 2020) [24]. The primary role of these online service stores is to help villagers who have difficulties using digital equipment to conduct online transactions or improve the villagers’ DL and train their DE ability. Since 2015, the Chinese State Council, the Ministry of Agriculture, the Ministry of Commerce, and the State Administration for Market Regulation have issued several policy documents aiming at fostering entrepreneurship. This action developed the digital environment, expanded the application of digital in rural areas, and promoted the development of rural e-commerce (Referring to “Opinions on Vigorously Developing E-commerce and Accelerating the Cultivation of New Economic Impetus” (Chinese Government [2015] No. 24), “Guiding Opinions on Actively Promoting the ‘Internet + ‘Action” (Chinese Government [2015] No. 40)). Furthermore, some poor mountainous areas have promoted the improvement of villagers’ digital entrepreneurial ability as an essential work of poverty alleviation. From the practical results, these policies have played a relatively prominent role in poverty alleviation. In some non-poor rural areas, DE has improved villagers’ income levels (Wang et al., 2020; Mei and Jiang, 2020) [24,55]. Therefore, the villager group’s research has good practical significance, theoretical significance, and sample representativeness.

Based on this consideration, we organized a research team to conduct interviews and questionnaires. The survey method and process were based on a list of poor counties provided by the agricultural department, and poverty-stricken counties and non-poverty-stricken counties were selected. One township was selected in each county, and then, the administrative village with the most mature e-commerce development was selected. In each administrative village, the village committee was invited to recommend 10–20 farmers (mainly engaged in agricultural products ascending) who had cooperated with the e-commerce service outlets for research activities. This survey aims to clarify the results of big data penetration in rural areas. To this end, the questionnaire includes information of the county (poverty stricken and non-poverty-stricken counties), local broadband development level, per capita income level, villagers’ sense of economic gain, farmers’ entrepreneurial orientation, FL level, DE level, farmers’ level of fertilizer and drug use, the output income level of main field crops and cash crops, and others. A total of 2200 villagers engaged in e-commerce entrepreneurship were collected through the survey. From this, 664 sets of data consistent with the objective of this study were extracted for empirical research. The sample area’s per capita income level is 15,673 RMB (about 2329 dollars), which is slightly lower than the national average. In this study, the T-test method was used to find that no significant difference exists between the samples and the national quality inspection. This result indicates that the survey samples had good representativeness.

Table 1. Statistics of survey samples.

Variable	Classification	Sample Size	Percentage
Gender	Male	468	70.48%
	Female	196	29.52%
Age	<30	292	43.98%
	31–40	215	32.38%
	41–50	138	20.78%
	>51	19	2.86%
Education background	Primary schools and below	265	39.91%
	Junior high school	359	54.07%
	High school	33	4.97%
	Bachelor and above	7	1.05%
Per-capita income level (year)	<5000 yuan	206	31.02%
	5001–10,000 yuan	164	24.70%
	10,001–20,000 yuan	252	37.95%
	>20,001 yuan	42	6.33%
Average time for migrant workers per year	<1 month	176	26.50%
	1–3 months	312	46.99%
	3–6 months	148	22.29%
	>6 months	28	4.22%
Broadband development levels	1–6 M	43	6.48%
	6–10 M	80	12.05%
	10–20 M	62	9.34%
	>20 M	468	70.48%
	Unable to answer	11	1.66%
County level of development	Poverty-stricken counties	332	50.00%
	Non-Poverty-stricken counties	332	50.00%
Family Background (Social Capital)	Relatively excellent	363	54.67%
	Relatively weak	301	45.33%

shows the sample details.

3.2. Measurement

The measurement tools used in this study refer to maturity scales and relevant recommendations. We also adopted a two-way translation program for the English version of the questions (Brislin, 1970) [56]. We optimized the text according to the specific usage scenarios to ensure that the respondents could fully understand and answer the questions accurately. All questions were scored by Likert’s five-point scale, ranging from strongly disagree to strongly agree. In this study, Mplus 8.0 (Mplus is a statistical modeling software developed by Swede Linda Muthén & Bengt Muthén)and SPSS 22.0 (SPSS is a statistical modeling software developed by Americans Norman H. Nie, C. Hadlai ( Tex ) Hull and Dale H. Bent) were used to calculate the detection indexes of latent variables and observed variables, and

Table 2. Results of variable definition and factor analysis.

