Factors Influencing Rural Households’ Decision-Making Behavior on Residential Relocation: Willingness and Destination

Abstract

All the traditional models of centralized residence based on “building a new socialist countryside” and “maintaining a balance between the increase and the decrease” are top-down in nature and require farmers to make responses and readjustment to all possible policies and changes. Therefore, it’s important to understand farmers’ preferences and take their willingness and needs into account when designing and implementing the relative planning programs of centralized residence. In this paper, with the numerical value 10 as the criterion of Events Per Variable (EPV) and Variance Inflation Factor (VIF), four different types of binary logistic regression were respectively applied to analyze factors that may influence farmer households’ relocation willingness and relocation destination in the following five aspects: Individual characteristics, household characteristics, housing characteristics, farmland characteristics, and implementation environment of centralized residence. As indicated in the results, people would show more willingness to relocate when they were younger, had higher household income, lived in an older building, possessed a bigger building area, owned farmland with higher quality, or lived in an environment with a higher infrastructure match rate. In addition, household income was a common factor influencing households’ choice between nearby relocation sites (NRS) and urban areas as their relocation destinations. The building area and occupancy rate negatively affected households’ choice of NRS, while building age negatively affected that of urban areas. Based on these influencing factors, some policy suggestions are proposed in this paper in terms of job creation, implementation of zoning and classification strategies, improvement of the quality of land transfer services, and reconstruction of the rural landscapes.

1. Introduction

Rural China has undergone a rapid and far-reaching socioeconomic and spatial transformation in the past years since 1978 [1,2,3], in the process of which significant changes took place in the rural living environment based on farmland and rural residential land [4,5]. The transition problems of farmland and rural residential land have received extensive attention from many scholars [6,7]. As the urbanization rate increased from 10.64% in 1978 to 56.1% in 2015, a large number of local farmers flooded from rural areas into urban areas for a better life, resulting in the loss and abandonment of farmland, which may affect national food security. Meanwhile, along with a massive population exodus from rural areas, many rural residential lands were left for the elderly to live in or remained unoccupied either seasonally or permanently, which were respectively known as “population aging” [8] and “Hollow Village” [9]. Additionally, under the constraints of the long-term urban-centered development strategy and the dual urban–rural system, many negative phenomena have emerged in the rural living environment, including the scattered and disorderly layout of villages, insufficient infrastructure and public service facilities, unsightly village appearance, and environmental pollution [3,10]. All these problems together have caused the overall decline of rural areas, becoming great challenges to realize sustainable rural development and rural revitalization.

Since the beginning of the new century, rural restructuring, which has already occurred in Europe [11,12,13], America [14,15], India [16], and Canada [17], has also taken place in China rural areas and has become an important part of global social and economic change [18]. An increasing number of scholars in different academic fields have explored the process of rural restructuring and relative rural issues in China [18,19], which mainly include the background, concept, and basic theories of rural restructuring [19,20], the relationship between land use transitions and rural restructuring [21,22], the pattern and process of rural restructuring [23], the impact mechanism of land use policy [1], and village relocation and spatial restructuring [2,24]. On the whole, rural restructuring is a systematic project that involves not only the construction of material elements supported by farmland and rural residential land but also that of non-material elements such as society, economy, and governance [19,20]. In this context, the Chinese government has gradually realized the importance and complexity of rural development and adopted various policies related to centralized residence [25], which can also be called concentrated rural resettlement [24,26] or community-based concentration [27]. In 2005, in order to narrow the gap between urban and rural areas, the strategy of “building a new socialist countryside” was proposed by the Chinese central government to build new concentrated residence communities (NCRC) and improve the quality of the rural living environment. Meanwhile, facing the conflict between the decrease of farmland and the increase of construction land, “the land use policy of “maintaining a balance between the increase and the decrease” was also proposed to optimize the layout of construction land through land consolidation and centralized residence. Since then, according to different types of rural planning, there has been a boom of resettlement and relocation in different regions of China through village-to-village, village-to-town, village-to-city, or peri-urban patterns [1,24,25], such as the “Beautiful Village Construction” in Zhejiang Province and “Beautiful City and Countryside Construction” based on “Town and Village Layout Planning (TVLP)” in Jiangsu Province.

However, on the whole, the pattern of centralized residence has played an important role in dealing with “Hollow Village”, alleviating rural poverty, optimizing land-use patterns, and realizing rural revitalization. Scholars held different opinions about the practice of centralized residence and relative relocation policies [24,25,26,27]. Some scholars argued that the current policies related to the centralized residence had many shortcomings, such as the lack of systematic spatial planning [28], disregard for farmers’ interests and future livelihood [27], lack of respect for farmers’ willingness [25,29] and the loss of traditional rural settlement culture [19]. Some scholars insisted that resettlement and relocation should be the last resort [30,31]. Some scholars believed that concentration might not be a panacea for rural planning due to the complexity and uncertainty of rural areas [24,27]. Almost all scholars considered that all the policies or strategies mentioned above are top-down in nature [10,27,32] and require the participation of relevant stakeholders, including farmers themselves [24,27], with particular attention to respecting their willingness and future needs [10,25,26,30,31,33]. Meanwhile, rural restructuring, including concentrated residence, is a complex systematic project that also requires farmers to make a response and readjustment to all possible policies and changes [19]. Therefore, it’s important to know farmers’ preferences and take their willingness and needs into account when designing and implementing the relative planning programs of centralized residence.

