Impact Paths of the Entrepreneurial Behavior of the Underclass Groups’ Involved in Urbanization: A Case Study of Zhejiang Province, China

Abstract

The aim of this paper is to clarify the influence mechanism and role paths of the entrepreneurial behavior of the underclass groups (EBUG) involved in urbanization from a microcosmic perspective and propose sustainable development paths for the transition of underclass groups’ entrepreneurship from the subsistence type to the opportunistic type. Based on the theories of planned behavior, the entrepreneurial event model, and social cognitive theory, this study constructs a theoretical framework of “intention–situation–behavior” of the EBUG involved in urbanization. Through a questionnaire survey conducted in three major urban agglomerations in Zhejiang Province, the theoretical model is validated by using structural equation modeling (SEM). On the one hand, perceived desirability, perceived feasibility, and land expropriation all have a significant positive influence on entrepreneurial behavior. On the other hand, land expropriation has a significant moderating effect on entrepreneurial intentions and behaviors. When the moderating role of land expropriation is not considered, underclass groups are more likely to engage in opportunistic-type entrepreneurship, which is primarily driven by perceived desirability such as achievement motivation and innovation orientation. In contrast, when land expropriation is considered, these groups tend to focus on survival entrepreneurship, which is mainly influenced by perceived feasibility factors such as social capital and market opportunities. The future survival and development of underclass groups is contingent upon urbanization, with the potential to influence the stability and sustainable development of society. The government should enhance the underclass groups’ perceived desirability through skill conversion, financial innovation, and digital empowerment; improve their perceived feasibility through the entrepreneurial resilience-building platform and the “Village Sage Mentorship System”; and refine the land expropriation policy by means of the securitization of collective assets, the multifunctional utilization of rural homesteads, and the cultivation of localized new business formats. By doing so, it can promote the transformation of the underclass groups’ entrepreneurship from the “subsistence type” to the “opportunistic type”.

1. Introduction

Since the reform and “opening-up” policy in 1978, urbanization has brought tremendous impetus to China’s economic and social development. However, rapid urbanization has also created large income gaps between different groups in urban and rural areas [1]. Therefore, how to improve the income of the underclass groups involved in urbanization is a key concern in narrowing the income gap [2]. But the underclass groups are increasingly separated from industrialized and modern societies under the shock of globalization, and they will face greater difficulties in survival, employment, and development [3,4]. In recent years, China has experienced accelerated transfer of rural labor employment, a peak of new labor force employment, and high reemployment of unemployed persons, as well as a dramatic imbalance between supply and demand [5,6]. Data from the National Bureau of Statistics (NBS, 2024) [7] show that in 2024, there were 12.56 million new people of working age living in cities, with a 5.1% unemployment rate. China’s urban areas currently host a rural-to-urban migrant population of 260 million, accounting for 18.7% of the nation’s total population and 27.7% of the permanent urban population. This massive population movement stems from two structural drivers. On the one hand, influenced by the strong pulling effect of the industrial adjustment between the primary industry in rural areas and the secondary and tertiary industries in urban areas on employment demand [8], the urban–rural income ratio has been maintained at above 2.5:1 for a long time. On the other hand, it is influenced by the flexibility of the household registration system [9,10] and the land expropriation compensation system [11,12]. For example, big cities set mobility thresholds through means such as the point-based household registration system, the required years of social security contributions, and housing restrictions. Meanwhile, the compensation standards for rural land expropriation have long been lower than the land appreciation income. Consequently, while employment demand among rural laborers intensifies, institutional thresholds continue to escalate. As the demand for low-skilled labor decreases, underclass groups have fewer and fewer opportunities to participate in the job market and weakening ability to participate in the market competition involved in urbanization. Underclass groups’ future survival and development will be the focus of urbanization, and the results will affect the stability and sustainable development of society [13].

However, most of the underclass groups involved in urbanization are subsistence-type laborers, and it is difficult for them to embark on an opportunistic-type path to sustainable development. So, encouraging the entrepreneurship of underclass groups is considered to be the most effective way through which they can achieve a sustainable livelihood. This has become a policy priority of the central government. For example, the 20th National Congress of the Communist Party of China called for improving the guarantee system for promoting entrepreneurship-led employment and supporting and regulating the development of new forms of employment. Similar local public policies followed. But the effectiveness of policies targeting the entrepreneurial behavior of the underclass groups (EBUG) has not been significant, and from an academic point of view, the main reason for this problem is the failure to identify the critical impact path on the EBUG involved in urbanization. The literature shows that on the one hand, most of the studies are conducted based on case analysis and experience summaries, which apply qualitative analysis methods to design entrepreneurship training systems, entrepreneurship financing systems, and related entrepreneurship countermeasures. On the other hand, the academic community recognizes that entrepreneurship is important for ensuring underclass groups’ long-term livelihood. However, existing studies are scattered and in the exploratory phase. Scholars have carried out qualitative analyses of influencing factors, but they have drawn different conclusions, and most conclusions lack data validation. For this reason, the influencing factors of underclass groups’ entrepreneurship call for further analysis and demonstration, and it is necessary to carry out detailed investigations from a microcosmic perspective to develop and improve the theory of the EBUG and establish a systematic, policy-oriented strategy for promoting underclass groups’ entrepreneurship. The contributions of this paper’s research are as follows: (1) The model of the EBUG involved in urbanization is constructed according to a theoretical analysis framework of “intention–situation–behavior” based on the theories of planned behavior (TPBs), the entrepreneurial event model (EEM), and social cognitive theory (SCT), and the influencing mechanism of the EBUG involved in urbanization is analyzed. (2) Through the investigation of different underclass groups, the theoretical model of the EBUG involved in urbanization is validated by using structural equation modeling (SEM), and the key influencing paths of perceived desirability and perceived feasibility as the main effects and land expropriation as the moderating effect are defined from a microcosmic perspective. (3) Corresponding policy suggestions are put forward based on the analysis of the influence paths.

Therefore, the purpose of this paper is to analyze the influencing mechanism and functional paths of the EBUG involved in urbanization from a microcosmic perspective, and the conceptual framework diagram is shown in Figure 1. First, the Section 1 describes the background, significance, and challenges of the research study. Second, the Section 2 describes the research progress on the connotation, entrepreneurial behavior, and influence paths of the EBUG involved in urbanization. Third, Section 3 sets forth the overall concept of the theoretical model and preconditioned hypothesis. Fourth, the Section 4 applies grounded theory approaches to establish a theoretical model of “intention–situation–behavior” for the EBUG involved in urbanization. Finally, through field surveys and in-depth interviews, the SEM method is used to verify the theoretical model with large sample data and clarify the influencing mechanism and functional paths of the EBUG involved in urbanization. Our goal is to develop and improve the theory of “intention–situation–behavior” for the EBUG, laying a theoretical foundation for promoting their entrepreneurship and providing a decision-making reference for the sustainable development and livelihood of underclass groups involved in urbanization and the long-term peace and stability of society.

2. Literature Review

2.1. The Underclass Groups Involved in Urbanization

The term “underclass” emerged as a sociological concept in the early 1960s and gained popularity in the 1980s as the rural–urban split widened and various economic reforms changed the urban population and environment [14]. The underclass concept has never been consistently defined despite constant use in developed countries. It was first used to describe American social problems by Myrdal [15] and was more influentially used by Wilson [16]. Wilson defined the underclass as people without formal education or experience, those who are unemployed for an extended period of time or do not participate in the labor force, and families that endure protracted periods of poverty and/or welfare dependency [17]. Two types of persons were identified as the main elements of China’s urban underclass in earlier studies. One includes workers who have been laid off by state-owned enterprises, and the other refers to migrant workers who are underpaid and underprivileged and have migrated from rural to urban areas [8,18]. However, in this paper, the underclass is defined as the population produced by China’s urbanization. This group is similar to but also different from the urban underclass in the previous social environment. It is concurrently farmer and worker but is not purely engaged in agriculture or industry. The emergence of this special group has resulted from imbalanced progress in the reforms of various systems in China’s economic transformation. There are two types of underclass groups involved in urbanization: voluntary and involuntary. Both groups feel hesitant and anxious about the future, with some becoming discouraged and marginalized, wandering between urban and rural areas.

(1) The voluntary type comprises underclass groups in active urbanization who take the initiative to stop engaging in agricultural production and start working in industrial and service industries. This type consists of migrant workers, one of the urban underclass groups. They are often described as “migratory birds” in urban areas largely due to the urban–rural dual system (such as China’s “hukou” system) [10,19]. Household registration regulations deny the “floating” population equitable access to government-provided services and basic welfare that urban residents enjoy [8]. In addition, this group is prevented from taking decent and high-paid jobs because of the lack of advanced skills in the urban labor market [20]. Thus, the situation of migrant workers is worse than that of laid-off workers in urban areas [18].

(2) The involuntary type includes underclass groups of passive urbanization, who are forced to participate in urbanization. This type includes landless farmers, who differ from urban underclass groups but are important constituents. With rapid urbanization and economic development, China’s government has expropriated a large amount of agricultural land from farmers for infrastructure and property development [21,22,23]. The landless farmers are forced to give up engaging in agricultural production and start working in industrial and service industries. Although they can receive one-off cash sums and other compensation from the government, which may maintain or even improve their lives in the short term, they are not guaranteed long-term livelihoods [22]. If they only engaged in agricultural production before land expropriation, they may face the same problems as migrant workers, such as in hukou, employment, education level, and welfare [24]. Therefore, this group should be included in the underclass groups involved in urbanization.