	Variable Name	Variable Measure	Factor Loading
Independent variable	Financial Literacy	1. I have a good knowledge of passbook, bank card, credit card, online banking, bank protection or financial products, and gold business.	0.707
		2. I have a clear and accurate understanding of financial returns and financial risks.	0.723
		3. Every year, I plan the proportion of money I will spend on consumption, saving, or investment.	0.850
		4. I can adjust my savings plan for changes in interest rates.	0.826
		5. I can use a legal weapon to protect financial rights and interests.	0.615
Mediator	Financial Inclusion	1. I feel that the number and types of financial institutions have increased in the past three years.	0.632
		2. It is more convenient for me to go to the financial institutions nearby.	0.789
		3. I think it’s easier to get bank approval for loans now.	0.639
		4. I think the interest rate is moderate, and we can afford it.	0.741
		5. My family has begun to attach importance to buying commercial insurance.	0.798
Moderator	Digital Literacy	1. I can use network equipment well and work on the network.	0.814
		2. I know how to conduct network broadcasts and am familiar with e-commerce sales.	0.888
		3. I can enjoy the convenience brought by digitalization and enjoy it.	0.889
		4. I can make fair use of mobile phone APP, through which I have made many good friends and obtained a lot of resources, which was unimaginable before.	0.574
		5. I think the current network and smartphones are easy to operate and safe, and I can make all kinds of transactions with confidence.	0.770
Dependent variable	Digital entrepreneurial willingness and behavior	1. I can use the digital platform to sell newly planted crops.	0.774
		2. I think I can sell my agricultural products at a good price through new methods such as live broadcasting and online sales, which can increase my income.	0.676
		3. For products with high risk but higher value, I prefer to plant or produce them because the sales can be guaranteed.	0.854
		4. Generally speaking, to make more money, I feel it is necessary to take risks to do something new because the big data environment has become more conducive to innovation.	0.871
		5. In the past five years, I have introduced new technologies to make the agricultural products or small handmade products different from the ones I sold before because I can sell them online.	0.895
		6. Through direct contact with consumers, my produce has created its own brand and generated more revenue.	0.841

shows the results in detail.

The explained (dependent) variable of this study is DE willingness and behavior. By referring to the definition, connotation, and description of DE by Nambisan (2017) [3], combined with Covin’s (1989) [57] design of questions related to entrepreneurial willingness and behavior in the “entrepreneurial-orientation” scale, the degree of “digital entrepreneurial willingness and behavior” was tested through six questions determined by the pretest. Examples include “I can use digital platforms to sell newly grown crops.” and “In the past five years, I have introduced new technologies to make agricultural products or small handmade products different from what they used to be because I can sell them online.” The Cronbach’s α of the scale is 0.922.

The explanatory (independent) variable of this study was FL. We partly referred to the interpretation and measurement methods of FL in the China Household Finance Survey (CHFS), which is a nationwide sample project conducted by the China Household Finance Survey and Research Center, aiming at collecting information about household finance at the micro-level. The main contents include housing assets and financial wealth, liabilities and credit constraints, income and consumption, social security and insurance, intergenerational transfer payments, demographic characteristics and employment, payment habits, and other relevant information. Then, we introduced the items designed by Zhang and Xiong (2017) [6], while using the processing techniques of Lynch and Netemeyer (2014) [58] for reference. Moreover, we determined the measurement items containing five dimensions through the test. Specifically, the dimensions were the understanding and application of financial knowledge, the cognition of financial risks and returns, financial planning awareness, financial product selection ability, and financial responsibility awareness. Questions included “I have a good understanding of the business including passbook, bank card, credit card, online banking, bank insurance or financial products, and gold” and “I plan the proportion of consumption, savings or investment every year”. The Cronbach’s α of the scale is 0.854.

The mediating variable (mediator) of this study was the sense of FI. We also referred to the survey design on FI in CHFS, mainly referring to the measurement method of Wang and Chen (2016) [38]. This study also divides the measurement scale into three dimensions and finally determines five items. Specific questions included “I feel that the number and types of financial institutions have increased in the past three years”, “It is more convenient for me to go to a nearby financial institution”, “I think it is easier to get bank approval for loans now”, “I think the interest rate of the bank is relatively moderate, which we can afford”, and others. The Cronbach’s α of the scale is 0.809.

The moderating variable (moderator) of this study was DL. Mainly referring to the measurement methods of Hargittai (2005) [59] and Ng (2012) [60], five measurement items were tested and determined. Specific questions included “I can use network devices and work with the network” and “I can use various mobile phone apps well, through which I have made many friends and obtained a lot of resources, which was unimaginable before”. The Cronbach’s α of the scale is 0.882.