Farmers’ behaviors and preferences of decision making are generally regarded as a complex process of bounded rationality [34,35,36,37], in which the outcomes are usually affected by their own cognitive factors and environmental factors. Zhang et al. [25] explored the factors influencing rural households’ choice of centralized residence between pure and nonpure farming areas in terms of their individual characteristics, family economy, policy perception, housing conditions, and social environment. Other research on farmers’ preference for centralized residence usually do not exist alone but are symbiotic with specific contexts of related policies and planning practices, such as rural community remediation [38], geological disasters, and ecological relocation [30,39,40,41], “maintaining a balance between the increase and the decrease” [42], the act of moving away from rural homesteads [43], the relationship between poverty and relocation [31,44], rural settlements [3,10], and land consolidation [33,45]. For example, Sun et al. [45] investigated the factors influencing farmers’ decision-making behavior in rural construction land transformation based on various aspects containing personal characteristics, family characteristics, construction land conditions, compensation for construction land consolidation, and expectation of transferring to cities. Notably, the changes in employment and residence after land expropriation [46] and the people-centered practices embedded in policy, planning, and implementation [47], rather than resettlement or relocation itself, may be the decisive factors influencing farmers’ decision-making behavior of relocation.

Due to the strong effects of self-reinforcing agglomeration and geographical barriers [48], the northern area of Jiangsu Province has historically been a poverty-stricken region in this province [49], facing huge challenges in infrastructure construction, labor outflow, and land transfer [50]. Up to now, many studies have been conducted on the region in China. However, few studies considered it as a typical county facing serious “rural disease” problems as well as centralized residence problems and carried out research on farmers’ behaviors and preferences of decision-making from such a perspective. This paper selected Sihong County, a typical poor and agriculture-led region in the northern area of Jiangsu Province, as the study area to explore the factors influencing households’ relocation willingness and destination.

2. Overview of the Study Area

Sihong County is a typical county in the northern area of Jiangsu Province, China (Figure 1), where the level of economic development is relatively lower than that of the central and southern parts. The county features plains and hills topographically and boasts rich water resources such as Hongze Lake, Tiangang Lake, and Huai River. In early 2018, it had 14 towns and 9 townships, with an area of 2731 kilometers and a registered population of over 1.10 million (Figure 1). In late 2018, Qingyang Town was divided into Qingyang Street, Zhonggang Street, and Dalou Street, with the county seat located in the center. In the latest edition of urban system planning, Qingyang Town was positioned as the main body of central urban area, Shuanggou Town and Jieji Town, two sub-central urban areas, and Bancheng town, Meihua Town, and Shangtang Town major towns.

Since 2005, under the guidance of new rural construction, town and village layout planning, and new urbanization construction, “three-concentration” construction model has been implemented by the local government to promote the concentration of population, landholding, and industrial projects, leading to the continuous improvement of the county’s urbanization level and the continuous decline of rural areas. Between 2005 and 2018, its resident population registered population and urbanization level respectively increased from 0.27 million to 0.90 million, from 0.99 million to 1.10 million, and from 29.10% to 57.70%, indicating the growing imbalance between the number of registered population and resident population and a greater outflow of the population as a result under the background of rapid urbanization development. The fact that rural laborers went out to work in cities all year round was one of the most important reasons for the mass exodus. For example, in 2016, there were 0.38 million populations of working age, of which about 23.15% were still farming or unemployed in their local hometowns, and about 58.37% were migrant workers away from their hometowns. Additionally, as the population flowed out, more and more rural areas were declining and were left with large numbers of old people and children, thus resulting in the phenomenon known as “Hollow Village”.

In order to ensure orderly population flow and sustainable development of rural areas in the process of urbanization and rural revitalization, according to the unified deployment of Jiangsu Province, three versions of “Town and Village Layout Planning (TVLP)” were successively completed in 2005, 2014, and 2018. In the first version of TVLP proposed in 2005, 1997 original villages were incorporated into 242 new concentrated residence communities (NCRCs). Due to the failure to follow the laws of rural development and respect farmers’ willingness, this version did not achieve the desired planning target and caused some new problems instead, such as superfluous NCRCs, poor quality NCRCs, incomplete supporting facilities, and unattractive NCRCs. In the second version of TVLP put forward in 2014, under the requirement of a “bottom-up” planning idea, 1794 natural villages were classified into 75 key villages, 51 characteristic villages, 33 key and characteristic villages, and 1635 general villages. The new version not only overcomes major disadvantages of the previous one but also adjusts measures to local conditions of different villages and adopts different supervision measures correspondingly. In the latest version of TVLP presented in 2018, according to systematic requirements of rural revitalization strategy, 1355 natural villages were reclassified into five types of villages, including 65 agglomeration and promotion types, 55 characteristic protection types, 15 suburban integration types, 782 relocation and consolidation types, and 438 general types.

By late 2018, under the guidance of previous TVLPs, about 305 different types of relocation sites or NCRCs have been planned and built in rural or urban areas, 222 of which were key construction sites built by the local government between 2011 and 2018. As mentioned in the introduction part, there are three relocation patterns in Sihong County: village-to-village, village-to-town, and village-to-city. For the village-to-village pattern, a village can be identified as an NCRC if it boasts such advantages as large size and population, superior geographic location, convenient transportation, good public service, and high economic level based on TVLP. Compared with traditional villages, as shown in Figure 2, NCRCs feature a more modern architectural appearance and living environment that allow rural residents to enjoy infrastructures and public services similar to those of cities. The village-to-village relocation is widely carried out in the form of “house for house”. Farmers and their households are required to move to nearby relocation sites (NRS), i.e., nearby NCRCs, which are usually not too far from their old houses and have already been allocated by the local government, according to TVLP. They can get a new house with a certain floor area that is proportional to the area of their old houses and homesteads. After the relocation, most farmers can still use their farmlands for agricultural production or be employed by others to do so. For village-to-town and village-to-city patterns, the approach of “house for money” is widely adopted by farmers and their households to get compensation for demolition and buy new houses in other places they want to go, including the seat of town and township, seat of county, and other cities. After the relocation, most of them would prefer to transfer their farmlands to others and switch to the secondary and tertiary industries with high added value.