2.2. The EBUG Involved in Urbanization

Research on the entrepreneurship of the underclass groups involved in urbanization is still limited. International studies on the EBUG involved in urbanization have focused on entrepreneurship among farmers [25,26,27], immigrants [28,29], and vulnerable groups [30,31]. Although the research objects of our study and other empirical studies have similarities, there are also large differences.

(1) In terms of farmer entrepreneurship, according to Pyysiäinen, Anderson, McElwee, and Vesala [32], the entrepreneurial tasks of traditional agriculture differ from those of diversification entrepreneurship, which requires market relations, professional skills, the ability to access resources, etc. These skills are directly related to social resources, relationships, and networks. Many scholars also note that farmers need policy support for entrepreneurship. Based on a survey of Shropshire farmers in the UK, Tate reported that changes in European agriculture and environmental policies from 1997 to 2009 influenced farmers’ entrepreneurial behavior [33]. Following a survey of ten remote rural areas in Germany, Greece, Poland, Portugal, and the UK, North and Smallbone noted that the rural entrepreneurship policies of different countries vary in terms of their effects, which is due mainly to the experience of local governments in policy interventions and their adaptability to the actual situations of local businesses [34]. Therefore, in different entrepreneurial environments, different regional goals and policy frameworks should be developed. Farmer entrepreneurship is the most similar to the entrepreneurship of underclass groups involved in urbanization, yet there are also differences. Farmers are generally considered agricultural producers who are dependent on land [35] and whose aim is subsistence, not reinvestment [36]. Studies on farmers’ entrepreneurship in China have focused primarily on internal determinants of entrepreneurship (such as individual and psychological characteristics) [37]. However, our study area is closely related to the process of urbanization. We believe that external factors, such as land use policy, also matter.

(2) In terms of immigrant entrepreneurship, it can help immigrants not only obtain jobs and move out of poverty but also become increasingly opportunity-oriented, bringing additional sources of innovation to the host region while creating jobs for both other immigrants and locals. Immigrant entrepreneurship has become an important factor in promoting socioeconomic development in many developed Western countries [38]. Western scholars are mostly concerned with the first generation of voluntary immigrants’ entrepreneurial activity [39]. For example, policies on social, financial, human, and cultural capital are more valuable for immigrant entrepreneurship. Overseas immigrants differ from internal immigrants because of their different cultural and ethnic backgrounds; thus, studies on the entrepreneurship of overseas immigrants focus on intergenerational differences, ethnic strategies, and cultural backgrounds [40,41]. However, underclass groups, as the main component of internal migration, have different characteristics and backgrounds in China’s economic and social environments [42]. Therefore, studies on the entrepreneurship of underclass groups reference different influencing factors.

(3) In terms of vulnerable groups’ (including women, recent immigrant youth, persons with disabilities, and people who are ill or immunocompromised) entrepreneurship, entrepreneurship is an effective way to end the cycle of poverty. It is significantly affected by the availability and quality of social capital [43]. For example, education (i.e., business training for microcredit recipients) plays an important role in encouraging entrepreneurship among women in Vietnam [31]. Entrepreneurship among vulnerable groups usually needs more government support and social protection programs [44]. Therefore, these types of groups usually require certain types of non-economic support. However, when underclass groups are unemployed, they become new vulnerable urban groups. The influencing factors of underclass groups’ entrepreneurship are significantly different, and such groups may require more economic support.

In this paper, the EBUG refers to the entrepreneurial activities carried out by the underclass groups in the urban-rural interface characterized by both survival-driven and development-oriented transitions. These activities are carried out by overcoming the constraints of institutional exclusions (the dual exclusions of the rural land system and the urban household registration system). Such entrepreneurship integrates two-way resource allocation between rural surplus labor and urban low-end market demands.

2.3. The Impact Paths Analysis of the EBUG

With respect to problems related to the entrepreneurial impact mechanism, previous studies have adopted a variety of perspectives, such as political, economic, and psychological. For example, some scholars have focused on the roles played by institutions and incentives for credit availability, property rights, regulation, and technology policy in entrepreneurship from the angle of economics [45]. From the sociology angle, some have found that the social networks, values, and social norms related to entrepreneurial activities are the main influencing factors [46]. Other scholars have explored how personality traits, including risk attitudes and self-confidence, may affect entrepreneurs’ behavior from a psychology perspective [47]. However, most research in this field has focused on the entrepreneurial impact mechanism in developed economies, and only a handful of studies address these issues in developing countries [48].

China’s rapid urbanization is usually closely linked to land expropriation [49,50]. Moreover, the process of urbanization has created a large population of underclass groups [49]. These groups remain subject to various uncertainties in sustaining their lives, competing in the labor market, and adapting to urban life [50]. Therefore, the impact mechanism of underclass groups’ entrepreneurship involved in urbanization is affected by land expropriation with rapid urbanization and industry development in China [22,23]. Land expropriation is a very important influencing factor and is closely related to underclass groups’ entrepreneurship in China. Therefore, this study differs from previous studies in terms of impact mechanisms and theoretical models.

3. Theoretical Framework and Testable Hypotheses

3.1. Theoretical Framework

The conceptual model of this study was developed by selecting 23 representative underclass groups engaged in entrepreneurship from pre-investigation areas (such as Tonglu County in Hangzhou City, Jiubao Town in Hangzhou City, Cixi City in Ningbo City, etc.). An in-depth individual interview approach was adopted to investigate their entrepreneurial situations. Then, by utilizing the grounded theory method, an inductive analysis was carried out on the interview data, which principally comprised four steps: open coding, axial coding, theoretical coding, and theoretical saturation testing. In this manner, concepts and categories were abstracted. Drawing on prior research related to the TPBs, the EEM, and SCT, and building on the general “intention–behavior” main-effect framework, considering the particularity of the underclass groups in the process of urbanization, the land expropriation situation was introduced, and the model was improved and innovated. Thus, a theoretical framework of “intention–situation–behavior” under the influence of the land expropriation situation was constructed, and an empirical test was conducted to explore the mechanism of the main effect and the moderating effect relationship that conforms to the characteristics of the underclass groups involved in urbanization, as shown in Figure 2.

As shown in Figure 2, according to the TPBs [51,52] and the EEM [53], underclass groups’ entrepreneurial intention is composed mainly of achievement motivation, innovation orientation, social capital awareness, and market opportunity awareness. These four influencing factors of underclass groups’ entrepreneurial intention fall into perceived desirability and perceived feasibility, which are two basic cognitive paths for entrepreneurial behavior.

Perceived desirability refers to the degree of attraction from the prospect of an entrepreneurial activity for underclass groups, which demonstrates whether entrepreneurship meets underclass groups’ wishes (value brought about by entrepreneurship). If the expected value of entrepreneurship is more consistent with the underclass groups’ entrepreneurship wishes, the underclass groups’ entrepreneurial intention is stronger and more likely to be devoted to entrepreneurship. Perceived desirability encompasses distinctive elements, including achievement motivation and innovation orientation. These personal traits, which are evident in every facet of a person’s life, can reveal whether underclass people have the capacity to start their own businesses.

Perceived feasibility refers to the extent to which underclass groups are confident in their ability to engage in entrepreneurial behavior, which indicates their judgement of their own ability (whether entrepreneurship is feasible). In this paper, perceived feasibility is understood as potential underclass entrepreneurs’ perceived judgement of their own entrepreneurial knowledge, skills, and experience. If perceived feasibility is stronger, entrepreneurship is more likely. Perceived feasibility includes perceived social capital, perceived market opportunities, and other factors at the resource level. These individual resources, which can be identified, can influence whether underclass groups engage in entrepreneurial behavior. This is consistent with the finding of Krueger [54], who perceived that feasibility has a significant positive correlation with entrepreneurial intention.

Although perceived desirability and perceived feasibility are two basic cognitive paths for entrepreneurial behavior according to the TPBs and the EEM, scholars have recognized that there may be an important moderator variable in the intention–behavior conversion, but the relationship between them has yet to be clarified [55,56,57]. One reason for this gap is that the “attitude variable” and “intention”, as proposed in the model to explain the entrepreneurial process, are “exante judgements” and cannot predict the occurrence, success, or failure of entrepreneurial behavior. The intention model does not predict behavior but rather predicts intention [58,59]. The other reason for the gap involves the explanation intensity of the intention for behavior. The intention model does not involve behavior. According to Sutton’s [60] empirical results for the TPBs, subjective norms, behavioral attitudes and behavior control power account for 40–50% of the explanation intensity of intention for behavior, which is much greater than the forecast rate of behavior. Therefore, this paper introduces the land expropriation variable into the intention–behavior conversion according to SCT. Land expropriation is most closely related to the underclass groups involved in urbanization in China [22,23].