Based on Alderete (2015) [32], He and Li (2019) [11], and Fan and Ying (2020) [61], this article used gender, age, education level, per capita income level, average annual migrant time, broadband, and traffic development level as control variables. The objective was to control the study on the characteristics of the investigated population and eliminate the interference of other factors. This study also aimed to compare the influence of different counties (poor and non-poor counties) and villagers’ family background factors on the research model, so the sample group was divided and investigated. Concerning the measurement indicators, the measurement items and factor loads of all variables in this study meet the basic requirements of statistics.

3.3. Reliability and Validity Analysis

Firstly, each scale’s reliability was tested by calculating the composite reliability (CR) of the scale. The results show that the CR values of FL, FI, DL, and DE willingness and behavior were 0.863, 0.845, 0.894, and 0.925, respectively, which are more significant than the critical point of 0.7 (Fornell and Larcker, 1981) [62]. Simultaneously, the internal consistency coefficient (Cronbach’s α) scales were 0.854, 0.809, 0.882, and 0.922, respectively, which are all more significant than the minimum standard of 0.7, indicating that the scales used have good reliability. Second, in this study, the questionnaire’s convergence validity was tested by average variance extracted (AVE), and the discriminant validity was tested by determining whether the latent variable’s square root of AVE is more exceeded than the correlation coefficient between latent variables. The results show that the latent variable’s AVE values were 0.561, 0.523, 0.633, and 0.675, respectively, which were all greater than the critical value of 0.5, indicating a good convergence validity of the scale. Moreover, the square root of AVE of the latent variable on the diagonal is greater than the direct correlation coefficient of each latent variable located in the same row or column. This result indicates that the discriminant validity of the scale also meets the statistical requirements.

Table 3. Reliability, convergent validity, and discriminant validity analysis.

Construct	Composite Reliability	Convergent Validity	Discriminant Validity
	CR	AVE	FL	FI	DL	DE
FL	0.863	0.561	0.749
FI	0.845	0.523	0.636	0.723
DL	0.894	0.633	0.446	0.536	0.796
DE	0.925	0.675	0.605	0.506	0.596	0.822

Note: The bold entries in the main diagonal are the square root of AVE.

shows the analysis results of reliability, convergent validity, and discriminant validity.

3.4. Descriptive Statistical and Correlation Analysis

Table 4. Descriptive statistics and correlation analysis.

	Mean	S.E	1.	2.	3.	4.	5.	6.	7.	8.	9.	10.
1. Gender	1.295	0.456	–
2. Age	1.825	0.856	0.024	–
3. Education	1.672	0.619	0.045	−0.285 **	–
4. Per capita income level	2.804	0.952	0.008	0.656 **	−0.155 **	–
5. Average time of migrant workers per year	3.221	1.183	0.108 **	0.074	−0.067	0.037	–
6. Broadband development levels	1.450	0.948	−0.147 **	0.235 **	0.206 **	0.310 **	−0.071	–
7. FL	1.502	0.649	0.014	−0.074	0.104 **	−0.071	0.015	−0.016	(0.854)
8. FI	2.194	0.782	0.027	0.008	0.041	0.084 *	−0.001	−0.02	0.544 **	(0.809)
9. DL	1.968	0.760	0.113 **	−0.042	0.012	−0.022	0.068	−0.108 **	0.401 **	0.472 **	(0.882)
10. DE	1.469	0.588	0.046	−0.101 **	0.127 **	−0.080 *	0.014	0.008	0.552 **	0.458 **	0.575 **	(0.922)

Note: * p < 0.05, ** p < 0.01. Bold entries in parentheses are the corresponding variable Cronbach’s α values.

shows the mean value, standard deviation, and correlation coefficient of this study’s main variables. Among these, gender, age, education level, per capita income level, average time for migrant workers per year, and broadband development level were the control variables. Variables show the following characteristics: first, the group participating in DE is still dominated by males, and in age, it is mainly concentrated around 40 years old. Second, the education level is generally low and mainly concentrated in junior middle school and below. Third, the per capita income level can reach the Chinese average level (in rural), but the income gap between different villagers is large (standard deviation = 0.952). Fourth, the average time of migrant workers is generally concentrated in about three months, and a massive difference is observed (standard deviation = 1.183). Fifth, most of the home broadband levels can meet business’s needs, reaching more than 20 M. Sixth, the levels of FL, DE willingness and behavior among the four main measured variables are relatively low; the FI level is the highest, followed by the DL level, but they are generally not high.

From the correlation results between variables, the following conclusions can be drawn from Table 4. First, FL is significantly positively correlated with DE willingness and behavior (β = 0.552, p < 0.05). Second, FL is significantly positively correlated with DE willingness and behavior (β = 0.552, p < 0.05). Third, FL was significantly positively correlated with FI (β = 0.544, p < 0.05). Fourth, the sense of FI was also significantly positively correlated with DE willingness and behavior (β = 0.458, p < 0.05). Descriptive statistical results provide a necessary premise for further verification of the relationship between variables.

