Does China’s Anti-Poverty Relocation and Settlement Program Benefit Ecosystem Services: Evidence from a Household Perspective

Abstract

To assess whether and to what extent the anti-poverty relocation and settlement program (APRSP) in China will be able to resolve the development dilemma of ecosystem conservation and human wellbeing, it is important to study the effects of policy on rural households in terms of the income generation from ecosystem services (ES). We constructed an index of dependence on ecosystem services (IDES) to evaluate the dependence of households’ net income generation on ecosystem services. Using data collected from South Shaanxi Province, we examined the effects of the relocation program on rural households’ IDES. We find that this relocation may benefit the ecosystem by significantly decreasing participants’ IDES. Relocation households have higher net incomes than non-relocation households from total ecosystem services, provisioning services, regulating services, and cultural services as well as socio-economic activities. There are significant differences in IDES between groups with different relocation and resettlement characteristics. The anti-poverty relocation program optimized the rural households’ income structure by increasing the proportion of income from socio-economic activities while reducing the proportion of income from ecosystem services. This study provides new evidence for evaluating eco-conservation and development policies by linking ecosystem services and human well-being at a micro scale. We also address the policy implications of our analysis for anti-poverty relocation programs.

1. Introduction

China plans to eradicate extreme poverty by 2020, in part by relocating 9.81 million people during the “13th Five-Year Plan” from barren areas that have fragile ecosystems or are prone to natural disasters. This ambitious program aims to solve the development dilemma that “the environment of a place cannot afford the inhabitants of that place” by achieving ecological protection and poverty alleviation in harsh environments. Considering the conflict between the public nature of ecological resources and the economic interests of individual rural households, on the one hand, the government hopes to improve rural households’ capability for self-development using the process of urbanization in order to increase incomes and alleviate poverty. On the other hand, the relocation and development of non-agricultural labor could reduce dependence on the ecosystem and enhance local ecological restoration and protection. During the policy implementation, rural households are both the main participants and important stakeholder groups, but they are also the basic unit of social production and consumption decision-making. The behavior of the relocated rural households reflects the processes of microeconomic behavior that they adopt to use natural resources and ecosystem services (ES), while determining the sustainability of the relationship between ecological protection and economic development [1]. Therefore, the success of the program depends on the extent to which the relocation and resettlement program can solve the development dilemma of rural households’ livelihood and ecological conservation. This paper aims to evaluate the policy’s dual goals, by analyzing the livelihoods of rural households from a micro perspective and addressing the impact of policies on ecologically relevant rural households’ income behavior.

Measurement of the ecological benefits of the policy or program depends on an effective assessment of the value of ES as well as the values of the changes in human well-being due to the restoration of livelihoods. Since the 1990s, the supply, consumption, function and value of ES have become important research foci. Since the development of ‘natural capital’ in both theory and practice [2], increasing attention has been paid to the interaction between ES and human activities. After the idea of ‘natural capital’ was recognized in China, the products provided by ecosystems and their functions in regulating life support, as well as their social and cultural functions, have been generally recognized in the academic community [3,4]. Scholars have explored and evaluated ES across different regions, time and scales. Their conclusions have promoted people’s awareness of ES and ecological protection [5]. With the launch of the “Millennium Ecosystem Assessment” project, researchers began to recognize the close relationship between ES and human well-being. Some scholars consider ES as a benefit to human beings [6], while there is a growing focus on how human well-being changes with the intensity of human consumption of natural resources [7]. The ‘Millennium Ecosystem Assessment’ project also addressed the close relationship between ES and human well-being [6], as ES have a direct impact on human well-being. In addition, people’s well-being also affects their consumption of natural resources and the function and supply of ES [7]. Therefore, one of the current directions for research is how to balance ecosystems and human well-being. This entails that the measurement and assessment of human income from ecosystems should be key to scientific and policy decisions. Using existing assessment models and tools, many scholars have assessed the benefits of ES for the entire human and regional population [8,9,10] at multiple scales. However, more measurement of effects at the micro-scale is also required; for instance, decision-making on use and dependence on ES at the household level. The functionality of ES cannot be separated from the impact of micro-level decisions and behaviors. Therefore, it is important to examine the dependency of rural households on ES.

This paper analyzes the impact of the anti-poverty relocation and settlement program (APRSP) on ES from the perspective of households. First, from previous studies [11], we build an Index of Dependence on Ecosystem Services (IDES) by integrating and quantifying the benefits to rural households from ES. Second, we calculate this dependence, and explore the effects of relocation on rural households’ dependence on ES.