The moderating effect is influenced by the characteristics (intensity and structure) of the entrepreneurship intention of the underclass groups. When the entrepreneurship intention of the underclass groups is weak (the entrepreneurship intention is unclear) or they only have the perceived feasibility of entrepreneurial behavior (the perceived desirability is relatively weak), the moderating effect of the land expropriation situation variable is relatively strong. For example, due to the development of scenic spots in Jingzhong Village, Cixi, Ningbo, the number of local tourists has increased, and the market and opportunities have also increased. Coupled with the policy guidance of the local government, the underclass groups with weak perceived desirability and perceived feasibility of entrepreneurship have also commenced to engage in scattered entrepreneurial activities. Subsequently, others will follow suit, forming a good entrepreneurial atmosphere. Another example is that the entrepreneurial support policies implemented for the underclass groups after land expropriation will make those underclass groups with strong perceived feasibility immediately perceive the increase in entrepreneurial resources and the improvement in entrepreneurial feasibility and then choose to engage in entrepreneurial behavior. Conversely, when the entrepreneurship intention of the underclass groups is strong (the entrepreneurial willingness is very strong, and the goal is clear) or the perceived desirability of entrepreneurial behavior is strong, the moderating effect of the land expropriation situation variable is relatively weak. It is also discovered in the investigation that in such cases, the underclass groups have often already been engaged in their own entrepreneurial undertakings before land expropriation.

Since the land expropriation process involves multiple aspects, the moderating effects of the four land expropriation situation factors, namely, the land expropriation mode, the resettlement mode, the compensation mode, and the complementary policy, on the entrepreneurship intention and behavior of underclass groups are different. For example, it is found in the investigation that the land expropriation mode is a relatively obvious moderating factor and the number of underclass groups engaged in entrepreneurship in remote rural areas is significantly fewer than that in the urban–rural fringe. Another example is that for those underclass groups who receive a large amount of land expropriation compensation, the compensation amount weakens their perceived desirability for entrepreneurship. Even if they possess certain achievement motivation and innovativeness, they will not apply them to entrepreneurial behavior.

Current research suggests many different formation mechanisms for the EBUG involved in urbanization. The mature variable scope, measurement scale, and theoretical hypotheses are not yet available for reference. It is not advised to directly adopt the research method of hypothesis–deduction. Therefore, grounded theory is applied to select variables for the EBUG in urbanization. Grounded theory is a qualitative research method created by Glaser and Strauss [61] and is not limited by theoretical hypotheses; it is used to make inductive analyses of data and focuses on the formation of a conceptual framework or theory [62,63]. In this paper, we choose eight significant variables, four of which are related to entrepreneurship intention, including achievement motivation, innovation orientation, social capital, and market opportunities, while the other four are related to the land expropriation contexts, including land expropriation mode, resettlement mode, compensation mode, and complementary policy, as shown in

Table 1. Description of model variables.

Model Paths	Significant Variables	Explanatory Notes
Formation Paths of Entrepreneurship Intention	Achievement motivation	Achievement motivation refers to an individual’s drive to pursue work they perceive to be significant and valuable, motivated by the goal of achieving success through self-imposed high standards. This motivational construct exerts a significant influence on the entrepreneurial intentions and risk tolerance of underclass groups.
	Innovation orientation	Innovativeness refers to the ability to develop original solutions to problems. This trait influences the entrepreneurial intentions of groups: specifically, groups characterized by innovativeness are more likely to develop entrepreneurial intentions.
	Social capital	Social capital refers to social networks, norms of reciprocity, and the trust generated therefrom, representing resources derived from individuals’ positions within social structures. The quantity of social capital influences the entrepreneurial intentions of groups.
	Market opportunities	Market opportunities refer to the ability of underclass groups to monitor the economic activities occurring around them, identify key elements such as target markets and customer segments relevant to entrepreneurial projects, and exploit these opportunities to initiate business ventures.
Moderating Paths in the Context of Land Expropriation	Land expropriation mode	One of the contextual factors of land expropriation that affects the underclass groups’ entrepreneurship. The land expropriation mode refers to the location and development type of the expropriated land.
	Resettlement mode	One of the contextual factors of land expropriation that affects the underclass groups’ entrepreneurship. The resettlement mode refers to in situ resettlement and off-site resettlement.
	Compensation mode	One of the contextual factors of land expropriation that affects the underclass groups’ entrepreneurship. For potential underclass group entrepreneurial individuals, land expropriation compensation fees can ease the financial pressure of entrepreneurship, but there is a tendency for this promotion to have a diminishing marginal effect.
	Complementary policy	One of the contextual factors of land expropriation that affects the underclass groups’ entrepreneurship. Complementary policies, including credit financing, platform construction, entrepreneurship training, and related services, are relevant to underclass groups’ entrepreneurship and affect the choices and judgements of the underclass groups.

(Source: authors’ own creation.)

.

3.2. Testable Hypotheses

(1) Hypotheses on the relationship between entrepreneurial intention and entrepreneurial behavior

Entrepreneurial intention has proven to be a fundamental and common variable in entrepreneurship studies [64]. The TPBs and the EEM are sophisticatedly used in the field of entrepreneurship cognition, in which it has been noted that “entrepreneurial intention is an underlying factor in entrepreneurial behaviour” [54]. Entrepreneurial intention reflects an individual’s motivation to put a conscious plan or decision into action, which is a prerequisite for engagement in entrepreneurial behavior [65]. To some extent, entrepreneurial intention highlights an individual’s desire and preference for entrepreneurship in their career development. Entrepreneurial intention ranges between attitude and behavior and has a strong ability to predict behavior [66]. Previous studies have demonstrated that entrepreneurship is a conscious and planned behavior [67]. The TPBs and the EEM highlight internal psychological mechanisms through which individual characteristic information and objective information influence individual entrepreneurial behavior, which implies the influence of individual cognition (including perceived desirability and perceived feasibility) on behavior [51,53]. On this basis, the following hypotheses are proposed in this paper:

H1. 

Perceived desirability has a significant influence on the EBUG.

H2. 

Perceived feasibility has a significant influence on the EBUG.

(2) Hypothesis on land expropriation as a moderator between entrepreneurial intention and entrepreneurial behavior

Scholars have recognized that there may be an important situation variable in the intention–behavior conversion and that this relationship has yet to be fully clarified. In relevant studies on environmental protection behaviors outside the field of entrepreneurship research, many scholars have shown that situational factors moderate the relationship between intention and behavior [55,56,57]. For example, Guagnano et al. [68] found that environmental behavior is the result of the interaction between the attitude variable and the situational factor. Johnston et al. [55] explored the role of dispositional and situational factors as moderators in promoting compliance behavior.

Since the land expropriation process involves many aspects, this paper regards four areas (land expropriation mode, resettlement mode, compensation mode, and complementary policy) with great influence, as found in field surveys, as situational factors of land expropriation. Inspiration can be obtained from studies on environmental behavior. First, perceived desirability and perceived feasibility are considered key attitude variables of entrepreneurial behavior. Second, the entire land expropriation process influences underclass groups’ judgement of their “desire” for and the “feasibility” of entrepreneurship. Field survey findings prove that the four situational factors of land expropriation moderate the relationship between entrepreneurial intention and entrepreneurial behavior. For example, for those underclass people who receive large amounts of compensation for land expropriation, the amount of compensation undermines the perceived desirability of entrepreneurship. In another example, due to the development of tourist attractions in villages, the number of local tourists has increased, along with the number of market opportunities. Coupled with the policy guidance of local governments, underclass people with weak perceived desirability and perceived feasibility also begin to engage in scattered entrepreneurial activities. All people subsequently follow suit and create a favorable business climate. On this basis, the following hypothesis is set forth:

H3. 

Land expropriation has a significant moderating effect on the relationship between entrepreneurial intention and entrepreneurial behavior.

4. Econometric Models

This study uses SEM [69] to test the theoretical hypotheses for the following reasons: First, when all the exogenous variables are latent variables and are measured by multiple indicators, SEM plays a very important role in the interactions among the variables. SEM is a multivariate technique that incorporates observed (measured) and unobserved variables (latent variables), whereas traditional statistical methods analyze only observed variables [70,71,72,73]. The variables in this study include latent variables and interactions among the variables. Second, SEM is a multivariate statistical analysis technique for establishing, estimating, and examining causal relationship models, which can explain not only the direct effects but also the indirect effects of variables [69,73]. SEM can handle multiple dependent variables at the same time and is more suitable for multicause analysis in this paper than the traditional regression model. Third, SEM is a highly flexible and comprehensive methodology that is appropriate for the study of entrepreneurial behavior [74,75,76]. Therefore, we use SEM to study the EBUG.

The model of the EBUG depicted in Figure 2 represents typical SEM. Entrepreneurship behavior, as an endogenous latent variable, is expressed as η. Perceived desirability and perceived feasibility, as the exogenous latent variables of entrepreneurial intention, are expressed as ξ₁ and ξ₂. Land expropriation is the moderator variable between entrepreneurial intention and entrepreneurial behavior and is expressed as ξ₃. Therefore, we need to address the interaction effect in the EBUG model, which is affected by the moderator variable. There are usually three key steps in dealing with SEM with interactive effects: indicator centering, strategies for generating product indicators, and parameter constraint methods [77,78,79]. First, indicator centering is a standard step before modeling interaction effects to reduce the multicollinearity problem and simplify the model [77,79]. Second, devising strategies for generating product indicators is the most important and complex problem when building the EBUG model. There are many methods for analyzing the interaction effects of latent variables via SEM [80]. Most of them are derived from the groundbreaking work by Kenny and Judd [78], who first used SEM with product indicators. Previous researchers have explored strategies for generating product indicators, such as all possible product indicators, matching product indicators [79,81], and single product indicators [82,83,84,85]. After considering the simplicity of modeling, the fit index of the model, estimation bias, and accuracy, previous research has shown that the matching product indicators are reasonable [79,86,87]. However, we note the following: (1) all indicators should be used to make full use of information; (2) no indicator should be reused to avoid high correlation, resulting in multicollinearity. More importantly, if the indicators are reused, the variance–covariance matrix of the error is a diagonal matrix, and many constraint equations are reduced in the constraint method [79,81,86]. In this study, according to the methods by Marsh et al. [79], we analyze all the indicators of perceived desirability, perceived feasibility, and land expropriation via confirmatory factor analysis; sort the load of the fully standardized solution from high to low; and match product indicators according to “large with large, small with small”. Finally, the product indicator of the moderator variable is expressed as MV, which includes X₂₂X₃₁, X₁₂X₃₄, X₁₁X₃₂, and X₂₁X₃₃. An unconstrained method is used to simplify the model according to Marsh et al. [79]. The path construction of the EBUG model is shown in Figure 3.