3.5. Common-Method Variance

In this study, all variables were obtained by villagers’ self-assessment, which may lead to the problem of common-method variance (CMV) among constructs. Therefore, two program control methods and statistical controls were adopted in this study to solve this potential effect. First, measures, such as questionnaire matching survey, anonymous response, reverse question design, and two-stage data collection, can reduce this influence to some extent. Second, the Harman single-factor test was used to test the CMV. After the unrotated exploratory factor analysis of all the latent variables, the first principal component factor’s variance interpretation rate was found to be 32.17%, less than the 50% standard suggested by Hair (2013) [63]. Therefore, the problem of CMV in this study is not severe.

4. Results

4.1. Main Effect Analysis

In this study, the hypothesis was verified through stepwise linear regression.

Table 5. The results of the main effect analysis.

Paths and Models		M1	M2	M3
		FL→DE	FL→FI	FI→DE
Path Coefficient	FL→FI		0.640 ***
	FI→DE			0.507 ***
	FL→DE	0.596 ***
Degree of Model Fit Indicators	X2/DF	2.183	1.735	1.841
	CFI	0.974	0.976	0.979
	TLI	0.970	0.972	0.975
	RMSEA	0.042	0.033	0.036
	SRMR	0.037	0.040	0.031

Note: *** p < 0.001.

shows the regression analysis results after controlling six variables, such as gender, age, education level, and per capita income level. Among these, M1 and M3 take DE willingness and behavior as dependent variables to examine the influence of FL and FI, respectively. M2 takes FI as the dependent variable and tests the impact of FL.

Table 5 shows the following: (1) the path coefficient of FL’s influence on DE willingness and behavior (FL→DE) is 0.596 (p < 0.001). This finding indicates that high FL promotes DE willingness and behavior, which supports H1. (2) The influence path coefficient of FL on FI (FL→FI) is 0.640 (p < 0.001), indicating that FL promotes FI. The test results are consistent with Wagner et al., (2018) [43], and H2 is verified. (3) The influence path coefficient of FI on DE willingness and behavior (FI→DE) is 0.507 (p < 0.001), indicating that FI promotes the willingness and behavior of DE, which supports H3.

4.2. Mediating Effect Analysis

In this study, the Bootstrap method was used to test the mediating effect of FI (FL→FI→DE), and

Table 6. The results of the mediating effect analysis.

Path	Effect	Coefficient	S.E.	Z	Bootstrapping
					Bias-Corrected 95% CI		Percentile 95% CI
					Lower	Upper	Lower	Upper
M4: FL→FI→DE	Total effect	0.597 ***	0.056	10.721	0.484	0.698	0.477	0.695
	Indirect effect	0.141 ***	0.036	3.899	0.069	0.215	0.067	0.212
	Direct effect	0.456 ***	0.076	6.019	0.305	0.603	0.297	0.595

Note: *** p < 0.001. The sample size of Bootstrap was 1000.

shows the analysis results. The results of the Bootstrap analysis of 1000 repeated samples show that within the confidence interval without zero (95% deviation correction CI between (0.069 and 0.215) and percentile 95% CI between (0.067 and 0.212), the mediating effect of FI (β = 0.141, SE = 0.036) was positively significant. This result indicates that FI plays a partial mediating role in the positive influence of FL on DE willingness and behavior. Thus, hypothesis H4 is verified.

4.3. Moderating Effect Analysis

Hierarchical regression was used to examine the moderating effect of DL. The latent variables were firstly centralized to avoid the adverse effects of multicollinearity, and then, the interaction terms were obtained on this basis.

Table 7. The results of the moderating effect analysis.

Variable	DE Willingness and Behavior
	M5	M6	M7
Control variables
Gender	0.044	0.002	0.010
Age	−0.058	−0.033	−0.058
Degree of education	0.101 **	0.054 *	0.044
Per capita income level	−0.033	−0.035	−0.008
Average time for migrant workers per year	0.023	−0.009	−0.024
Broadband development levels	0.019	0.068 **	0.048
Independent variable: FL		0.370 ***	0.303 ***
Moderator variable: DL		0.431 ***	0.414 ***
Interaction item: FL × DL			0.198 ***
F	2.687	71.467	72.181
R2	0.024	0.466	0.498
∆R2	0.015	0.460	0.491

Note: * p < 0.05, ** p < 0.01, *** p < 0.001.

shows the test results. Moreover, Table 7 shows that the interaction term between FL and DL has an influence coefficient of 0.198 (p < 0.001) on digital entrepreneurial willingness and behavior. This result indicates that DL positively moderates the influence of FL on digital entrepreneurial willingness and behavior. To further explain this moderating effect, according to the Simple Slope test recommended by Aiken and West (1991) [64], we drew the moderating effect analysis diagrams, as shown in Figure 2. The results show the following: at a low DL level, the influence curve of FL on DE willingness and behavior is relatively flat. By contrast, at a high level of DL, the impact curve of FL on DE willingness and behavior is steep. This finding indicates that the group with higher DL has a more prominent role in promoting DE’s willingness and behavior.