2. Theoretical and Empirical Background

A comprehensive understanding of the relationship between ES and human well-being is the basis for optimizing relations between humans and the land and solving the problem of ‘public ponds’ and negative externalities. However, since the United Nations Millennium Ecosystem Assessment, most research on this relationship is still based on a theoretical framework and large-scale description. In order to clarify policy needs for how resources and development will affect the poor and vulnerable groups, we need to examine more accurately the relationship between ES and human well-being. Otherwise, we risk making the plight of the poor population worse in the process of pursuing economic growth and human development [12]. However, relevant research in this area still lacks a systematic approach.

First, the quantification of ES has been concerned with the overall contribution of natural capital to human well-being, and there is a lack of research methods and tools to quantify the value of ES under dynamic and uncertain conditions. Recently, there have been qualitative improvements in the understanding and assessment of ecological services. Ecologists have attempted to mainstream the concept of ecosystem services (ES), while relevant economic policies have promoted the transformation of theory into action [13]. It has become accepted that scientific decision-making depends on effective assessment of the value of ES. Many assessments of the value of ES have tended to focus on the value of nature itself [14], rather than the distribution and consequence of natural services. With the concept of ecosystem services proposed by Daily (1997), people are becoming aware that full understanding of ES across human groups, time, and spatial scales is the real core of the pricing of nature [15]. However, practical integration of ES as ‘natural capital’ into decision-making about resource utilization is still in the early stages of development. While existing assessments based on the type and unit price of ES increase awareness of their importance, these methods have limitations that lead to underestimation or avoidance of pricing of ES [16,17,18].

Second, in-depth analysis of rural households, which are the key link in connecting ES with the well-being of residents, can provide new perspectives on the relationships between ES, human well-being and regional sustainability. Yet, characterization of the dependence of human well-being on ES from a micro perspective remains to be accomplished.

Third, domestic and foreign scholars have studied the relationship between quantitative ES and human well-being, and proposed many indicators that provide useful quantitative analysis of the relationship between the two. Malte Busch (2011) associated the tangible and intangible ecosystem benefits with a range of material and nonphysical factors that constitute human well-being by studying the linkages between offshore and coastal human well-being [19]. Using economic analysis and questionnaires, he found that offshore wind power production affected ES and changed human well-being, providing a new perspective on the relationship between ES and HWB. Research methods for a quantitative assessment of ES and the study of the relationship between ES and HWB have been developed from the perspectives of cost-benefit analysis tools [20], land-use change [21,22], web-based applications [23], and the Millennium Ecosystem Assessment framework [24]. Yang et al. proposed a method to quantify the index of dependence on ecosystem services (IDES), which quantifies the relationship between ES and HWB, and which they used to study the dependence of local farmers in the Wolong Nature Reserve in China on ES [11].

To implement the relocation and resettlement program effectively, integration of ecology and the effects of poverty reduction from a micro-perspective is necessary; here we study this integration at the level of households. On the one hand, domestic evaluation of ecological benefits from the relocation and resettlement program is mostly in terms of ecological migration, but such analyses of ecological effects have concentrated on macro ecological effects, through a macro ecological index system, regional soil and water conservation, ecological restoration and so on. On the other hand, with research on rural households’ livelihood attracting attention, many scholars have shifted research on farmer’s livelihood from economics to ecology. The relationship between human livelihood and the ecological environment has become a hot topic in research on humans and ecological systems science [25,26]. Land use and land cover change, as well as energy use by relocated and resettled households, are considered important indicators of how rural households’ livelihood strategies affect the environment. Rural households often spend a long time adapting to significant changes in their geographical location and livelihood capitals in the process of migration. In order to achieve the goal of livelihood restoration, strategies adopted by rural households have a significant impact on the ecological environment by affecting both production and consumption [27]. A powerful engine for the transformation of rural households’ livelihoods in ecologically fragile regions in the context of urbanization is their gradual shift from traditional agriculture to non-agriculture and diversification. Correspondingly, their consumption behavior, dependence on natural resources, and utilization of natural resources, has changed their dependence on the ecological environment [28].