3. Methodology

3.1. Data Source

As part of the project of Sihong County’s 2014 and 2018 versions of TVLP, this study is organized as special research on “New Urbanization” from the perspective of farmers and their households, aiming to learn the current situation of rural development and its other possible problems. The questionnaire was designed based not only on existing related research and team’s advice but also on the county’s current problems of society, economy, and governance, such as Hollow Village” and “population outflow”. Under many teachers’ supervision and guidance, the questionnaire finally consisted of four aspects, i.e., basic information of individuals and their families, housing environment and relocation willingness, living facilities and travel behaviors, and farming conditions and transferring of farmland rights. To ensure the quality and accuracy of the questionnaires, between 10 and 16 January 2015, they were handed out randomly to local farmers in Meihua Town and Shuanggou Town and modified several times to better serve the purpose. Then, in the next three months around Chinese New Year, our team was divided into four groups to hand out the questionnaires in 14 towns and 9 townships. Given the characteristics of current farmer distribution and the small proportion (1‰) of rural households in different towns and townships, about 1034 interviewees, who were chosen proportionally from as many regions as possible to cover every administrative village, took part in the questionnaire survey as the representatives of their households. Considering the lower education level of farmers, semi-structured interviews, structured interviews, and mixed face-to-face interviews were applied flexibly by interviewers of our team to different types of farmers. For farmers who couldn’t read and understand our questions well, the interviewers would translate the questions into vernacular languages and give oral explanations. For others who want to talk more about their living conditions and relative possible problems, the interviewers would write these down and revise the questionnaires in time. Additionally, all the data in this paper, including geospatial data, socioeconomic data, and chart data, are from the project of Sihong County’s 2014 and 2018 versions of TVLP.

3.2. Variable Selection

Previous research explored the factors influencing farmers’ or householders’ relocation willingness, consisting of individual characteristics, household characteristics, housing characteristics, farmland characteristics, and relative policy environment. For example, Zhang and Han [38], Chen, et al. [44] proposed that individual characteristics, which can also be called demographic characteristic or basic information, should include age, gender, education, occupation, and so on, and there’s evidence that younger farmers showed a greater relocation willingness than older farmers. However, other relative scholars [25,39,41] argued that age, gender, and education were not significantly correlated with household relocation willingness. For household characteristics, household/family income played an important role in farmers’ behavior of decision making and positively influenced farmers’ relocation willingness among different regions [25], different groups [30], and different individuals/households [39,44,45]. However, Zhang and Han [38] proved that higher-income families expressed less willingness. Besides household/family income, household size [38,39,41], children and older people [39], and primary source of income [44] were also analyzed as household characteristics or household factors. For housing characteristics, building area [44], and the age and amount of construction land [45] were proved to have negative effects on farmers’ decision making, but Zhang et al. [25] argued that farmers who owned more residential land area were more willing to accept centralized residence in pure farming areas. For farmland characteristics, since farm work was one of the main sources of income, the more [45] and the larger [44] the cultivated land, the less willing people would be to leave for other places. Evidence from Chen et al. [43] also showed that the distance between farms and new houses was an important factor for farmers’ willingness to withdraw from their rural homesteads. For relative policy environment, scholars focused on trust in institutions [30] or local government [38], understanding of policy [44] and implementation quality of policy [25,45], most of which had positive correlation with farmers’ willingness in and after the process of relocation, resettlement or transformation.

On the whole, the decision-making process of relocation willingness could be affected by many different types of factors, such as age, household income, household size, building area, and farmland area. According to the theory of bounded rationality [51] and evidence from related studies [36,37,45], farmers’ willingness choice and relative decision making were processes of making decisions on the basis of the satisfaction principle rather than the optimization principle. The behavior of decision makers was affected by a specific complex environment, and their cognitive level was determined by a variety of factors. As shown in

Table 1. Definition and descriptive statistical results of independent variables.

Type	Variable	Definition	Mean	S.D.
Individual	Age	under 40 years old = 1; 41–50 years old = 2; 51–60 years old = 3; over 60 years old = 4	2.487	1.100
	Gender	female = 0; male = 1	0.599	0.490
	Education	primary school and below = 1; junior high school = 2; senior high school and above = 3	1.576	0.682
	Migrant worker	No = 0; Yes = 1	0.185	0.388
Household	Household size	1–2 people = 1; 3–6 people = 2; over 6 people = 3	2.049	0.521
	Children under 14 years	No = 0; Yes = 1	0.703	0.457
	Old people over 64 years	No = 0; Yes = 1	0.291	0.454
	Household income	under 10k CNY = 1; 10k–30k CNY = 2; 30–50k CNY = 3; over 50k CNY = 4	2.444	0.959
Housing	Building age *	under 10 years = 1; 10–20 years = 2; over 20 years = 3	1.978	0.837
	Building area *	within 100 sq.m= 1; 100–150 sq.m = 2; over 150 sq.m= 3;	1.740	0.768
Farmland	Original farmland area	under 5 mu = 1; 5–8 mu = 2; 8–11 mu = 3; 11–15 mu =4; over 15 mu = 5	2.855	1.339
	Current farmland area	within 1 mu = 1; 1–5 mu =2; 5–10 mu =3; over 10 mu = 4	2.581	1.058
	Farmland transfer type *	no transfer = 1; part transfer = 2; all transfer = 3	1.492	0.782
	Farmland quality *	within 800 CNY = 1; 800-1, 000 CNY = 2; over 1, 000 CNY = 3	2.163	0.819
	Farmland distance **	within 0.5 km = 1; 0.5–1.0 km = 2; 1.0–1.5 km = 3; over 1.5 km = 4	2.432	1.099
Implementation Environment	Number of relocation sites	4~21	9.809	3.673
	Completion rate (%)	80.25~100.00	96.238	3.240
	Occupancy rate (%)	49.83~83.44	60.602	7.661
	Infrastructure match rate (%)	43.33~91.43	70.869	11.949

Note: Variables with * have no more than 35 missing cases. Variables with ** have 81 missing cases. In this study, 1 CNY = 0.16 USD.

, combined with the actual development situation of the research area, 4 aspects and 15 specific indicators were finally selected and normalized by a unified category code in our study to represent characteristics of farmers themselves. Additionally, considering that relocation willingness was a dynamic process that highly depended on the implementation environment of local policy in a certain period, 4 specific indicators, extracted from 2015 construction reports of 23 town and township governments, were also selected to represent the current policy attraction and governance capacity of local government.