4.4. Moderated-Mediating Effect Analysis

Based on Ye and Wen’s (2013) [65] practice of testing the mediated-moderating effect model, the present study verified whether DL’s moderating effect on FL and digital entrepreneurial willingness and behavior could be realized through the mediating effect of FI. The specific steps are as follows:

DE = ɑ₀ + ɑ₁FL + ɑ₂DL + ɑ₃(FL × DL) + e₁,

(1)

FI = β₀ + β₁FL + β₂DL + β₃(FL × DL) + e₂,

(2)

DE = ɡ₀ + ɡ₁FL + ɡ₂DL + ɡ₃FI + ɡ₄(FL × DL)+ɡ₅(FI × DL) + e₃,

(3)

If the coefficient ɑ₃ of regression in Equation (1) is significant, then a foundation is laid for subsequent inspection with the mediation effect’s adjustment. Furthermore, the significance of regression coefficients β₃ and ɡ₃, β₃ and ɡ₅, β₁, and ɡ₅ needs to be examined. If any of the above sets of the regression coefficient are significant, then the impact of FL on digital entrepreneurial willingness and behavior is at least partially realized through the mediator of FI. Finally, the regression coefficient ɡ₄ of Equation (3) (FL and DL interaction’s coefficient) needs to be tested. If the regression coefficient ɡ₄ is not significant, then the moderating effect is fully mediated. Otherwise, a partial mediated-moderating effect exists when ɡ₄ is significant.

Table 8. Analysis results of moderated-mediating effect.

Variable	DE Willingness and Behavior	Sense of FI	DE Willingness and Behavior
	M8	M9	M10
Control variables
Gender	0.013	−0.032	0.016
Age	−0.04 *	−0.025	−0.037
Degree of education	0.041	0.018	0.040
Per capita income level	−0.005	0.12 ***	−0.016
Average time for migrant workers per year	−0.012	−0.016	−0.010
Broadband development levels	0.030 *	−0.015	0.031 *
Independent variable: FL	0.275 ***	0.555 ***	0.228 ***
Moderator variable: DL	0.320 ***	0.318 ***	0.290 ***
Interaction item: FL× DL	0.146 ***	0.088 ***	0.145 **
Mediating variable: sense of FI			0.088 ***
Interaction item: FI × DL			0.014
F	72.181	47.805	60.831
R2	0.498	0.397	0.506
∆R2	0.491	0.389	0.498

Note: * p < 0.05, ** p < 0.01, *** p < 0.001.

shows the specific analysis results.

Table 8 shows that after controlling for demographic variables, such as gender, age, and education level, Equation (1) suggests that the interaction of FL and DL plays a significant role in predicting entrepreneurial willingness and behavior (ɑ₃ = 0.146, p < 0.05) (M8). This result indicates that DL plays a moderating role between FL and digital entrepreneurial willingness and behavior (H5). This result provides a basis for the subsequent tests.

Furthermore, M9 in Table 8 (calculated by Equation (2)) shows that the interaction term between FL and DL has a significant positive effect on FI (β₃ = 0.114, p < 0.05). Therefore, DL positively moderates the positive relationship between FL and FI. Results show that in model M10 (calculated by Equation (3)), after controlling for FL, DL, the interaction term of FL and DL, the interaction term of FI and DL and the control variables, the mediator FI exerts a significantly positive influence on digital entrepreneurial willingness and behavior (ɡ₃ = 0.088, p < 0.05). Ye and Wen (2013) [65] proposed that if the regression coefficients β₃ and ɡ₃ are significant, then the mediated moderating model is established. Moreover, the coefficient of DL and FL’s interaction term ɡ₄ (ɡ₄ = 0.145, p < 0.05) is significant in Equation (3). This result indicates that the moderated-mediating effect is partial mediation, thereby verifying H6.