Existing research provides a theoretical basis for the analysis in the present study, but also leaves room for further research. There have been few studies of the relocation and resettlement program, from the perspective of the value of ES. There is no evaluation system or analysis that links ES with the well-being of micro-farmers through the mechanisms that farmers use in the process of obtaining benefits. This paper analyzes the impact of relocation and resettlement program on the dependence of rural households on ES by in terms of an Index of Dependence on Ecosystem Services (IDES), which we use to evaluate the effect of policies at the household level.

3. Data and Methods

3.1. Data Sources

The data in this paper are from an investigation of rural households’ livelihoods in AnKang Municipality in southern Shaanxi Province, China. AnKang is located in an extremely poor area in the QinBa concentrated serial poverty mountain area. There are three million permanent residents, of whom about 1/3 are poverty-stricken and about 52% are in poor villages. The relocation and resettlement program in southern Shaanxi involved 226,000 rural households, or about 880,000 rural people from Ankang, who are all living in remote mountainous areas that are prone to natural disasters. The five counties selected for the survey in Shaanxi province are Pingli County, which is the key county for poverty alleviation and development at the Shaanxi Province level, Hanbin District, Ningshan County, Ziyang County and Shiquan County all of which are the key counties for poverty alleviation and development at the national level.

During the survey, we first selected three townships in each target county (district), which met the following conditions: They implemented the relocation and resettlement program; they are in or near nature reserves, and have implemented ecological conservation, such as sloping land conservancy program (SLCP). Next, the research team selected 25 villages in each township, including ten communities targeted for resettlement, three villages for implementing the SLCP and 12 were randomly selected. Then, two village groups were randomly selected from each administrative village (community) according to the Hukou registration list. Finally, we chose all permanent residents of each sampled villager group. The survey respondents were heads of households or their spouses aged 18–65 and the survey included basic information about their families, various livelihood capital conditions, livelihoods activities, relocation and settlement status. During the field investigation, we used a series of quality control measures in order to ensure the quality of the data. We first trained the investigators and ask them to pay attention to how they administered the questionnaires. Then, we assigned supervisors for each survey group to monitor and adjust their operation at any time during survey. Third, the supervisors selected part of the completed questionnaires and rechecked them with these respondents each day. These results were sent back to the investigator on the same day. After the survey, the entire questionnaire was turned into numeric code and entered into the dataset. Finally, we conducted logical and numerical data cleaning procedures to further ensure the quality of the data entered. The survey obtained 1306 valid questionnaires, of which 29.1% were relocated households and 70.9% were non-relocated households.

3.2. Indicator Construction

The Construction of an Index of Dependence on Ecosystem Services (IDES)

We refer to Yang et al. (2013) [11] for the construction of IDES. The index system includes an IDES total index and three sub-indices. IDES is the ratio of the net income earned by a family from the ecosystem to the absolute value of the total net income of that family derived from the ecosystem and other economic activities (including remittances from out migration work in urban areas and small businesses that are not related to ES, see

Table 1. Detailed classification of household net income by type of related ecosystem services.

Category	Subclasses	Item	Type of Related Ecosystem Services
Operating Income	Crop income	Cabbage	P0
		Carrot	P0
		Potato	P0
		Corn	P0
		Other crops	P0
	Husbandry income	Bacon	P1
		Pig	P0
		Ox	P0
		Yak	P0
		Horse	P0
		Poultry and eggs	P0
		Bee	P0
		Other farming	P0
	NTFPs income	Non-timber forest product	P0
	Other agricultural income	Other agricultural income	P0
	Non-agricultural income	Hotel and restaurant	C1 or NA†
		Eco-tourism	C1 or NA†
		Transportation	C1 or NA†
		Contract work	NA
		Other small business	C1 or NA†
Wages Income		Wages and bonuses	NA
		Local labor income	NA
		Immigrant labor income	NA
Property Income	Lease of land and houses	Lease of land and houses	C1 or NA†
	Other property income	interest income	NA
		Land compensation	NA
		Other rental	NA
Transfer Income	Revenue received from relatives and friends	Gift income	NA
	Income pay for ES	Subsidies for returningfarmland to forest	R0
		Subsidies for ecologicalpublic welfare forest	R0
		Agricultural subsidies	R0
	Social security benefits	The lowest income subsidy	NA
		Pensions	NA
		Other subsidies	NA
Other Income		Other social and economic income	NA
Avoided Costs		Firewood collection	P0

Note: † means that only when this income relates to ecotourism, could it be allocated to the revenue related to cultural services; Otherwise, it is unrelated to ES.