3.3. Samples Characteristics

The data in this study were derived from the previous survey and a subset of the total survey response. After eliminating unavailable cases, 801 cases were finally selected into our study as samples, a small portion of which might have a few missing variables (Table 1.).

As described in the title, the relocation willingness and relocation destination were our research topics and were chosen as the dependent variables in our study. For the former, it means, in the next five years, whether rural farmers want or plan to leave their current houses and move to another site or not, they have two options: No and Yes. For the latter, it means, if rural farmers choose “Yes”, they will face four destination options: NRS, seat of town and township, seat of county and another city. Through data analysis of the samples, there were only 315 participants who opted for “Yes”, among which 133 people preferred NRS as their destination, 73 people preferred the seat of town and township, 97 people seat of county, and 12 people another city. Owing to the aging of the population and rapid urbanization, the age in the samples was mainly over 40 years old, accounting for 75.53% of the total. The proportion of males in the sample was 59.93%, a bit higher than that of females. In terms of education level, 53.43% of the participants only finished primary school and below. About 18.48% were migrant workers who left their hometowns to work in the cities. The household size of 3–6 people was the most common case, accounting for 72.66%. About 70.29% of households had at least one child under 14 years old, and 29.09% of households had at least one older family member over 64 years old. As one of the typical underdeveloped areas of Jiangsu Provence, more than half of the participants lived on a household income of no more than CNY 30,000 (about USD 4800) per year, and only 16.85% earned more than CNY 50,000 (about USD 8000) per year. Due to policy requirements of TVLP and government regulation, 63.85% of the housing buildings reached 10 years of age, the rest 36.15% of which were built no more thanten0 years mainly under the local governments’ planning. Regardless of the area of the living yard, the building area of samples was mainly over 100 square meters, accounting for 54.17%. Since farmland was the primary source of income for farmers, the original farmland area of households mainly ranged from 5 to 15 mu, accounting for 83.77%. With the implementation of the farmland transfer policy, 31.13% of the participants chose to partly or totally transfer their farmland to their local governments or other people. Driven by the new urbanization policy and farmland transferring, the current farmland area of 44.19% of the participants dropped below 5 mu. As to farmland quality, which can be described through crop income per mu unit, 56.96% of the samples reached no more than CNY 1000 (about USD 160). The farmland distance referred to the distance between farmers’ houses and their farmland, and it was found that 78.61% of the samples presented a distance of over 1500 meters.

In addition to the above categorical variables, four factors that reflected the implementation environment of local policy were continuous variables and were counted in Table 1. The number of relocation sites ranged from 4 to 21 and had an average of 9.81 among 23 towns and townships. The completion rate of new relocation sites was more than 80% and had an average of 96.24%. The occupancy rate of new relocation sites that directly reflected their attraction to the local farmers ranged from 49.83% to 83.44% and reached an average of 60.60%. Compared with the completion rate and occupancy rate, the infrastructure match rate had the biggest standard deviation and ranged from 43.33% to 91.43%. The mean value and standard deviation of the infrastructure match rate were 70.87 and 11.95, respectively.

3.4. Model Selection

Logistic regression is the most frequently used regression model for the analysis of discrete outcome variables with two or more possible values [52]. To examine the influence of many independent variables on a nominal dependent variable, logistic regression is usually divided into binary, disordered multinomial, and ordered multinomial regression [25] and has been widely used in farmers’/households’ relocation willingness [25,31,38,39,40] and other relative human behaviors of decision making. Generally speaking, multicollinearity [53] among independent variables and Events Per Variable (EPV) [54] are two key factors in the performance of logistic regression and are easily overlooked in practical application. For the former, there’s no finite solution to maximum likelihood estimate [53], and therefore, it needs to be tested with the corresponding linear regression model [55] and can be overcome through stepwise regression or Categorical Principal Component Analysis (CATPCA) [56]; 10 Variance inflation factor (VIF) is commonly used to test multicollinearity in linear regression model [57,58], with the value 10 as maximum criterion. For the latter, EPV means the minimum number of events in outcome groups or dependent variables per independent variable [54], with the value 10 as a widely advocated minimal criterion for sample size considerations in logistic regression analysis [54,59,60,61].

In our study, there are two dependent variables: relocation willingness and relocation destination. For relocation willingness, as described in the Section 3.3, the minimum number of events in dependent variables, the number of independent variables, and EPV are respectively 315, 19, and 16.58 (higher than 10), meaning that the sample size is sufficient for our study. For relocation destination, the three types of value are respectively 12, 19, and 0.63 (lower than 10), meaning that the sample size is not sufficient and multinomial logistic regression can’t be employed to analyze influence factors of relocation destination directly. In order to meet the requirements of sample size, two approaches are adopted. One is to reduce the events’ number of dependent variables. Considering that NRS plays an important role in rural reconstruction and TVLP and huge differences between urban and rural areas [62], the four choices of relocation destination are reorganized into two groups: NRS and other destinations (Type1), rural areas and urban areas (Type2). Rural areas, including NRS and seats of town and township, are usually defined as the vast areas below the county seat in China, as compared to urban areas, which include the county seats and cities. The other one is to screen out the variables that might affect the dependent variables or have little or no impact. Stepwise selection methods are widely applied to identify a limited number of covariates for inclusion in regression models, which certainly involve the logistic regression model [63]. In this study, two methods, including the forward-LR method and the backward-LR method, are adopted as the stepwise selection methods for that they are based on partial maximum likelihood estimation and are seen as the most reliable methods [55].