4.5. Analysis of PSC vs. NPSC

Bruton et al., (2013) [1] believed that entrepreneurship, as a high-risk business activity, brings high-yield results that can improve individual economic and non-economic benefits, and, in addition, have a positive impact on the living standards of the poor. However, some scholars have pointed out that the poverty reduction effect of entrepreneurship varies from system to social environment, especially in the case of differences in regional economic circles. In reality, due to their lack of FL, it is difficult for farmers in poor areas to form large-scale entrepreneurship. Based on this, this study compared the difference between FL’s influence on digital entrepreneurial willingness and behavior in PSC and NPSC. The objective was to further explore the internal mechanism between FL and DE in different social and economic macro environments, specifically when differences among regional economic circles exist.

Table 9. Results of poverty-stricken counties vs. non-poverty-stricken counties.

Model and Path	PSC (β)	NPSC (β)	Differences
FL→FI	0.644 ***	0.537 ***	0.107 **
FI→DE	0.556 ***	0.707 ***	−0.151 **
FL→DE	0.531 ***	0.460 ***	0.071 *
FL→FI→DE	0.137 ***	0.124 ***	0.013 **
DL: Moderating effect	0.427 ***	0.781 ***	−0.354 **
DL: Moderated-Mediating effect	0.377 *	0.724 ***	−0.347 **

Note: * p < 0.05, ** p < 0.01, *** p < 0.001. The inter-group difference test refers to the correction deviation test based on 2000 boots.

shows the comparative analysis results.

According to the results, we found the following: first, among PSC (β = 0.644, p < 0.05) and NPSC (β = 0.537, p < 0.05), FL can significantly promote the willingness and behavior of DE, and there is a significant difference between the two influence coefficients (∆β = 0.107; p < 0.05). Thus, the robustness of hypothesis H1 was further confirmed. Compared with the non-poor counties, the villagers’ FL in developing countries plays a vital role in promoting DE willingness and behavior. Second, the influence coefficient of FL on FI of villagers in PSC is 0.556 (p < 0.05), whereas that of villagers in NPSC is 0.707 (p < 0.05), the significant difference between the two coefficients is −0.151 (p < 0.05), verifying the robustness of H2. The empirical data analysis results also show that villagers’ FL in PSC has a more substantial effect on promoting the sense of FI than that in NPSC. Third, from the perspective of the influence of FI on DE willingness and behavior, the promotion effect of poor and non-poor counties is positively significant, with impact coefficients of 0.531 (p < 0.05) and 0.460 (p < 0.05), respectively. The difference between them is very small (∆β = 0.071; p < 0.1), indicating the robustness of H3. Although the difference between the two is not significant, the promotion effect was more substantial for villagers in PSC. Fourth, in NPSC, the mediating effect of villagers’ sense of FI is significant, and the influence coefficients of the two are 0.137 (p < 0.05) and 0.124 (p < 0.05) respectively, and there is a significant difference between them (∆β = 0.013; p < 0.01), which confirms the robustness of H4. Fifth, based on the comparison of the moderating effect of DL on FL and DE willingness and behavior, the moderating effect of DL of villagers in PSC is lower than that of NPSC, which is 0.427 (p < 0.05) and 0.781 (p < 0.05), respectively. The difference value between the two is 0.354 (p < 0.05), which confirms the robustness of H5. Sixth, in terms of the moderated-mediating effect of DL, PSC (β = 0.377, p < 0.05) are weaker than NPSC (β = 0.724, p < 0.05), with a significant difference of 0.347 (p < 0.05). This result proves the robustness of H6.

4.6. Comparison of Family Background Differences among Entrepreneurial Groups

Studies have shown that rich social capital can make more entrepreneurial resources in rural areas, thus promoting the farmers’ entrepreneurial intention. First of all, the higher the family support, the more conducive to their entrepreneurial confidence (Yu et al., 2015) [66]. Secondly, the more social networks, the more information and experience can be mastered, which can improve the alertness of farmers’ DE (Yang et al., 2017) [67]. Thirdly, having close connections is conducive to identifying farmers’ entrepreneurial opportunities. To explore the internal mechanism between FL and DE under different social capital levels, this paper further compares the differences in the impact of FL on DE behavior under high social capital and low social capital samples.

In most studies, family background is divided into three dimensions, namely, economic capital, cultural capital, and social capital (Tong, 2020) [68]. For this study, economic capital (household income level) and cultural capital (education level) were more suitable for control variables to discuss the basic model. Therefore, we applied social capital to represent the family backgrounds of different entrepreneurial individuals. Referring to Li’s (2013) [69] classification of social capital and the corresponding measurement method, this research mainly used the government network, business network, and total degree of family education to measure the social capital and family background of entrepreneur group, ultimately coming down to a high and low level (to save space, the specific formular is not listed). After controlling the control variables, we compared and analyzed the two sample groups corresponding to the two levels (

Table 10. Results of family background comparison among entrepreneurial groups.