) The three sub-indices include a provisioning services index, a regulating services index, and a cultural services index, which are calculated as above [29]

We use cost-benefit analyses (CBA) to assess net benefits of a unit of analysis at the household level. The data from a variety of sources are aggregated into an integrated assessment. Where market price of the surveyed items is available, for example the crop item and the husbandry item, it is quite simple to assess the gross benefits by multiplying the quantity that is obtained from the questionnaire survey. However, when market data are not available, non-market valuation methods such as the contingent valuation method, the travel cost method, and the stated preference method were used. We combined them together to calculate relevant costs of crop and husbandry. For example, we evaluated the total amount of fuel wood used in one year by first estimating a family’s one day amount by using the unit-day value method. Then we used the benefit transfer method to estimate its economic value as the household’s avoided cost.

From the definition of IDES, the higher its value and that of each sub-index, the more humans are dependent on the related ES. Lower values indicate that human beings are less dependent on ES. The calculation of IDES and the three sub-indices are as follows:

$$IDE{S}_i=\frac{EN{B}_i}{\left|{{\displaystyle \sum }}_{i=1}^{3}EN{B}_i+SNB\right|},$$

(1)

$$IDES={\displaystyle \sum }_{i=1}^{3}IDE{S}_i,$$

(2)

where i is ecosystem service types. IDES is a general measure of human dependence on ecosystem services. $IDE{S}_i$ indicates a sub-index of human dependence on i-type ecosystem services. $EN{B}_i$ represent the total net revenue humanity receives from i-type ecosystem services, and $SNB$ represents the total net benefit that human beings derive from socio-economic activities (e.g., migrant work, and small business unrelated to ecosystem services).

According to the types of household net income and avoided cost, this paper classifies the items from ES in the construction of the index system as follows in Table 1. The letters P, R, C, and NA represent the benefits of provisioning services, regulating services, cultural services and ecosystem-unrelated services, respectively. For a rural household, there are normally five kinds of income category, which are operating income, wages income, property income, transfer income, other income and avoided cost. We labeled each income item of the sub-category with the relevant type of related ecosystem service. Crop income, animal husbandry income, non-timber forest product income, and other agricultural operating income were aggregated as the provisioning service. In addition, households’ other socioeconomic activities were identified as the income unrelated to ES. However, we classified the item as cultural services only when the benefit is related to ecotourism; otherwise it is regarded as a benefit unrelated to ES. In our survey site, Payment of Ecosystem Service (PES) programs were designed mainly for regulating services (e.g., water conservation, soil erosion control, carbon sequestration, and air purification). Thus, we classified PES as benefits related to regulating services. For example, subsidies for returning farmland to forests, ecological public welfare forest subsidies and agricultural subsidies are classified as regulatory services. In this paper, we only evaluate the direct and first-order indirect use values from ES. Therefore, we identify these by the numbers ‘0’ and ‘1’ after ‘P, R, or C’, representing direct and indirect benefits from ES, respectively. Since the gross benefit obtained from ES might be lower than the sum of costs for generating the corresponding ES and costs of ecosystem dis-services, the total net benefits value might be negative.3.2.2 Factors Influencing IDES

According to the literature and the situation at the survey site, the factors affecting IDES are classified into four categories: relocation factors, livelihood assets, family population characteristics and geographical features.

Table 2. The setting and value of the independent variable.