In this study, four types of binary logistic regression model are employed to identify key factors and guarantee the stability of the model results. The first model (Model 1) is based on enter method, in which all 19 independent variables would enter into the binary logistic regression model. The second model (Model 2) and the third model (Model 3) are respectively based on forward-LR method with 0.05 entry probability and backward-LR method with 0.10 removal probability to obtain key variables. The final model (Model 4) is also based on enter method, but in which only statistically significant variables in Model 1, Model 2, and Model 3 would enter into the binary logistic regression model. Model 1 is only used for dependent variables with the value of VIF at less than 10 and that of EPV at more than 10. Model 4 is used to integrate all possible key variables that have statistical significance or are considered to be very important from a professional perspective.

It is worth noting that among these 19 variables, 4 variables of implementation environment are continuous variables. The remaining 15 variables are categorical variables with hierarchical characteristics, which can be treated as continuous variables reflecting hierarchical characteristics and used to measure positive and negative effects by many scholars [25,39,41,64,65] or as unordered categorical variables reflecting characteristics of different category groups [38]. In the analysis process of households’ willingness to relocate, all 19 independent variables were used as continuous variables to explore their positive or negative effects on relocation willingness. In the analysis process of households’ relocation destination, except 4 continuous variables of implementation environment, the remaining 15 categorical variables were used as dummy variables to identify possible variable events with significant differences and get the results of Model 2, Model 3, and Model 4 (marked as Model 4A here). After that, treat the valid variables identified in Model 4A as continuous variables and rerun Model 4 to get Model 4B.

4. Results

According to the binary logistic regression models mentioned above, 801 samples and 315 samples with “yes” as the answer to the question of relocation willingness were respectively imported into IBM SPSS Statistics 23 software to run Model 1, Model 2, Model 3, and Model 4. The results of the analysis of factors influencing rural households’ relocation willingness are shown in

Table 2. Logistic regression results of rural households’ willingness on residential relocation.

Options	Model 1		Model 2		Model 3		Model 4
	B	Exp(B)	B	Exp(B)	B	Exp(B)	B	Exp(B)
Constant	−4.942	0.007	−3.920 ***	0.020	−4.476 ***	0.011	−4.482 ***	0.011
Age	−0.371 ***	0.690	−0.325 ***	0.722	−0.352 ***	0.703	−0.326 ***	0.722
Gender	0.444 **	1.559			0.348 *	1.417	0.346 **	1.414
Education	−0.140	0.869
Migrant worker	−0.182	0.833
Household size	0.139	1.149
Children under 14 years	0.174	1.191
Old people over 64 years	−0.050	0.951
Household income	0.122	1.130					0.149 *	1.161
Building age	1.178 ***	3.246	1.090 ***	2.973	1.160 ***	3.191	1.113 ***	3.043
Building area	0.198	1.219			0.212 *	1.236	0.178	1.194
Original farmland area	−0.073	0.930
Current farmland area	0.092	1.097
Farmland transfer type	0.055	1.057
Farmland quality	0.322 ***	1.380	0.343 ***	1.410	0.343 ***	1.409	0.296 ***	1.344
Farmland distance	−0.001	0.999
Number of relocation sites	−0.019	0.981
Completion rate	0.009	1.009
Occupancy rate	−0.010	0.990
Infrastructure match rate	0.015 *	1.015	0.019 **	1.019	0.017 **	1.018	0.015 **	1.015
Omnibus tests	χ2 = 133.197 p = 0.000		χ2 = 119.866 p = 0.000		χ2 = 126.441 p = 0.000		χ2 = 134.339 p = 0.000
−2 Log likelihood	760.161		773.593		767.017		864.110
Percentage correct	67.7%		68.6%		67.7%		67.9%
Effective cases	665		665		665		744

Note: *, **, and *** represent 0.1, 0.05, and 0.01 levels of significance, respectively. Relocation willingness: Yes = 1, No = 0.

. Since household income had been proved by many scholars to be one of the key factors influencing rural households’ decision-making behaviors [30,38,39,44,45] and the study area is an economically underdeveloped region, household income was incorporated into the Model 4, even though it was not recognized in Model 1, Model 2, and Model 3. The results of the analysis of factors influencing rural households’ relocation destinations are shown in

Table 3. Logistic regression results of rural households’ destination for Type 1.

Options	Model 2		Model 3		Model 4A		Model 4B
	B	Exp(B)	B	Exp(B)	B	Exp(B)	B	Exp(B)
Constant	2.830 **	16.949	3.765 ***	43.155	3.495 ***	32.955	3.245 ***	25.666
Household size = 2			−1.204 **	0.300	−1.169 **	0.311
Household size = 3			−0.924	0.397	−0.964 *	0.381
Household size							−0.227	0.797
Household income = 2	−0.707 *	0.493	−0.455	0.634	−0.551	0.576
Household income = 3	−1.432 ***	0.239	−1.195 ***	0.303	−1.082 ***	0.339
Household income = 4	−0.866 **	0.421	−0.597	0.551	−0.516	0.597
Household income							−0.227 *	0.797
Building area = 2	−0.066	0.936	−0.084	0.919	−0.181	0.834
Building area = 3	−1.036 ***	0.355	−1.077 ***	0.341	−0.847 **	0.429
Building area							−0.324 **	0.723
Occupancy rate	−0.036 **	0.965	−0.037 **	0.964	−0.033 *	0.968	−0.033 **	0.968
Omnibus tests	χ2 = 23.506 p = 0.001		χ2 = 28.934 p = 0.000		χ2 = 27.019 p = 0.001		χ2 = 14.126 p = 0.007
−2 Log likelihood	335.765		330.338		385.708		398.601
Percentage correct	65.9%		64.8%		64.0%		61.7%
Effective cases	264		264		303		303

Note: *, **, and *** represent 0.1, 0.05, and 0.01 levels of significance, respectively. Relocation destination: NRS = 1, others = 0.

and

Table 4. Logistic regression results of rural households’ destination for Type 2.