In the EBUG model, the latent structural equations are a set of linear regression equations that explain the internal structural relationships among the constructs. The measurement equations are factor analysis equations or other types of equations used to describe the external relationships between each construct and its indicators. The product indicators can be explained by the multiplication equation of the two measurement equations, called the moderator effect equations. In this study, the type of estimation technique used is the maximum likelihood (ML)-based covariance structure analysis method, which is documented in software such as SPSS Amos 24 [88,89]. In addition to the internal structural equations, external measurement equations, and moderator effect equations, weight relations equations are required to complete ML modeling. ML estimates the case values of each construct by the weighted aggregate values of its indicators, where the weights are factor loadings for reflective indicators and regression coefficients for formative indicators after rescaling [90]. In

Table 2. Structural equation model for EBUG model.

General Model	Specific Model
Latent structural equations ($\mathsf{\eta }=\mathrm{B}\mathsf{\eta }+\mathsf{\Gamma }\mathsf{\xi }+\mathsf{\zeta }$)	$\mathsf{\eta }={\mathsf{\gamma }}_1{\mathsf{\xi }}_1+{\mathsf{\gamma }}_2{\mathsf{\xi }}_2+{\mathsf{\gamma }}_3{\mathsf{\xi }}_3+{\mathsf{\gamma }}_4{\mathsf{\xi }}_1{\mathsf{\xi }}_3+{\mathsf{\gamma }}_5{\mathsf{\xi }}_2{\mathsf{\xi }}_3+\mathsf{\zeta }$
	Centralized equation: $\mathsf{\eta }={\mathsf{\gamma }}_1^{\prime }{\mathsf{\xi }}_1^{\mathrm{c}}+{\mathsf{\gamma }}_2^{\prime }{\mathsf{\xi }}_2^{\mathrm{c}}+{{\mathsf{\gamma }}_3^{\prime }{\mathsf{\xi }}_3^{\mathrm{c}}+\mathsf{\gamma }}_4^{\prime }{\mathsf{\xi }}_1^{\mathrm{c}}{\mathsf{\xi }}_3^{\mathrm{c}}{+\mathsf{\gamma }}_5^{\prime }{\mathsf{\xi }}_2^{\mathrm{c}}{\mathsf{\xi }}_3^{\mathrm{c}}+{\mathsf{\zeta }}^{\prime }$ ${\mathsf{\xi }}_1^{\mathrm{c}}={\mathsf{\xi }}_1-\overline{{\mathsf{\xi }}_1}$ ${\mathsf{\xi }}_2^{\mathrm{c}}={\mathsf{\xi }}_2-\overline{{\mathsf{\xi }}_2}$ ${\mathsf{\xi }}_1^{\mathrm{c}}{\mathsf{\xi }}_3^{\mathrm{c}}=\left({\mathsf{\xi }}_1-\overline{{\mathsf{\xi }}_1}\right)\left({\mathsf{\xi }}_3-\overline{{\mathsf{\xi }}_3}\right)$ ${\mathsf{\xi }}_2^{\mathrm{c}}{\mathsf{\xi }}_3^{\mathrm{c}}=\left({\mathsf{\xi }}_2-\overline{{\mathsf{\xi }}_2}\right)\left({\mathsf{\xi }}_3-\overline{{\mathsf{\xi }}_3}\right)$
Reflective measurement equations $\left(\mathrm{x}={\mathsf{\Lambda }}_{\mathrm{x}}\mathsf{\xi }+{\mathsf{\delta }}_{\mathrm{x}}\right)$	${\mathrm{x}}_{1\mathrm{i}}={{\mathsf{\lambda }}_{\mathrm{X}}}_{1\mathrm{i}}{\mathsf{\xi }}_1+{{\mathsf{\delta }}_{\mathrm{X}}}_{1\mathrm{i}} \mathrm{i}=1,2,3,4$ ${\mathrm{x}}_{2\mathrm{i}}={{\mathsf{\lambda }}_{\mathrm{X}}}_{2\mathrm{i}}{\mathsf{\xi }}_2+{{\mathsf{\delta }}_{\mathrm{X}}}_{2\mathrm{i}} \mathrm{i}=1,2,3,4$ ${\mathrm{x}}_{3\mathrm{i}}={{\mathsf{\lambda }}_{\mathrm{X}}}_{3\mathrm{i}}{\mathsf{\xi }}_3+{{\mathsf{\delta }}_{\mathrm{X}}}_{3\mathrm{i}} \mathrm{i}=1,2,3,4$
Moderator effect equations $\left[{\mathrm{x}}_{\mathrm{i}}{\mathrm{x}}_{\mathrm{j}}=\left({\mathsf{\Lambda }}_{\mathrm{x}\mathrm{i}}{\mathsf{\xi }}_{\mathrm{i}}+{\mathsf{\delta }}_{\mathrm{x}\mathrm{i}}\right)\left({\mathsf{\Lambda }}_{\mathrm{x}\mathrm{j}}{\mathsf{\xi }}_{\mathrm{j}}+{\mathsf{\delta }}_{\mathrm{x}\mathrm{j}}\right)\right]$	${\mathrm{x}}_{1\mathrm{i}}{\mathrm{x}}_{3\mathrm{i}}={{\mathsf{\lambda }}_{\mathrm{X}}}_{1\mathrm{i}3\mathrm{i}}{\mathsf{\xi }}_1{\mathsf{\xi }}_3+{{\mathsf{\lambda }}_{\mathrm{X}}}_{1\mathrm{i}}{\mathsf{\xi }}_1{{\mathsf{\delta }}_{\mathrm{X}}}_{3\mathrm{i}}+{{\mathsf{\lambda }}_{\mathrm{X}}}_{3\mathrm{i}}{\mathsf{\xi }}_3{{\mathsf{\delta }}_{\mathrm{X}}}_{1\mathrm{i}}+{{\mathsf{\delta }}_{\mathrm{X}}}_{1\mathrm{i}3\mathrm{i}} \mathrm{i}=1,2,3,4$ ${\mathrm{x}}_{2\mathrm{i}}{\mathrm{x}}_{3\mathrm{i}}={{\mathsf{\lambda }}_{\mathrm{X}}}_{2\mathrm{i}3\mathrm{i}}{\mathsf{\xi }}_2{\mathsf{\xi }}_3+{{\mathsf{\lambda }}_{\mathrm{X}}}_{2\mathrm{i}}{\mathsf{\xi }}_2{{\mathsf{\delta }}_{\mathrm{X}}}_{3\mathrm{i}}+{{\mathsf{\lambda }}_{\mathrm{X}}}_{3\mathrm{i}}{\mathsf{\xi }}_2{{\mathsf{\delta }}_{\mathrm{X}}}_{2\mathrm{i}}+{{\mathsf{\delta }}_{\mathrm{X}}}_{2\mathrm{i}3\mathrm{i}} \mathrm{i}=1,2,3,4$
Weight relations $\left(\widehat{\mathsf{\xi }}={\mathsf{\Omega }}_{\mathsf{\xi }}\mathrm{x},\widehat{\mathsf{\eta }}={\mathsf{\Omega }}_{\mathsf{\eta }}\mathrm{y}\right)$	${\widehat{\mathsf{\epsilon }}}_1={\mathsf{\omega }}_{\mathsf{\xi }11}{\mathrm{x}}_{11}+{\mathsf{\omega }}_{\mathsf{\xi }12}{\mathrm{x}}_{12}+{\mathsf{\omega }}_{\mathsf{\xi }13}{\mathrm{x}}_{13}+{\mathsf{\omega }}_{\mathsf{\xi }14}{\mathrm{x}}_{14}$ ${\widehat{\mathsf{\epsilon }}}_2={\mathsf{\omega }}_{\mathsf{\xi }21}{\mathrm{x}}_{21}+{\mathsf{\omega }}_{\mathsf{\xi }22}{\mathrm{x}}_{22}+{\mathsf{\omega }}_{\mathsf{\xi }23}{\mathrm{x}}_{23}+{\mathsf{\omega }}_{\mathsf{\xi }24}{\mathrm{x}}_{24}$ ${\widehat{\mathsf{\epsilon }}}_3={\mathsf{\omega }}_{\mathsf{\xi }31}{\mathrm{x}}_{31}+{\mathsf{\omega }}_{\mathsf{\xi }32}{\mathrm{x}}_{32}+{\mathsf{\omega }}_{\mathsf{\xi }33}{\mathrm{x}}_{33}+{\mathsf{\omega }}_{\mathsf{\xi }34}{\mathrm{x}}_{34}$ ${\widehat{\mathsf{\epsilon }}}_4={\mathsf{\omega }}_{\mathsf{\xi }41}{\mathrm{x}}_{11}{\mathrm{x}}_{31}+{\mathsf{\omega }}_{\mathsf{\xi }42}{\mathrm{x}}_{12}{\mathrm{x}}_{32}+{\mathsf{\omega }}_{\mathsf{\xi }43}{\mathrm{x}}_{13}{\mathrm{x}}_{33}+{\mathsf{\omega }}_{\mathsf{\xi }44}{\mathrm{x}}_{14}{\mathrm{x}}_{34}$ ${\widehat{\mathsf{\epsilon }}}_5={\mathsf{\omega }}_{\mathsf{\xi }51}{\mathrm{x}}_{21}{\mathrm{x}}_{31}+{\mathsf{\omega }}_{\mathsf{\xi }52}{\mathrm{x}}_{22}{\mathrm{x}}_{32}+{\mathsf{\omega }}_{\mathsf{\xi }53}{\mathrm{x}}_{23}{\mathrm{x}}_{33}+{\mathsf{\omega }}_{\mathsf{\xi }54}{\mathrm{x}}_{24}{\mathrm{x}}_{34}$ $\widehat{\mathsf{\eta }}={\mathsf{\omega }}_{\mathsf{\eta }1}{\mathrm{y}}_1+{\mathsf{\omega }}_{\mathsf{\eta }2}{\mathrm{y}}_2+{\mathsf{\omega }}_{\mathsf{\eta }3}{\mathrm{y}}_3$