Variables	Variables Setting	Value
		Mean	Standard Deviation
Relocation Factors
Relocated-family	Dummy variable. Relocated-family takes 1, otherwise takes 0	0.28	0.45
Relocation Feature
Relocation type	Dummy variable. Involuntary relocation takes 1, voluntary relocation takes 0	0.73	0.44
Settlement mode	Dummy variable. Centralized settle takes 1, scattered settle takes 0	0.62	0.49
Relocation time	Dummy variable. New stage (2011 and after) takes 1, early stage takes 0	0.35	0.48
Livelihood Assets
Land per capita	Continuous variables. Ratio of total land area to total population (unit: mu/person)	1.22	1.58
Forest per capita	Continuous variables. Ratio of total forest area to total population (unit: mu/person)	10.34	18.17
Own assets	Continuous variables. The total number of household-owned production tools, vehicles, and durable goods	2.82	1.72
Housing valuation	Continuous variables. Estimated value of family housing (unit: ten thousand yuan)	9.97	7.01
Social support net	Continuous variables. The number of households that can provide help urgently when the family needs a large sum of money	4.36	4.03
Specialized cooperative	Dummy variable. Join in specialized cooperative takes 1, otherwise takes 0	0.05	0.22
Cadre relative	Continuous variables. The number of household’s relative who is cadres at village or township level	0.54	1.3the
Head of Household
Gender (nominal)	Dummy variable. Female takes 1, otherwise takes 0	0.11	0.31
Gender (actual)	Dummy variable. 1 for the household that the male head of household went out for more than six months, the actual family decision-making power is transferred to the female, otherwise takes 0	0.13	0.34
Age	Continuous variables. Age of head of household	50.62	12.64
Average education years	Continuous variables. The ratio of the total number of education years to the family size	6.21	2.75
Total labor force	Continuous variables. Number of members over the age of 16 and under 65	2.74	1.39
Family Structure
Elderly + adult	Dummy variable. Elderly and adult workers	0.17	0.38
Adults	Dummy variable. Adult workers take 1, otherwise takes 0	0.36	0.48
Adults + children	Dummy variable. Adult workers and children take 1, otherwise takes 0	0.28	0.45
Elderly + adult + child	Dummy variable. Adult workers, children and elderly takes 1, otherwise takes 0	0.13	0.34
Geographical Features
Distance to the town	Continuous variables. The distance from the village where the farmer lives to the township (unit: ten kilometers)	10.34	7.96
Adjacent reserve	Dummy variable. Close to or in nature reserve takes 1, otherwise takes 0	0.35	0.48

shows the specific settings and values of variables.

The relocation factors, which represent the different scenarios of relocation, are the main independent variables that we focus on. These factors include whether the household has been involved in the relocation and resettlement program, the type of relocation, the mode of resettlement and the time of immigration. In the empirical analysis, we first explore the impact of participation in the relocation and resettlement program on the household’s IDES by including the variable “relocated-family”, as well as other control variables in the whole sample regression model. For the second stage, in order to further assess the effects of different types of relocation features on the household’s IDES, we use the relocated households sample to carry out the estimation. At this stage, three relocation features, namely, “relocation type”, “settlement mode”, as well as the “relocation time”, are included in the regression model one by one, together with the control variables. The types of relocation mainly include voluntary relocation and involuntary relocation (due to construction of engineering projects); among the 355 relocated households, 261, or 73.52%, of households voluntarily relocated. According to the resettlement mode, the relocated families can be divided into two types: centralized resettlement and scattered resettlement. Two-hundred twenty-four (224) relocated households are included in centralized resettlement, accounting for 61.71%. The relocation program in south Shaanxi was initiated in 2011, before which many resettlement relocation activities were carried out sporadically, but the scale, subsidies and support were far less than the later ones. Therefore, we term New Stage Relocaters those who participated in relocation after the implementation of the relocation policy of Southern Shaanxi in 2011, and Early Relocaters those who relocated before the new policy. We compare the impact of different policies and efforts, and at the same time assess rural householders’ ecological behavior changes.

In the regression model, the control variables, namely, livelihood assets, family demographic characteristics, and geographical features, were included. The livelihood assets of farmers include three items: natural capital, material capital, and social capital. Among these, per capita arable land and forest area represent natural capital; the two indexes of farmer’s own assets and house valuation represent the material capital of farmers; the scale of the social support network, participation of specialized cooperative organization, as well as cadre relative of rural households are chosen to represent the family’s social capital. In general, the more natural capital households own, such as land and forest, the more services they will obtain from the ecosystem. Social capital is measured in terms of the size of the social support network, membership of a specialized cooperative and number of cadre relatives. The size of the social support network is represented by the number of households that a farmer can call on for help in emergencies. We use the number of cadre relatives to measure the quality of a household’s social network. In rural China, households’ social networks can play a buffering role when farmers encounter risks and shocks, thus avoiding the predatory use of ecology by farmers. We also added another indicator, “membership in a specialized cooperative”, which consists of “whether the household belongs to a specialized cooperative organization.”

Family demographic characteristics include those of the head of household, average years of schooling of the family, number of workforce members in the family and family demographics. The head of household is usually the highest decision-maker of a family, whose age and gender play a decisive role in the family’s income and needs. In general, the older the head of a household, the higher the reliance on ES [30], and households with female heads tend to be more sensitive to ecological change and rely more on ES [31]. This paper examines the gender of the head of a household from in both nominal and actual terms. The nominal head of household is registered in the household register, and 89% of the sample households have a male as the nominal head of household. Considering rural-urban mobility of labor, this paper also considers the actual head of household; that is, the transfer of actual decision-making power to other members of the household after the nominal head of household migrates out. The family composition of rural households is also an important factor for households’ decision-making. Based on previous studies [32], we divide the family structure of peasant households into four types: the elderly and adult working-force in the family, the adult working-force in the family, the adult working-force and children in the family, the elderly and children and the adult working-force in the family.