Options	Model 2		Model 3		Model 4A		Model 4B
	B	Exp(B)	B	Exp(B)	B	Exp(B)	B	Exp(B)
Constant	−0.601	0.548	−0.356	0.700	−0.575	0.563	−0.208	0.812
Age = 2			−0.844 **	0.430	−0.806 **	0.446
Age = 3			−0.053	0.949	−0.148	0.862
Age = 4			−0.350	0.704	−0.490	0.613
Age							−0.146	0.865
Household income = 2	0.492	1.636	0.562	1.755	0.691	1.996
Household income = 3	1.634 ***	5.122	1.665 ***	5.287	1.500 ***	4.480
Household income = 4	1.014 **	2.757	1.016 **	2.762	0.794 *	2.213
Household income							0.268 **	1.308
Building age = 2	−0.933 **	0.393	−0.945 **	0.389	−0.624 (p = 0.101)	0.536
Building age = 3	−1.152 ***	0.316	−1.071 **	0.343	−0.755**	0.470
Building age							−0.347 **	0.707
Omnibus tests	χ2 = 27.986 p = 0.000		χ2 = 34.268 p = 0.000		χ2 = 28.648 p = 0.000		χ2 = 13.157 p = 0.004
−2 Log likelihood	314.602		308.320		362.389		377.880
Percentage correct	68.6%		69.3%		68.0%		66.7%
Effective cases	264		264		303		303

Note: *, **, and *** represent 0.1, 0.05, and 0.01 levels of significance, respectively. Relocation destination: Urban area = 1, Rural area = 0.

. The original dependent variables of Type 1 and Type 2 failed to reach the value 10 in terms of EPV; therefore, Model 1 was not executed in this study. As shown in Table 2, Table 3 and Table 4 the rest of the models wouldn’t be affected by the numerical criterion of VIF and EPV.

4.1. Factors Influencing Rural Households’ Choice of Relocation Willingness

As shown in Table 2, the four models are statistically significant. Through the omnibus tests, it can be found that the Chi-square values of Model 1, Model 2, Model 3, and Model 4 are respectively 133.197, 119.866, 126.441, and 134.339, all at the significance level of 0.001. The accuracy rates of Model 1, Model 2, Model 3, and Model 4 are respectively 67.7%, 68.6%, 67.7%, and 67.9%. In the final model, i.e., Model 4, 6 of the total 19 original independent variables are identified as possible factors influencing rural households’ relocation willingness, 4 of which also appear in the other three models and share similar values in the B coefficient and the odds ratio. The outcomes of the model parameter indicate that the results are credible and relatively stable. Specifically, age, gender, household income, building age, farmland quality, and infrastructure match rate are key factors influencing households’ relocation willingness at a significance level not exceeding 0.1.

Age and gender are two individual factors influencing households’ relocation willingness. At the significance of 0.01, every one unit increase in age is linked to a 27.8% decrease in the odds of relocation, which means that the older the age, the less willing people would be to relocate. The elderly people demonstrate less willingness to relocate than young people do. Under the influence of urbanization and industrialization, more and more young people prefer to study, work, and live in cities, leaving large numbers of elderly people, women, and children in their hometown. Having lived in the countryside for a long time makes it difficult for the elderly to change their original lifestyle and adapt themselves to city life. This, coupled with the various costs involved in relocation and urban living, further adds to the unwillingness of the elderly to relocate. For gender, men and women tend to behave differently in the decision-making process due to traditional cultural ideas and different modes of thinking. At the significance of 0.05, the probability that male farmers choose to relocate is 1.414 times higher than that of female farmers. That is to say, compared with females, male farmers show more willingness to the idea of relocation. The main reason may be that men are generally considered the head of household in China, show greater enterprise, and are more receptive to new things, while women tend to pay more attention to details and stability in life.

Household income is the only household factor influencing households’ relocation willingness. At the significance level of 0.10, each additional unit in household income increases the probability by 16.1% that the household chooses to relocate. In other words, the higher the household income, the greater the ability to cope with relocation and the more willingness people would express to do so. In rural areas featuring traditional agriculture, the backward economic development and low income of farmers are two key factors restricting the implementation of relocation. According to the statistics yearbook, the per capita disposable income of rural residents in Sihong County in 2016 amounted to CNY 13,625 (about USD 2180), which was CNY 3981 (about USD 637) lower than that of Jiangsu Province. During the field survey period, we discovered that some households would pay another CNY 100,000 (about USD 16,000) at least on new houses once receiving the relocation compensation, and this expense can cause great pressure for them. Rural households with high family income have more financial ability and show a stronger desire to improve their living conditions.

Building age is the only housing factor influencing households’ relocation willingness. With the B coefficient greater than 1, it can be inferred that the age of the building has a particularly strong positive correlation with households’ relocation willingness. Similar results have also been found in the research of Zhang et al. [25]. According to the evidence from Boggess and MacDonald [66], who are two members of StataCorp LLC, we can conclude that the aforementioned result is effective and believable. The cross-tabulation analysis results of the relation between building age and relocation willingness showed that there were respectively 15.5%, 52.2%, and 53.6% people willing to relocate among three groups of under-10-year building age, 10–20-year building age, and over-20-year building age. The probability to relocate for the group of under-10-year building age was far lower than that of the other two groups. With the building ages as unordered categorical variables and other variables remaining unchanged, the results of binary logistic regression presented that among farmers who preferred relocating, those who had their houses built less than 10 years ago and between 10 and 20 years ago accounted for, respectively, 8.2% and 66.2% of those with their houses built over 20 years ago, with the significance level at 0.000 and 0.052, respectively. The cliff-like decrease in this probability was what caused the value of B coefficient to be greater than 1. As a matter of fact, most of the houses with under 10 years of building age were built by the local government or farmers themselves, depending on the types of construction projects, such as ecological migration project and project of centralized residence. Households with under-10-year building age show less willingness to relocate to other destinations as their current living environment is far better than that of traditional rural houses, most of which have been forbidden to be rebuilt for many years. Accompanied by the outflow of rural population and formation of “Hollow Village”, a large number of traditional rural houses were abandoned, and some NRS were built because older houses could no longer meet farmers’ needs of pursuing a safe and high-quality living environment.