Model and Path	Family Background (Superior) High Level of Social Capital	Family Background (Inferior) Low Level of Social Capital	Comparison of Differences
FL→DE	0.649 ***	0.550 ***	0.099 **
FL→FI	0.682 ***	0.607 ***	0.075 *
FI→DE	0.556 ***	0.451 ***	0.105 **
FL→FI→DE	0.162 ***	0.117 ***	0.045 **
DL: Moderating effect	0.555 ***	0.678 ***	−0.123 **
DL: Moderated-Mediating effect	0.530 **	0.680 ***	−0.150 **

Note: * p < 0.05, ** p < 0.01, *** p < 0.001. The inter-group difference test refers to the correction deviation test based on 2000 boots.

) and reached the following conclusions.

First, in terms of the effect of FL on the promotion of digital entrepreneurial willingness and behavior, the influencing coefficients of groups with high and low levels of social capital were 0.649 (p < 0.05) and 0.550 (p < 0.05) respectively, with a difference value of 0.099 (p < 0.05). This result shows that if the family background of villagers is superior (a higher level of social capital), then the digital entrepreneurial willingness and behavior will be positive. The results prove the robustness of H1 again. Second, regardless of whether the family background is superior, FL has a significant positive effect on the promotion of FI, with influence coefficients of 0.682 (p < 0.05) and 0.607 (p < 0.05), respectively. The significant difference between the two coefficients is 0.075 (p < 0.1), which verifies the robustness of H2. Third, regardless of whether the villagers’ family background is superior, a sense of FI can significantly promote digital entrepreneurial willingness and behavior, with influence coefficients of 0.556 (p < 0.05) and 0.451 (p < 0.05), respectively. The significant difference between the two coefficients is 0.105 (p < 0.05), which confirms the robustness of H3. The result of difference analysis shows that villagers with better family backgrounds have a stronger promotion effect of FI on their digital entrepreneurial willingness and behavior. Fourth, although differences in the degree of superiority of family background exist, the mediating effect of FI is almost the same in the two sample groups, with influence coefficients of 0.162 (p < 0.05) and 0.117 (p < 0.05), respectively. The significant difference between the two coefficients is 0.045 (p < 0.05), manifesting the robustness of H4. Fifth, the moderating effects of DL are significant under different family backgrounds, which confirms the robustness of H5. However, differences exist between the two, that is, for villagers with superior family background, the influence coefficient of DL’s moderating effect is 0.555 (p < 0.05). On the contrary, for villagers with weak family backgrounds, the influence coefficient of the moderating effect is 0.678 (p < 0.05). The moderating effect of DL is stronger for villagers with inferior family background than for those with weak family background. This finding also reflects the way in which, when villagers have insufficient families (weaker social relations), DL (interaction with FL) can strongly promote their entrepreneurial willingness and behavior. Sixth, the mediated-moderating effect of DL plays a significant role in different family backgrounds, which indicates the robustness of H6. Similar to the moderating effect of DL, some differences in their effects exist, that is, the mediated-moderating effect of DL of villagers with a superior family background is lower than that of villagers with weak family backgrounds. The influence coefficients of the two are 0.530 (p < 0.05) and 0.680 (p < 0.05), respectively, and there is significant difference between them (∆β = −0.150; p < 0.05), which verifies the robustness of H6.

Based on the conclusions above, this study summarizes the role of FL and DL in the county economic development level and entrepreneurial group family background (divided into four regions: high–weak, high–strong, low–weak, and low–strong, as shown in Figure 3). Among these, the mediating role of FI is the strongest in high–strong regions, whereas the moderating role of DL is strongest in the low–weak regions. The effects of the two are different in other sample combinations.

5. Discussion and Conclusions

5.1. Main Conclusions

Taking 664 villagers who have participated in DE as samples, this study investigated the effect of FL on the willingness and behavior of DE from the perspectives of behavioral finance and DE, including the mediating role of FI and the moderating mechanism of DL. The results show the following: (1) FL has a substantial effect on the willingness and behavior of DE. (2) The sense of FI plays a partial mediating role in the influence of FL on the willingness and behavior of DE. (3) DL can positively moderate the influence of FL on digital entrepreneurial willingness and behavior. (4) The moderating effect of DL can also play a role through the mediator FI, thus promoting the influence of FL on the willingness and behavior of DE. (5) The influence of FL on the willingness and behavior of DE differs between poverty stricken and non-poverty-stricken counties, including families participating in entrepreneurship. In addition, differences exist among the mediating effect of FI and the moderating and mediated-moderating effects of DL. This study has contributed to theories related to FL, DE, FI, and DL, specifically expanding the academic understanding of the relationship among the four.