(Source: authors’ own creation.) Note: η represents entrepreneurial behavior; ξi represent latent variables; B is a matrix of coefficients relating the latent endogenous variables to each other; Γ is a matrix of coefficients relating the endogenous variables to the exogenous variables; ${\mathsf{\gamma }}_{\mathrm{i}}$ represent path coefficients; ζ is a constant term; λi refer to path coefficients, which indicate the degree of influence of observed variables on the corresponding latent variables; ${\mathrm{x}}_{\mathrm{i}}$ refer to random factors of the corresponding variables; Λx are matrices of factor loadings; δx are vectors of measurement errors in x; Ω is a matrix that reflects the weight relationships among variables; ω is a coefficient that reflects the weight relationships among variables.

, we present the complete ML formulation of the general EBUG model.

However, according to the principles and methods of SEM construction with the product indicators above, the moderator variable is expressed as four product indicators—X₂₂X₃₁, X₁₂X₃₄, X₁₁X₃₂, and X₂₁X₃₃; consequently, the moderator effect equation is defined as

$${\mathrm{x}}_{22}{\mathrm{x}}_{31}=\left({{\mathsf{\lambda }}_{\mathrm{X}}}_{22}{\mathsf{\xi }}_2+{{\mathsf{\delta }}_{\mathrm{X}}}_{22}\right)\left({{\mathsf{\lambda }}_{\mathrm{X}}}_{31}{\mathsf{\xi }}_3+{{\mathsf{\delta }}_{\mathrm{X}}}_{31}\right)$$

(1)

$${\mathrm{x}}_{12}{\mathrm{x}}_{34}=\left({{\mathsf{\lambda }}_{\mathrm{X}}}_{12}{\mathsf{\xi }}_1+{{\mathsf{\delta }}_{\mathrm{X}}}_{12}\right)\left({{\mathsf{\lambda }}_{\mathrm{X}}}_{34}{\mathsf{\xi }}_3+{{\mathsf{\delta }}_{\mathrm{X}}}_{34}\right)$$

(2)

$${\mathrm{x}}_{11}{\mathrm{x}}_{32}=\left({{\mathsf{\lambda }}_{\mathrm{X}}}_{11}{\mathsf{\xi }}_1+{{\mathsf{\delta }}_{\mathrm{X}}}_{11}\right)\left({{\mathsf{\lambda }}_{\mathrm{X}}}_{32}{\mathsf{\xi }}_3+{{\mathsf{\delta }}_{\mathrm{X}}}_{32}\right)$$

(3)

$${\mathrm{x}}_{21}{\mathrm{x}}_{33}=\left({{\mathsf{\lambda }}_{\mathrm{X}}}_{21}{\mathsf{\xi }}_1+{{\mathsf{\delta }}_{\mathrm{X}}}_{21}\right)\left({{\mathsf{\lambda }}_{\mathrm{X}}}_{33}{\mathsf{\xi }}_3+{{\mathsf{\delta }}_{\mathrm{X}}}_{33}\right)$$

(4)

Through the deformation formula, the moderator effect equation can be expressed as

$${\mathrm{x}}_{22}{\mathrm{x}}_{31}={{\mathsf{\lambda }}_{\mathrm{X}}}_{2231}{\mathsf{\xi }}_2{\mathsf{\xi }}_3+{{\mathsf{\lambda }}_{\mathrm{X}}}_{22}{\mathsf{\xi }}_2{{\mathsf{\delta }}_{\mathrm{X}}}_{31}+{{\mathsf{\lambda }}_{\mathrm{X}}}_{31}{\mathsf{\xi }}_3{{\mathsf{\delta }}_{\mathrm{X}}}_{22}+{{\mathsf{\delta }}_{\mathrm{X}}}_{2231}$$

(5)

$${\mathrm{x}}_{12}{\mathrm{x}}_{34}={{\mathsf{\lambda }}_{\mathrm{X}}}_{1234}{\mathsf{\xi }}_1{\mathsf{\xi }}_3+{{\mathsf{\lambda }}_{\mathrm{X}}}_{12}{\mathsf{\xi }}_1{{\mathsf{\delta }}_{\mathrm{X}}}_{34}+{{\mathsf{\lambda }}_{\mathrm{X}}}_{34}{\mathsf{\xi }}_3{{\mathsf{\delta }}_{\mathrm{X}}}_{12}+{{\mathsf{\delta }}_{\mathrm{X}}}_{1234}$$

(6)

$${\mathrm{x}}_{11}{\mathrm{x}}_{32}={{\mathsf{\lambda }}_{\mathrm{X}}}_{1132}{\mathsf{\xi }}_1{\mathsf{\xi }}_3+{{\mathsf{\lambda }}_{\mathrm{X}}}_{11}{\mathsf{\xi }}_1{{\mathsf{\delta }}_{\mathrm{X}}}_{32}+{{\mathsf{\lambda }}_{\mathrm{X}}}_{32}{\mathsf{\xi }}_3{{\mathsf{\delta }}_{\mathrm{X}}}_{11}+{{\mathsf{\delta }}_{\mathrm{X}}}_{1132}$$

(7)

$${\mathrm{x}}_{21}{\mathrm{x}}_{33}={{\mathsf{\lambda }}_{\mathrm{X}}}_{2133}{\mathsf{\xi }}_2{\mathsf{\xi }}_3+{{\mathsf{\lambda }}_{\mathrm{X}}}_{21}{\mathsf{\xi }}_2{{\mathsf{\delta }}_{\mathrm{X}}}_{33}+{{\mathsf{\lambda }}_{\mathrm{X}}}_{33}{\mathsf{\xi }}_3{{\mathsf{\delta }}_{\mathrm{X}}}_{21}+{{\mathsf{\delta }}_{\mathrm{X}}}_{2133}$$

(8)

The weight relation equation of the moderator variable is defined as

$${\widehat{\mathsf{\epsilon }}}_4={\mathsf{\omega }}_{\mathsf{\xi }41}{\mathrm{x}}_{22}{\mathrm{x}}_{31}+{\mathsf{\omega }}_{\mathsf{\xi }42}{\mathrm{x}}_{12}{\mathrm{x}}_{34}+{\mathsf{\omega }}_{\mathsf{\xi }43}{\mathrm{x}}_{11}{\mathrm{x}}_{32}+{\mathsf{\omega }}_{\mathsf{\xi }44}{\mathrm{x}}_{21}{\mathrm{x}}_{33}$$

(9)

5. Empirical Implementation

This paper mainly adopts data from field surveys which cover three major urban agglomerations in Zhejiang Province. First, as the nation’s first Demonstration Zone for Common Prosperity, Zhejiang demonstrates remarkable development balance. The income ratio between urban and rural residents stands at 1.96—significantly lower than the national average of 2.56. Additionally, the income ratio between residents in its highest- and lowest-income cities is 1.67. Notably, it is the only province in China where the per capita income of all prefecture-level cities exceeds the national average. Second, as one of the regions with the most developed private economy, Zhejiang fosters a strong entrepreneurial atmosphere. Through the synergistic integration of cultural genes, policy innovation, industrial ecology, and resource endowments, the province has established the regional advantages of “low threshold and high tolerance” for the entrepreneurship of underclass groups. Field survey targets include underclass groups under active urbanization and underclass groups under passive urbanization. The underclass groups under active urbanization include a new generation of migrant workers under ex situ urbanization and returning migrant workers under in situ urbanization. The underclass groups in passive urbanization are mainly under in situ urbanization. This study focuses on the formation of intention–behavior. Underclass groups with entrepreneurial intention are set as the core research subjects of this model. To avoid sample bias, the samples selected in this study cover both the “intentioned but non-acting” groups and the “acting” groups. Before the formal questionnaires were distributed, small-scale interviews and pre-tests were conducted in a small sample to allow potential respondents to design and evaluate them (including criticisms and suggestions for questions, content, wording, sequence, form, and layout). In this way, a pilot questionnaire was designed, and the participants then gave feedback on its operability. The initial questionnaire was distributed in Jiubao town, Hangzhou city (community shops, merchants in shopping malls, etc.). The formal questionnaire was distributed in Hangzhou, Ningbo, Yiwu, Wenzhou, and other places. The questionnaire item design is presented in

Table 3. Questionnaire item design.