Farmland quality is the only farmland factor influencing households’ relocation willingness. At the significance level of 0.01, every one unit increase in farmland quality means a 34.4% increase in the probability of relocation. Generally speaking, the lower the quality of farmland, the more willingness farmers would show to transfer their farmland to the local government or others and to relocate. Farmland quality has been recognized as an important factor of farmland transferring [67]. This result is contrary to our expectations. One possible reason is that households with higher farmland quality are more financially capable of paying the relocation costs as agricultural income makes up the main source of income for some households. Another possible reason is related to the policy on farmland transfer. According to the feedback of respondents, if they chose to transfer their farmland, households with higher farmland quality would receive CNY 800 (about USD 128) or even higher compensation per mu every year from the local government or related companies, while those with lower farmland quality could get only CNY 500 (about USD 80) at most. Farmers who are willing to transfer their farmlands are most likely to relocate to a new destination.

The infrastructure match rate is the only implementation environment factor influencing households’ relocation willingness. At the significance level of 0.05, every unit increase in infrastructure match rate is followed by a 1.5% increase in the probability of relocating. It means the higher the infrastructure match rate, the more willingness would be shown to relocate. The infrastructure match rate refers to the matching degree of various facilities, including road facilities, green infrastructure, water supply and drainage facilities, and power supply facilities. As revealed in the evidence presented in the research of Chen et al. [43], complete infrastructural facilities and good public services would prompt people to move out of their rural homesteads. As of August 2015, the average infrastructure match rate of Sihong County was just about 70%, which may discourage farmers from relocating to some extent. Therefore, strengthening infrastructure construction still remains the top priority of the local government.

4.2. Factors Influencing Rural Households’ Choice of Relocation Destination

As described in the part of Model Selection, the two kinds of relocation destinations, NRS and urban areas, are also our focus in this study. For Type 1, as shown in Table 3, the NRS and other destinations are respectively coded as 1 and 0. The chi-square values of the omnibus tests of Model 2, Model 3, Model 4A, and Model 4B are 23.506, 28.934, 27.019, and 14.126, respectively, all with a significance level of 0.01. The accuracy rates of these four models are 65.9%, 64.8%, 64.0%, and 61.7%, respectively. For Type 2, as shown in Table 4, the urban area and rural area were respectively coded as 1 and 0. The chi-square values of the omnibus tests of Model 2, Model 3, Model 4A, and Model 4B are respectively 27.986, 34.268, 28.648, and 13.157, all with a significance level of 0.001. The accuracy rates of these four models are 68.6%, 69.3%, 68.0%, and 66.7%. In both Type 1 and Type 2, affected by how the models are dealt with and the existence of missing values, different outcomes of model parameter are presented in Model 3 and Model 4, although they share the same factors. For the same variables that appear in different models, it is found that their B coefficients and odds ratios are of similar values for each same dummy variable. The outcomes of the model parameter indicate that the results of all the models are credible and relatively stable.

Household income is a common factor influencing households’ choice of destinations between NRS and urban areas. According to Model 4B in Table 3 and Table 4, at the significance level of 0.1 and 0.05, every one unit increase in household causes respectively a 20.3% drop in the odds of moving to NRS and a 30.8% increase in that of moving to urban areas. This means the more the household income, the less willing people would be to choose NRS and the more willing to choose urban areas. The most likely reason is that housing prices of urban areas are generally higher than those of NRS; therefore, only those households with higher family income can afford a new house in cities. Besides housing price, the increasing extra cost of urban living is another barrier for households with lower family income as expenses like property management fees, water and electricity fees, and gas fees are far lower in NRS than in urban areas.

The building area and occupancy rate are two important factors influencing households’ choice of NRS as destination. For building area (Table 3), at the significance level of 0.05, every one unit increase in building area relates to a 27.7% drop in the probability of relocating to NRS. This means the larger the building area, the less likely it is to choose NRS as destination. It is perhaps due to the fact that, compared with people who have a smaller house, those who own more building areas can get more compensation or spend less extra money on a new house elsewhere. For occupancy rate, at the significance level of 0.05, every one unit increase in occupancy rate is linked to a 3.2% decrease in the probability of moving to NRS. It is an unexpected result for us. Generally speaking, a higher occupancy rate signifies greater attraction of NRS and a higher level of governance, which will prompt farmers to relocate in pursuit of a higher quality living environment. However, with the gradual implementation of TVLP and the continuous increase in the occupancy rate of NRS since 2005, the land developer of NRS, in better conditions, began to increase the plot ratio of new houses and construct taller buildings, which were very different from the houses with only two or three floors built in the early years. Such behavior of land developers reduced the attractiveness of NRS, causing more households to settle in urban areas instead.

Building age is an important factor influencing households’ choice of urban areas as a destination. As shown in Table 4, at the significance level of 0.05, each additional unit increase in building age signifies a 29.3% drop in the probability of moving to urban areas. This means the older the building age, the less likely it is to choose cities as a destination. Specifically, among people who preferred moving to cities, those with a house built over 20 years ago and between 10 and 20 years ago accounted for respectively 47.0% and 53.6% of those with a house built less than 10 years ago, with the significance level at 0.05 and 0.101. People with a house of less than 10 years mostly enjoy a more modern living environment, in which some public facilities may not work well and meet their living needs. For these people, their cultural concepts and consumption concepts had changed after relocating to their current houses. People with a house of over 20 years and between 10 and 20 years still prefer to stay in rural areas and to maintain the original production and lifestyle as much as possible.

5. Discussion

On the whole, among the factors influencing the households’ decision-making behaviors of relocation willingness, household income, building age, farmland quality, and infrastructure match rate had a positive effect. The analysis results of household income and building age were largely consistent with those from previous relative research [38,39,44,45]. Zhang and Han [38] presented that higher-income families showed less willingness to remediate their residential land, which also means such families expressed less willingness to relocate, for relocating to NRS or other destinations is the current main method to implement the policy of community remediation in China. One possible reason why the conclusion of Zhang and Han [38] differed from that of this study was that samples used by Zhang and Han [38] came from three typical regions with different economic development levels, while samples of this study came from some underdeveloped areas in the east of China. It was the difference of economic development levels, not the difference among individuals or groups with different household income that played the leading role in the research of Zhang and Han [38]. As demonstrated in the results of this study, household income was also a key factor influencing households’ choice of relocation destination. The higher the household income of farmers, the more willingness they would show to relocate to urban areas instead of rural areas. Therefore, improving the level of regional economic development and household income is an important way to realize rural revitalization and solve the problem of population exodus.