Considering the geographical conditions and the actual situation of rural households in south Shaanxi, we judged whether there were adjacent protected areas and we used the distance to the town to measure the ’householder’s geographical location’. The distance to the town not only indicates whether the traffic conditions in the communities where the farmers are located are convenient, but also measures the degree of households’ access to the market, which is an important factor affecting the expansion of non-resource income channels for farmers. Production behavior of rural households that are in or adjacent to the nature reserves, tends to be somewhat limited, and whether there is an adjacent reserve also affects rural households’ access to natural resources to some extent [33].

3.3. Analysis Method

Using the Index of Dependence on Ecosystem Services (IDES), this paper explores the impact of the relocation and resettlement program on the dependence of rural households on ecosystem services. The dependent variable, the overall IDES, is right-censored. It fits the requirement of a Tobit model proposed by Tobin (1958) [34]. The model is expressed by the following equation:

$$IDES=\{\begin{array}{l}1 if y_i^{*}>1,\\ y_i^{*} if y_i^{*}\le 1,\end{array}$$

(3)

$$y_i^{*}=\beta x_i+{\mu }_i, {\mu }_i~\mathrm{N}\left(0, {\sigma }^2\right),$$

(4)

where, $y_i^{*}$ is a latent variable, $x_i$ is the vector of explanatory variables, namely relocation factors, family livelihood assets, family demographics and geographical features. $\beta$ is the vector of estimated parameters, and ${\mu }_i$ is the disturbance term.

The regression method is as follows: First, we analyze the factors influencing IDES in the overall sample, and verify the contribution of relocation to IDES. Then, taking relocated households as the sample, we analyze the impact of different relocation characteristics (including the type of relocation, resettlement mode and immigration time) on rural households’ IDES.

4. Results and Discussion

4.1. Comparison of Rural Households’ Benefits from Ecosystem Services

Table 3. Net income from related ecosystem services.

Net income (Unit: Yuan)	Relocated Households		Non-Relocated Households		The Total Sample
	Mean (Standard Deviation)	Minimum: Maximum	Mean (Standard Deviation)	Minimum: Maximum	Mean (Standard Deviation)	Minimum: Maximum
Total Net Ecosystem Income	15,145 (26,330)	−50,000:193,281	13,646 (23,709)	0:252,725	14,084 (24,502)	−50,000:252,725
Social Economic Activities	12,244(14,335)	0:94,000	5489 (8342)	0:70,000	7475 (10,904)	0:94,000
Provisioning Services	10,463 (19,716)	−1027:192,501	9747 (17,195)	−362:201,805	9948 (17,934)	−1027:201,805
Regulating Services	980 (977)	0:7625	819 (892)	0:7080	864 (919)	0:7625
Cultural Services	3642 (16,159)	−50,000:150,000	2875 (16,653)	0:250,300	3099 (16,507)	−50,000:250,300

Note: A negative value means that the total benefit derived from ecosystem services is lower than the sum of costs for generating the corresponding ES and costs of ecosystem dis-services; All unit of the number is Yuan.

reports the net gains from different sources for both relocated and non-relocated households. The table shows that relocated households received more total net ecosystem income than non-relocated households; the average total net benefit for the relocated households from ES was 1,498.67 yuan higher than that of non-relocated households. We see from the sub-indicators that the net incomes of relocated households from provisioning services, regulating services and cultural services are also higher than those of non-relocated households. In addition, the net income of relocated households from social and economic activities is about 6754.19 yuan higher than the average for non-relocated households.

4.2. Comparison of Rural Households’ Dependence on Ecosystem Services

Table 4. Comparison of the overall IDES and sub-indicators.