Category Name	Variable Name		Corresponding Survey Questions
Entrepreneurship behavior	Entrepreneurship behavior (EB)		EB1: I have started a business. EB2: I have always worked for someone else, never started a business. EB3: I have other jobs while starting my own business.
Entrepreneurship intention	Perceived desirability (PD)	Achievement motivation (AM)	AM1: My desire to start a business is very strong. AM2: I truly want to become an entrepreneur.
		Innovation orientation (IO)	IO1: I accept new things quickly and I am suited for entrepreneurship. IO2: Innovation is necessary to entrepreneurship.
	Perceived feasibility (PF)	Social capital (SC)	SC1: I can obtain entrepreneurship resources from my friends and family. SC2: I can build close relationships with potential customers.
		Market opportunities (MO)	MO1: I think market conditions are ripe and there is a potential market space for entrepreneurship. MO2: I think I will start a business when I find market opportunities.
Land expropriation situation	Land expropriation (LE)	Land expropriation mode (LM)	The way expropriated land is developed will affect entrepreneurship.
		Resettlement mode (RM)	Settlement mode will have an impact on entrepreneurial behavior.
		Compensation mode (CM)	There is a need for adequate compensation to reduce the financial burden on entrepreneurs.
		Complementary policy (CP)	Guidance and training will promote entrepreneurship.

(Source: authors’ own creation.)

. It is designed in the form of a Likert scale with five ordered response levels, and options are presented according to the degree of agreement. “Strongly agree” is equal to five points, “agree” is equal to four points, “neutral” is equal to three points, “disagree” is equal to two points, and “strongly disagree” is equal to one point. Before the former questionnaire was administered, 20 underclass groups agreed to participate in the pre-test to provide opinions on the meaning and grammatical expression of the questions, and the questionnaire was then modified based on this feedback. The questionnaire consists of 15 measurement entries. The distributed formal questionnaire was mainly in paper form. In the field survey, 500 questionnaires were distributed, and 486 were collected (a collection rate of 97.2%). Eighteen questionnaires with undesirable answers were deemed invalid. There were 468 valid questionnaires, accounting for more than 96.3% of the total number. We found a minimal gender gap among the survey respondents. Most of the respondents were under 50 years old. Their education level was generally no higher than high school; this finding implies that underclass groups do not receive enough education. The average annual family income was less than CNY 100 thousand. The underclass groups in Hangzhou, Yiwu, and Wenzhou have relatively high incomes related to urbanization. However, as consumption in these three cities is relatively high, some underclass people still live in poverty.

Table 4. Descriptive analysis of individual factors.

Characteristic Variable	Frequency (%)
Gender
Female	237 (50.6)
Male	231 (49.4)
Age
≦30	148 (31.6)
31–40	143 (30.6)
41–50	126 (26.9)
51–60	51 (10.9)
Education
Primary school and below	44 (9.4)
Middle school	151 (32.3)
High school	177 (37.8)
University and above	96 (20.5)
Income
≦20,000	40 (8.5)
20,000–30,000	94 (20.1)
30,000–50,000	131 (28)
50,000–100,000	138 (29.5)
≧100,000	65 (13.9)

(Source: authors’ own creation.)

shows the descriptive analysis of the interviewees’ background information.

Table 5. Descriptive analysis of indicator variables.

Indicator Variables	Frequency (%)
	Strongly Disagree	Comparatively Disagree	Neutral	Comparatively Agree	Strongly Agree
EB1	32 (6.84)	100 (21.37)	178 (38.03)	114 (24.36)	44 (9.4)
EB2	26 (5.6)	96 (20.5)	163 (34.8)	131 (28)	52 (11.1)
EB3	65 (13.9)	165 (35.3)	144 (30.8)	63 (13.5)	31 (6.6)
AM1	17 (3.6)	51 (10.9)	108 (23.1)	167 (35.7)	125 (26.7)
AM2	16 (3.4)	61 (13)	131 (28)	129 (27.6)	131 (28)
IO1	37 (7.9)	74 (15.8)	120 (25.6)	148 (31.6)	89 (19)
IO2	22 (4.7)	48 (10.3)	79 (16.9)	146 (31.2)	173 (37)
SC1	36 (7.7)	79 (16.9)	123 (26.9)	135 (28.8)	95 (20.3)
SC2	33 (7.1)	88 (18.8)	148 (31.6)	138 (29.5)	61 (13)
MO1	31 (6.6)	67 (14.3)	153 (32.7)	149 (31.8)	68 (14.5)
MO2	46 (9.8)	87 (18.6)	153 (32.7)	122 (26.1)	60 (12.8)
LM	13 (2.8)	52 (11.1)	149 (31.8)	190 (40.6)	64 (13.7)
RM	54 (11.5)	138 (29.5)	183 (39.1)	81 (17.3)	12 (2.6)
CM	53 (11.3)	133 (28.4)	185 (39.5)	74 (15.8)	23 (4.9)
CP	32 (6.8)	71 (15.2)	131 (28)	135 (28)	99 (21.2)

(Source: authors’ own creation.)

presents the descriptive findings of the indicator variables, including entrepreneurship behavior, achievement motivation, innovation orientation, social capital, market opportunities, land expropriation mode, resettlement mode, compensation mode, and complementary policy.

Entrepreneurship behavior refers to the cognition or judgement of whether entrepreneurial behavior occurs. To measure the status of entrepreneurship behavior, three questions were designed, as shown in Table 3. We conduct a preliminary statistical analysis. Table 5 shows that the underclass groups have extremely strong entrepreneurial intentions in the survey area, but they are more inclined to work than to engage in entrepreneurship. If they start a business, most of them present a low service industry and subsistence-type entrepreneurship, which serves to satisfy the needs of their families and improve their quality of life. The businesses involved include fruit and vegetable wholesale, catering, and logistics businesses. Their entrepreneurial success rate and level are not satisfactory.

Entrepreneurial intention includes perceived desirability and perceived feasibility, which affect whether entrepreneurial behavior occurs. To measure the status of perceived desirability and perceived feasibility, eight questions were designed, as shown in Table 3. Table 5 reveals that the underclass groups have strong perceived desirability (including achievement motivation and innovation orientation) and perceived feasibility (including social capital and market opportunities) whether or not they have started a business. A field survey reveals that the underclass groups with strong perceived desirability and perceived feasibility tend to have a more positive attitude towards entrepreneurship. In other words, those underclass persons who “try to make things better”, “know better and adapt to circumstances”, and “dare to take an adventurous approach” remain aware of market information and are good at making full use of interpersonal network resources, have success stories in entrepreneurship, or carry out large-scale entrepreneurial activity, although some of them experience business failure. For example, the village of the “flower landscape” led the development of the tourism industry; many people in Fuyang, Xiaoshan, and Tonglu in Zhejiang Province like to visit, and bars, bed and breakfasts (B&Bs), and farmhouses have sprung up one after another.

Land expropriation includes land expropriation mode, resettlement mode, compensation mode, and complementary policy, as shown in Table 3. Among them, the land expropriation mode refers to the distance from the location of the expropriated land to the main urban zone (such as an urban village, an urban fringe, or a remote rural area) and the development type of the expropriated land (such as a village in a tourist attraction, a resort, or an economic or technical development zone). The resettlement mode refers to in situ settlement or housing settlement for underclass groups, which can make houses rentable and able to generate rental income. The compensation mode is largely based on the location of land expropriation and different types of land expropriation projects. Differences in the level of regional economic development also influence the amount of compensation. Complementary policies, including credit financing, platform construction, entrepreneurship training, and related services, are relevant for underclass groups’ entrepreneurship. To measure the status of land expropriation, four questions were designed, as shown in Table 3. Table 5 shows that the underclass groups are more aware of the impact of the land expropriation mode and complementary policy on entrepreneurship than that of the resettlement and compensation modes.

6. Variable Measurement

The reliability and validity of the data structure are analyzed via exploratory factor analysis (EFA) and confirmatory factor analysis (CFA). EFA and CFA are two inseparable and important components of factor analysis. In developing the theory, the model should be established via EFA; then, the concept and calculation tools of the analytical model are provided to verify and correct it. The results provide an important basis and assurance for the establishment of hypotheses for CFA [91].

6.1. Reliability Analysis

The reliability and stability of the data are the basis for exploratory factor analysis and SEM checking. In other words, the reliability test is the basis for subsequent data analysis. Cronbach’s alpha values and composite reliability are important indices of reliability analysis. First, Cronbach’s alpha values are adopted for reliability analysis of the data via EFA. It is generally believed that values ranging between 0.70 and 0.80 are good and values ranging between 0.80 and 0.90 are very good. Through testing, it is found that all the Cronbach’s α values of the factors are above 0.7 in this paper, as shown in

Table 6. The reliability test results of questionnaire variables.