Farmers’ age was inclined to affect households’ decision-making behaviors in a negative way. Male farmers expressed more willingness to relocate than female farmers did. The analysis result of farmers’ age was basically consistent with that from previous relative research [38,44], but evidence from Xu et al. [39] and Zhang et al. [25] showed that the age and gender of participants had no correlation with their relocation willingness. In China, population aging and migration between rural and urban areas have become two major challenges affecting social and economic development [8], leading to huge differences and uncertainties in the rural population structure. Therefore, the regional differences may be the main reason for the different results.

The building area and occupancy rate had no significant correction with relocation willingness but were proved to be related to the choice of relocation destination. In this study, at the significance level of 0.119, every one unit increase in building area means a 19.4% rise in the odds of relocating. Though the building area had no significant effect on relocating willingness, it shared similar positive characteristics as the results of Zhang et al. [25]. In this study, the term occupancy rate was beyond its literal meaning, implying farmers’ pursuit of rural residential styles and living conditions, both of which were different from those in urban areas.

Factors including education level, the proportion of migrant workers, household size, number of children under 14 years and old people over 64 years, original farmland area, current farmland area, farmland transfer type, farmland distance, number of relocation sites, and completion rate all had significant effects on households’ decision-making behaviors of relocation willingness and destination. The analysis result of education, household size was basically consistent with those from previous relative research [25,39,41]. According to the research of Tao et al. [68], migrant workers were a special group with a particular preference of housing choice. However, in our study, only 18.48% of the samples were from migrant workers as most of them rarely stayed at home. The research of Zhang et al. [25] presented that arable operation means would also affect households’ attitude towards the centralized residence, i.e., those who chose to rent out farmland showed more willingness to accept centralized residence policy. Such conclusion differed from what was reached in this study that households were more likely to transfer their farmlands instead of renting them out.

6. Conclusions

Centralized residence has played an important role in dealing with “Hollow Village”, alleviating rural poverty, optimizing land-use pattern, and realizing rural revitalization. However, all of the traditional centralized residence patterns based on “building a new socialist countryside” and “maintaining a balance between the increase and the decrease” are essentially top-down and require farmers to respond and readjust to all possible policies and changes. In this study, a household survey was conducted in 23 towns or townships in Sihong County, Jiangsu Province, China. According to the existing criteria of EPV and VIF used by most scholars, four different two groups of binary logistic regression were used to analyze factors that influenced households’ decision-making behavior. The results indicated that there might be heterogeneous groups among these participants in Sihong County. Household income might be a fundamental factor that influenced farmers’ relocation willingness and destination, which explained why some participants with financial difficulty refused to answer our questions. The factors of implementation environment, especially the occupancy rate and infrastructure match rate, can basically reflect the differences in the governance capabilities of local governments.

Based on the results of the above analysis and actual development of the study area, this paper tries to provide some policy suggestions for TVLP in the study area from the following five aspects: Firstly, create more jobs. Since household income is positively correlated to households’ relocation willingness to choose urban areas as destination, the local government should launch more supporting industrial projects around NRS to ensure stable family income and attract more households to relocate here. Secondly, implement the strategies of zoning and classification. For example, as building age and infrastructure match rate are both positively correlated to household relocation willingness, it’s necessary to strengthen infrastructure construction for houses built within 10 years and NRS and implement dilapidated housing renovation projects for houses of over 20 years to guarantee their basic living conditions. Thirdly, improve the quality of land transfer services, which includes improving the service level of rural land transfer agencies and establishing a land transfer security deposit system to ensure that farmers can receive timely compensation for land transfer, especially for those with high-quality farmland. Finally, reconstruct the rural landscapes. Rural areas have different landscape features from cities, such as low density, low floors, and idyllic scenery. The construction of NRS must conform to the characteristics of the rural landscape.

On the whole, this paper concentrates on the case study of factors influencing rural households’ decision-making behaviors and compares it with other similar studies in China. The main conclusions not only help to improve China’s centralized residence policy but also offer useful reference and theoretical support to other countries confronting similar rural problems. On the one hand, this study places the policy of centralized residence in the international context of rural restructuring. It is very helpful to build a bridge between China’s centralized residence and relative research topics, such as the behavioral preference of housing choice, the heterogeneity of different socioeconomic groups, and top-down or bottom-up framework index planning. On the other hand, economic conditions and the governance capacity of local governments may also be two common factors influencing the decision-making behaviors of the global population. Economic status has been a key indicator that affects the Human Development Index (HDI) in different degrees since the 20th century [69]. It could be one of the key factors influencing the decision-making behaviors of residents, especially for those living in developing countries or poor areas. For the governance capacity of local governments, people-centered practices embedded in policy, planning, and implementation play a greater role than resettlement itself [47]. Finally, all the above discussions and in-depth studies will help to clarify debated issues about centralized residence in rural areas from a global perspective. As a policy expectation, the centralized residence has greatly improved the living conditions of Chinese farmers, but more evidence is still needed to illustrate its effects on rural revitalization.

There are inevitably some limitations in this study. As the research of centralized residence is still in its early stages [26], and scholars have not yet reached a consensus on relevant policy perceptions, this study is just preliminary research to explore the influence of different socioeconomic characteristics and governance capabilities of local government on relocation willingness and destination. The factors of various demolition compensation policies should be introduced in future research. Additionally, due to the limitation of data sources, the results of this paper only reflect the preference of the current survey area. Future studies should select more samples from different areas for comparison.