Indices	Whether Relocated		t-test	Relocation Type		t-test	Settlement Mode		t-test	Relocation Time		t-test
	Yes	No		Voluntary	Involuntary		Centralized	Scattered		New Stage	Early Stage
Total IDES	0.51	0.67	7.31 ***	0.51	0.92	4.36 ***	0.45	0.60	3.15 ***	0.42	0.56	2.95 ***
Provisioning services	0.37	0.52	7.44 ***	0.37	0.35	−0.54	0.34	0.42	2.45 **	0.33	0.40	1.92
Regulating services	0.08	0.09	1.18	0.06	0.14	6.22 ***	0.06	0.11	4.34 ***	0.04	0.10	4.37 ***
Cultural services	0.06	0.06	−0.5	0.07	0.03	−1.14	0.06	0.07	0.32	0.05	0.07	0.57

Note: * p < 0.1; ** p < 0.05; *** p < 0.01.

provides a comparison of IDES and various sub-indicators by different types of relocation factors. According to the IDES index, there are significant differences in the dependence of farmers on ES under different relocation conditions. First, the IDES of relocated households is significantly lower than that of non-relocated households. The reasons for this difference are mainly concentrated in the sub-indices of provisioning services of the two groups. The index of provisioning services for relocated households is significantly lower than that of non-relocated households, while the two samples show no significant differences in the indices of regulating services and cultural services. Second, the IDES of voluntarily relocated households was significantly lower than that of involuntary relocated households. This difference is due to the regulating service indices of both groups; the index of regulating service for voluntary movers is significantly lower than that of involuntary movers. Although the indices of provisioning services and cultural services for voluntary movers are higher than those of involuntary households, the differences between the two are not significant. In addition, the IDES for centrally located households is significantly lower than that for scattered settled households. From the sub-indices, the provisioning services and regulating services indices of centrally located households are significantly lower than those of scattered settled households. However, the two show no significant difference in cultural services. Finally, the IDES of relocated households at the new stage is significantly lower than that of early relocated households, and the index of regulating services of the former is also significantly lower than that of the latter. However, there is no significant difference between the two in the indices of provisioning and cultural services.

4.3. Determinants of IDES

The overall IDES is used as the dependent variable in estimating factors that affect IDES. The results are shown in

Table 5. The impact of relocation factors on rural households’ IDES.

Variables	(1)	(2)	(3)	(4)
Relocation Factors
Relocated-family	−0.096 ***
	(0.025)
Relocation Feature
Relocation type	\	0.195 ***
		(0.064)
Settlement mode	\	\	™0.188 ***	\
			(0.05)
Relocation time	\	\	\	−0. 198 ***
				(0.046)
Livelihood Assets
Land per capita	0.036 ***	0.085 ***	0.075 ***	0.057 **
	(0.007)	(0.027)	(0.025)	(0.024)
Forest per capita	−0.001	−0.001	−0.001	−0.001
	(0.001)	(0.002)	(0.002)	(0.002)
Own assets	0.028 ***	0.061 ***	0.062 ***	0.058 ***
	(0.007)	(0.013)	(0.013)	(0.013)
Housing valuation	−0.006 ***	−0.009 ***	−0.009 ***	−0.009 ***
	(0.001)	(0.003)	(0.003)	(0.003)
Social support net	0.001	0.001	0.002	0.004
	(0.002)	(0.005)	(0.005)	(0.005)
Specialized cooperative	0.011	−0.058	−0.079	−0.013
	(0.047)	(0.079)	(0.079)	(0.076)
Cadre relative	−0.013 *	−0.021	−0.018	−0.026 *
	(0.008)	(0.016)	(0.016)	(0.016)
Family Demographic Characteristics
Average education years	−0.001	0.004	0.005	0.006
	(0.004)	(0.01)	(0.009)	(0.009)
Total labor force	−0.014 *	−0.019	−0.019	−0.021
	(0.01)	(0.02)	(0.02)	(0.02)
Head of Household
Gender (nominal)	−;0.077 **	−0.027	0.016	−0.015
	(0.039)	(0.111)	(0.111)	(0.109)
Gender (actual)	−0.225 ***	−0.223 ***	−0.207 ***	−0.239 ***
	(0.031)	(0.06)	(0.06)	(0.058)
Age	0.005	0.026 **	0.025 **	0.024 **
	(0.006)	(0.012)	(0.012)	(0.012)
Age2	−0.00006	−0.0003 ***	−0.0003 ***	−0.0003 **
	(0.0001)	(0.0001)	(0.0001)	(0.0001)
Family Structure
Elderly + adult	0.027	0.011	−0.069	−0.033
	(0.062)	(0.142)	(0.139)	(0.138)
Adults	−0.023	−0.072	−0.112	−0.077
	(0.065)	(0.144)	(0.143)	(0.141)
Adults + children	−0.034	−0.177	−0.243 *	−0.193
	(0.067)	(0.146)	(0.145)	(0.143)
Elderly + adult+ child	−0.049	−0.071	−0.140	−0.085
	(0.067)	(0.149)	(0.146)	(0.145)
Geographical Features
Distance to the town	0.003 ***	0.0004	0.001	0.005 *
	(0.001)	(0.003)	(0.003)	(0.003)
Adjacent reserve	0.049 **	0.190 ***	0.206 ***	0.157 ***
	(0.024)	(0.048)	(0.048)	(0.047)
Constant	0.572 ***	0.083	0.161	0.047
	(0.157)	(0.337)	(0.335)	(0.328)
LR chi2(20)	230.18 ***	96.62 ***	102.89 ***	107.59 ***
Pseudo R2	0.403	0.486	0.518	0.542
Sample size	1,074	282	282	282