Latent Variables	Measure Item	Standardized Indicator Loading	Non-Standardized T Value	Cronbach’s α	Composite Reliability
EB	EB1	0.637	-	0.823969	0.723148
	EB2	0.861	10.378
	EB3	0.529	7.832
PD	AM1	0.784	-	0.904667	0.819297
	AM2	0.775	16.221
	IO1	0.646	13.481
	IO2	0.706	14.789
PF	SC1	0.627	-	0.914188	0.797314
	SC2	0.712	12.093
	MO1	0.746	12.489
	MO2	0.729	12.306
LE	LM	0.638	-	0.770431	0.634943
	RM	0.503	7.008
	CM	0.647	7.153
	CP	0.519	2.539

(Source: authors’ own creation.)

. Second, composite reliability (construct reliability) is measured for the intrinsic quality of the latent variables via CFA. If the composite reliability of potential variables is above 0.6, the model has ideal intrinsic quality. In CFA testing, the composite reliability values of entrepreneurship behavior, perceived desirability, perceived feasibility, and land expropriation all exceed 0.6 (Table 6). The results of the reliability analysis prove that each scale has high reliability and that the variables have good internal consistency.

6.2. Validity Analysis

After the reliability of the data is guaranteed, the consistency between the results of the measurement data and the expected results should be measured.

First, partial samples are randomly selected from the total samples. The KMO sample test and Bartlett’s test of sphericity are adopted, and EFA is carried out to verify the construct validity of the data. As shown in

Table 7. The inspection with KMO and Bartlett’s test.

Kaiser–Meyer–Olkin (KMO) Measure of Sampling Adequacy		0.857
Bartlett’s Test of Sphericity	Approx. Chi-Square	1665.846
	df	105
	Sig.	0.000

(Source: authors’ own creation.)

, the KMO value is 0.857. When this value exceeds 0.7, the model is generally assumed to be suitable for factor analysis, whereas a KMO value below 0.5 is not suitable. The p-value of Bartlett’s test is 0.000. A p-value less than 0.01 indicates that the null hypothesis at the 1% significance level is rejected. The correlation coefficient matrix of the set cannot be a unit matrix; that is, there is a correlation between the original variables. Therefore, the questionnaire has structural validity and is suitable for factor analysis. By using principal component analysis and the variance maximum method (

Table 8. Rotated factor loading matrix.

Measure Variables	1	2	3	4	5
EB1	0.182	0.121	0.804	0.269	−0.041
EB2	0.154	0.212	0.866	0.130	0.014
EB3	0.065	0.000	0.832	−0.094	0.099
AM1	0.765	0.319	0.178	0.084	−0.015
AM2	0.843	0.186	0.127	0.067	−0.010
IO1	0.770	0.281	0.023	0.070	0.142
IO2	0.755	0.210	0.147	0.130	0.066
SC1	0.206	0.813	0.052	0.228	−0.009
SC2	0.232	0.853	0.111	0.015	0.080
MO1	0.431	0.737	0.083	0.037	0.097
MO2	0.420	0.670	0.232	0.117	0.218
LM	0.244	0.334	0.196	0.652	0.144
RM	0.060	0.134	0.052	0.276	0.708
CM	0.053	0.042	0.015	−0.045	0.851
CP	0.060	0.024	0.057	0.877	0.085
Eigenvalues	5.763	1.795	1.388	1.022	1.013
Variance interpretation%	38.417	11.965	9.255	6.816	6.754
Cumulative variance interpretation%	38.417	50.382	59.637	66.453	73.206

(Source: authors’ own creation.) Note: The extraction method is principal component analysis. The rotation converged in 5 iterations by using the varimax method with Kaiser normalization. The bold-faced font indicates that the rotated factor loading is greater than 0.5, which shows that the item has a strong correlation with the factor it belongs to, can well reflect the characteristics of this factor, and has good validity.

), five indicators with eigenvalues greater than 1 are extracted, and 73.206% of the total variation is explained. Each indicator is clearly loaded on the expected indicators, the load is above 0.652 (above 0.5), and the indicator structure is clear. This proves that the data can be subjected to CFA.

Second, after the EFA is verified, CFA is carried out for further validity analysis. The other remaining samples are used for CFA. The normalized factor loading at 0.619–0.838 reaches a significant level (p < 0.001) and shows good convergence. The factor loads corresponding to the individual measurement items are 0.619 or greater and can be used to effectively measure the potential factors of the overall survey data (Table 8). This proves that the data structure in this study is valid for building the EBUG model.

7. Results Analysis

In this paper, the EBUG model is analyzed via a two-step path to test the hypothetical theoretical framework. In the first step, when the EBUG model does not contain the product indicator of the moderator variable, Model I is constructed for the impact of each major effect variable (perceived desirability, perceived feasibility, and land expropriation) on entrepreneurship behavior. In the second step, when the EBUG model contains the product indicator of the moderator variable, Model II is constructed for the impact of the major effect and interaction effect variables (MV) on entrepreneurship behavior. The two models are debugged, identified, and tested through maximum likelihood estimation on Amos 24. The similarity goodness indices of the two models are shown in

Table 9. Fitness analysis results of models.

Fit Index	Standard	Model
		Model I	Model II
CMIN/DF	<2.0	1.388	1.025
RMSEA	<0.05	0.029	0.007
IFI	>0.90	0.995	0.999
NNFI	>0.90	0.982	0.998
CFI	>0.90	0.995	0.999
CN	>200	491	573

(Source: authors’ own creation.) Note: (a) CMIN/DF = ${\mathsf{\chi }}^2/\mathrm{d}\mathrm{f}$ (the ratio of chi-square to the degrees of freedom). Carmines and McIver [92] argued that the conditions under the inspection should be less than 2.0. (b) RMSEA = root mean square error of approximation. Browne and Cudeck [93] argued that the index should be less than 0.1, and the smaller the better, with less than 0.08 for better fitting and less than 0.05 for very good fit. (c) IFI = incremental fit index. L. T. Hu and Bentler [94] argued that the conditions under the inspection should be greater than 0.9. (d) NNFI = non-normed fit index (=Tacker–Lewis index, TLI). Bentler [95] argues that the condition under inspection should be greater than 0.9. (e) CFI = comparative fit index; (f) CN = critical N. Hoelter [96] suggested that the indicator value should be greater than 200. Hu and Bentler [97] argued that the model should be able to accept a minimum CN of 250.

.

In Model I, the chi-square value of the path model is 41.647, and the value of the significance probability is 0.077 (above the significance level of 0.05). The null hypothesis is accepted, which indicates that the theoretical model can be consistent with the sample data. Table 9 shows that the CMIN/DF is 1.388 and below 2. The value of the RMSEA is 0.029, which is below 0.05. The value of the IFI is 0.995, the value of the NNFI is 0.982, and the value of the CFI is 0.995, which are all greater than 0.900. The value of CN is 491, which is greater than 200. The fitness analysis results of Model I indicate favorable consistency between the hypothesis model and the actual sample data, which indicates that H1 and H2 are supported. In Model II, the chi-square value of the path model is 92.226, and the value of the significance probability is 0.415 (above the significance level of 0.05). The null hypothesis is accepted, which indicates that the theoretical model can be consistent with the sample data. Table 9 shows that the CMIN/DF is 1.025 and below 2. The value of the RMSEA is 0.007, which is below 0.05. The value of the IFI is 0.999, the value of the NNFI is 0.998, and the value of the CFI is 0.999, which all exceed 0.900. The value of CN is 573, which exceeds 200. The fitness analysis results of Model II indicate favorable consistency between the hypothesis model and the actual sample data. Moreover, the values of CMIN/DF and RMSEA of Model II are below those of Model I, and the values of IFI, NNFI, CFI, and CN of Model II are greater than those of Model I. The above comparisons show that the simulation fitness of Model 2 is better than that of Model 1, which supports H3.

Further verifying the role of the moderator variable, as shown in

Table 10. Result analysis (regression coefficients and R²).

Model	Proposed Relationship	Effect Type	Estimate	S.E.	C.R.	R2	Study Results
Model I	EB←PD	Direct effect	0.147 ***	0.042	3.473	0.182	Supported
	EB←PF	Direct effect	0.271 **	0.092	2.965		Supported
	EB←LE	Direct effect	0.238 ***	0.048	4.945		Supported
Model II	EB←PD	Direct effect	0.112 **	0.042	2.700	0.297	Supported
	EB←PF	Direct effect	0.235 **	0.076	3.091		Supported
	EB←LE	Direct effect	0.343 ***	0.059	5.774		Supported
	EB←MV	Direct effect	0.505 ***	0.07	7.196		Supported

(Source: authors’ own creation.) Note: *** p < 0.001; ** p < 0.01.