5.2. Theoretical Contribution and Implications

The marginal contribution of this study is mainly reflected in five aspects. First, the study distinguishes two types of DE. One is to start a digital enterprise (institution) to develop and operate digital products and services. The second is to participate in DE, that is, to start businesses around the products and services of digital enterprises. Previous studies have mainly discussed issues related to the first type of DE (e.g., Yu et al., 2018; Duan and Li, 2020) [70]. On the contrary, research on the second type is insufficient. This study enriches the research content of DE of “double low” groups by investigating related issues of DE.

Second, this study hypothesizes and verifies the promoting effect of FL on DE willingness and behavior and expands the research on the antecedents of DE. Although existing studies based on entrepreneurship theory mainly discuss the influence of financial policy, environment, system, geography, and other factors on entrepreneurial behavior (Sigfusson, 2013; Mainela et al., 2014) [71,72], the studies of FL’s effect on entrepreneurial willingness and behavior are limited. The results of this study show that with social progress, specifically the continuous diffusion and consolidation of financial knowledge, skills, and tools to society, people’s FL have also changed (mainly improved), which has a substantial effect on the willingness and behavior of DE.

Third, this study hypothesizes and verifies that FI partially mediates the influence between FL and digital entrepreneurial willingness and behavior and enriches the research on the mechanism of FI. Previous literature mainly took FI as an independent or dependent variable (Fernandes et al., 2014) [58] and discussed its mediating effect less.

Fourth, the influence of digitalization (particularly in China) is widespread, significantly affecting people’s DL and changing their understanding of mobile devices, computers, and the internet. Existing literature mainly discusses the positive influence of DL on entrepreneurship orientation (Oggero, 2019; Islami et al., 2019) [53,54], specifically emphasizing the promoting effect of DL on the establishment of digital enterprises but has rarely examined the effect of DL on participation in DE. This study finds that with the continuous improvement of the convenience of software and equipment application, people’s DL is no longer an obstacle to participating in DE, but more of a moderating role.

Fifth, this study compared and analyzed the samples of poverty-stricken and non-poverty-stricken counties, as well as entrepreneurial groups with various family backgrounds. The influence mechanism of FL on digital entrepreneurial willingness and behavior was further explored through the grouping examination of the research samples, and the robustness test of the basic model is also carried out.

The findings above are rarely involved in previous studies and have certain contributions to similar theoretical models and variable relations. Moreover, this research has some enlightenment and guiding significance to management practice. The results of data analysis show that FL has a substantial effect on people’s willingness and behavior to participate in DE, specifically for the BOP. Vigorously popularized financial knowledge (particularly FI) and improved FL is one of the prerequisites for promoting DE willingness and behavior. Therefore, first, the government and financial institutions should jointly strengthen the financial knowledge education of the BOP. Second, the mediation and transmission of FI are strong. People’s sense of FI should be further enhanced, and full consideration should be given to the cost (affordability), convenience, and availability of loans. Third, digital technologies, platforms, and facilities should be actively used to provide more policy support for people to improve their opportunities to participate in DE, promote the entrepreneurial motivation of the BOP groups, and improve their success rate of entrepreneurship. Finally, in terms of improving people’s entrepreneurial motivation, policymakers should consider the priorities of policy implementation from two aspects, namely, the region where the group is located and the family background. For example, in relatively affluent regions, the sense of FI should be improved for groups with superior family backgrounds (Tong, 2020) [68]. For the poor areas, we should start to improve the DL of the group with weak family backgrounds. The effects of policies can be better utilized through such measures.

5.3. Research Limitations and Future Prospects

This study also has some limitations. First, the sample collection of this study is mainly from western villages, which fails to take the overall picture of China’s rural areas into account. Hence, the research results are more a reflection of the western regions’ experience. The regional cultural (educational) and economic and social development levels differ greatly due to China’s vast territory and numerous ethnic groups. The empirical framework obtained needs to be further improved, enriched, and developed. Second, although farmers are a representative part of the BOP groups, a large number of BOP groups still exist in and around cities. Nevertheless, these samples have not been included in the study due to the difficulty of obtaining a wider range of data for this study. Future studies should conduct empirical analysis, comparison, and discussion of these groups. Finally, with the continuous development of information technology, commercialization (digital commerce) led by big data is likely to penetrate every corner of people’s social life worldwide. Therefore, DE, specifically large-scale participation in DE, is bound to become a long-term topic worthy of attention in the future. In addition to FL, DL, FI, and other variables, some potential factors may also affect participation in DE. These problems need to be further discussed with more practical research methods, data, and cases.