, all the direct hypotheses are supported. In Model I, all the major effect variables of perceived desirability (0.147 ***, p < 0.001), perceived feasibility (0.271 **, p < 0.01), and land expropriation (0.238 ***, p < 0.001) are found to have a significant positive influence on entrepreneurship behavior, supporting H1 and H2. Thus, perceived desirability is found to have greater influence than perceived feasibility. This finding indicates that when the EBUG model does not contain the product indicator of the moderator variable, the EBUG is more affected by perceived desirability than by perceived feasibility. Moreover, the underclass groups’ entrepreneurship is more inclined to be of the opportunistic type (where one aims to pursue one’s own development, such as accomplishment, honor, and innovation) because of the impact of achievement motivation and innovation orientation. In Model II, all the major effect variables of perceived desirability (0.112 **, p < 0.01), perceived feasibility (0.235 **, p < 0.01), and land expropriation (0.343 ***, p < 0.001) and the interaction effect variables of MV (0.505 ***, p < 0.001) are also found to have a significant positive influence on entrepreneurship behavior, supporting H1, H2, and H3. By comparing the two models, we find that the value of R² increases by 0.115 (Model I: R² = 0.182; Model II: R² = 0.297), which indicates that land expropriation has a significant moderating effect on the relationship between entrepreneurial intention and entrepreneurial behavior. Notably, perceived desirability is found to have less influence than perceived feasibility. This finding indicates that when the EBUG model contains the product indicator of the moderator variable, the EBUG is more affected by perceived feasibility than by perceived desirability, which is caused by the addition of the moderator variable. In addition, underclass groups’ entrepreneurship is more inclined towards the subsistence type (where one aims to meet one’s own survival needs) because of the impact of social capital and market opportunities.

8. Conclusions and Discussion

In this study, the EBUG model is constructed and analyzed according to a theoretical analysis framework for the EBUG in urbanization in China. A structural equation model with a moderator variable is applied to construct the EBUG model. This involves major effect variables (perceived desirability, perceived feasibility, and land expropriation) and interaction effect variables (MV). The two models are compared to verify the theoretical assumptions and analyze the influence and intrinsic relationships of perceived desirability, perceived feasibility, and land expropriation on entrepreneurship behavior. The results provide a decision-making reference for underclass groups’ development and livelihood in urbanization in China. We draw the following conclusions from the results: (1) Through the comparison of the two models (Model I does not contain the product indicator of the moderator variable, and Model II contains the product indicator of the moderator variable), we verify the theoretical hypotheses. Perceived desirability, perceived feasibility, and land expropriation all have a significant positive influence on entrepreneurial behavior, supporting H1, H2, and H3. (2) The moderator variable of the EBUG model significantly influences the relationship between entrepreneurial intention and entrepreneurial behavior. The effect of the moderator variable is reflected in two main aspects. First, land expropriation has a significant moderating effect on entrepreneurial intention to engage in entrepreneurship behavior. This study shows that the EBUG model with the moderator variable has a more significant influence than the EBUG model without the moderator variable. Second, land expropriation as a moderator variable has an effect on the path of entrepreneurial intention to entrepreneurship behavior. The study demonstrates that underclass groups’ opportunistic entrepreneurship is affected mainly by perceived desirability when land expropriation, a moderator variable, is not considered. In the survey, we find that opportunistic-type entrepreneurs might start a business before land expropriation or not face land expropriation problems. The purpose of their entrepreneurship is to pursue their achievement motivation and innovation orientation, which belong to perceived desirability. In contrast, underclass groups’ entrepreneurship of the subsistence type is affected mainly by perceived feasibility when land expropriation, as a moderator variable, is considered. In the survey, we find that subsistence-type entrepreneurs might start a business after land expropriation, which is forced on them by life pressures, such as landlessness, unemployment, and lack of income. Their entrepreneurship mainly considers social capital and market opportunities, which are aspects of perceived feasibility. Therefore, although land expropriation as a moderator variable has a significant positive effect on the results of the analysis, the change in different types of land expropriation is due to the negative impact of land expropriation.

The underclass groups in the study area have very strong entrepreneurial intentions, but most of them present subsistence-type entrepreneurship, so their entrepreneurial success rate and level are not satisfactory. Different types of entrepreneurship and their influencing factors should be distinguished. This study proposes three sustainable development paths for the shift from subsistence-type to opportunistic-type entrepreneurship for underclass groups.

(1) Promoting opportunistic-type entrepreneurship among underclass groups through perceived desirability. Opportunistic-type entrepreneurship should focus on underclass groups’ perceived desirability (including achievement motivation and innovation orientation). This type of entrepreneurship is not common among underclass groups. Most underclass groups are poorly oriented, with low entrepreneurial quality. Therefore, the local government should enhance guidance for the entrepreneurship of underclass groups from the perspective of achievement motivation and innovation orientation and promote their entrepreneurship to transform it from the subsistence type to the opportunistic type. First, skill conversion-based support in urban community services can reduce entry barriers by establishing rent-free entrepreneurial spaces and order-matching mechanisms, prioritizing sectors like home care, neighborhood maintenance, childcare, and elderly care. Second, financial innovation mechanisms should integrate credit scoring loans (CNY 50,000–CNY 200,000 collateral-free loans based on employment tenure, vocational certifications, and social security contributions), pilot Entrepreneurship Failure Compensation Insurance (providing six months of minimum living allowance after one year of operational failure), and subsidized support (up to 50% equipment leasing subsidies and government-mediated procurement). Finally, digital empowerment initiatives could scale impact through live-streaming e-commerce training, connecting entrepreneurs with commodity supply chains (e.g., Yiwu and Linyi), and tiered logistics subsidies.

(2) Promoting subsistence-type entrepreneurship among underclass groups through perceived feasibility. Subsistence-type entrepreneurship should strengthen underclass groups’ perceived feasibility (including social capital and market opportunities). Underclass groups face structural constraints such as limited cognitive resources, restricted access to market intelligence, and underdeveloped entrepreneurial literacy. Coupled with a lack of entrepreneurial experience, they may engage in poorly conceived entrepreneurial behavior, easily make bad decisions, miss opportunities, and fail to grasp the laws of the market or identify entrepreneurial opportunities. Therefore, it is recommended to establish an entrepreneurial resilience-building platform, integrating an “Entrepreneurship Clinic” staffed with advisory panels comprising entrepreneurs, accountants, and legal professionals to provide monthly pro bono diagnostic sessions. Through education and communication platforms, the perceived feasibility of entrepreneurship for underclass groups can be enhanced. Education should consist of pretraining selection and entrepreneurial support. For those who do not have entrepreneurial intention, it is advised to clarify employers’ demands on the labor market and then develop job training programs, confirm employment training content, and promote employment in the job market. For underclass groups with entrepreneurial intention, entrepreneurial guidance and training should embody relevance and hierarchy. Given that underclass people differ in terms of age, gender, and cultural level, training for expertise and technical professions with scientific and technological content, skills training for community services, and skills training for cottage crafts in a differentiated manner are advised. In addition, according to the survey, village heads’ social capital is greater than ordinary villagers’ social network resources. Therefore, underclass groups with broader social capital have a stronger sense of entrepreneurship, enjoy more entrepreneurial opportunities, and establish larger enterprises. This demonstrates that local governments should mobilize social capital by implementing a “Village Sage Mentorship System”, where each successful entrepreneur is assigned to support five entrepreneurial ventures. The government should provide corresponding tax deduction incentives to facilitate this initiative.

(3) Promoting the transition from subsistence-type to opportunistic-type entrepreneurship through the land expropriation policy. It is necessary to alleviate the negative influence of land expropriation on the underclass groups involved in urbanization. The settlement mode, the amount of compensation, and other social security factors can play a positive role in promoting underclass groups’ entrepreneurship but can also have a negative influence. According to the survey, on the one hand, if an underclass group has a larger settlement area, it is more likely to choose accommodations, catering, and other industries for subsistence-type entrepreneurship. If the settlement area is located in an economically developed area with a good external market, it will attract more nonlocal tenants. However, the simple renting of vacant houses to tenants apparently gradually decreases market competitiveness, leading to failure in entrepreneurship. On the other hand, underclass people who obtain compensation fees for land expropriation and “make a great fortune overnight” are characterized by individual differences. For those who are good at financial management, these fees can be efficiently utilized and play a positive role in their entrepreneurial success. For those conservative underclass groups who lack funds and entrepreneurship ability, it is difficult to find suitable entrepreneurship projects, and their lives can become even more difficult. A few people live wantonly and decadently. Such simple security cannot ensure underclass people’s future livelihood. Therefore, guarantee-type policies for “blood transfusion” should make way for opportunistic-type policies for “development” or “blood generation”. It is recommended that local governments encourage the implementation of localized entrepreneurship for resource development, taking into account the characteristics of the underclass groups of passive urbanization and the land expropriation policy situation. First, it is necessary to carry out securitization of collective assets, and convert land expropriation compensation into a village collective entrepreneurship fund, giving priority to the support of the formation of cooperatives (e.g., organic vegetable cultivation and cold chain storage), villagers with capital/land shares, and government-supported subsidized loans. Secondly, it is necessary to perform the multifunctional utilization of rural residential land, as well as allowing the transformation of unused farmhouses to develop B&Bs and handicraft workshops, simplifying the fire approval process, and giving a CNY 20,000 subsidy for each transformed room. Finally, it is necessary to cultivate localized new economic models, for example, roof photovoltaic entrepreneurship, and promote the “farmers out of the roof + enterprises out of the equipment + power generation revenue sharing” model, with grid companies’ guaranteed tariff purchase. Another example is becoming a carbon trading intermediary, training forest land carbon sink measurement skills, and set up specialized service teams to undertake carbon neutral businesses.

This paper explores and gains insights into the influencing mechanism and functional paths of the EBUG involved in urbanization and provides a reference for promoting it. This study also has certain limitations. In the future, we will adopt longitudinal tracking design or cross-sectional mixed-sample design methods, dividing the samples into “intentioned but non-acting” and “acting” groups. By analyzing the differences in variables like perceived behavioral control and subjective norms and clarifying the applicable scope of the conclusions, the research depth will be further enhanced.