Note: * p < 0.1; ** p < 0.05; *** p < 0.01. Standard errors in parentheses.

. Column (1) is the estimated effect of farmers’ participation in relocation on IDES using the total sample. After accounting for missing and singular values, the final total sample size is 1074; For further analysis of the impact of different relocation characteristics on IDES, we included relocation type in the regression reported in Column (2), Column (3) with settlement mode included, and Column (4) with relocation time included.

Regression based on the total sample (Column 1) shows that participation in the relocation and resettlement program has a significant negative impact on farmers’ IDES. It appears that relocated households have less reliance on ES than non-relocated households.

As can be seen from regressions for relocated households, Table 5, Columns (2), involuntary resettlement has a significant positive effect on IDES. It shows that the voluntary relocation is more conducive to reducing farmers’ dependence on ES. From in-depth interviews with local villagers and officials during the survey, we found that voluntary relocation households were always more willing to move from the deep mountains. With the training program, as well as other financial support provided by the local government, they could participate more non-farm activities than the involuntary resettlement households, such as for engineering and disaster prevention. The former are willing to take the initiative to adapt to the changes brought about by the relocation and build new non-resource-dependent livelihoods.

From regressions for relocated households, Table 5, Columns (3), centralized resettlement has a significant negative effect on IDES. Centralized resettlement effectively reduces the dependence of farmers on ES. Compared with scattered resettlement such as “entering cities and towns” and “flower arrangement”, farmers in concentrated resettlement area have less reliance on ES. On the one hand, the support facilities available to the resettled communities are usually relatively better planned and implemented than for other communities, with highly accessible markets, matched industrial or modern agricultural park, and wider access to employment and business information resources, thus helping farmers to broaden their channels of income. On the other hand, centralized resettlement makes it easier for the government to arrange intensive training, employment guidance, increase the chances of migrant workers going out to work and the chance of developing non-agricultural business activities, all of which help to reduce the dependence on ES.

In addition, from regressions for relocated households, Table 5, Columns (4), immigration in the new stage has a significant negative effect on IDES. It shows that, compared with the early spontaneous and small-scale relocation, the new resettlement relocation project effectively reduces the farmers’ dependence on ES. Participation in this relocation project embodies government-led, systematic progress, and with related industrial and policy support measures will help to achieve non-resource-dependent changes in livelihoods of farmers.

For the control variables, among the household livelihood assets, household’s per capita arable land and their own assets have a significant positive effect on IDES, while housing valuation has a significant negative effect on IDES. The results for household livelihood assets show that the amount of natural capital owned by farmers affects their ability to obtain income from ecosystems. The more arable land per capita of households, the more provisioning services that farmers gained from the ecosystem, so that the proportion of non-social net income in total revenue increased to some extent. In the full sample, households with higher housing valuation rely more on ES. From the survey, the value of local residents’ houses is generally higher than that of relocated households. On the one hand, the indigenous inhabitants tend to have a more favorable geographical location. On the other hand, indigenous residents also have more land area, so they usually have more access to provisioning services. However, the result for the relocated households is reversed: the higher the valuation of their houses, the lower their reliance on ES. In addition to the structure and quality of housing, geographical location is also an important factor. Since the farmers who have a better location often have the opportunity to develop more non-agricultural activities, it reduces their dependence on ecological resources. In addition, the number of self-owned assets in the whole sample and relocation samples have opposite effects. Non-relocated households with more self-owned assets tend to have more productive tools or means of transport and tend to have the capacity to engage in local non-agricultural or to work outside. As a result, the provisioning services obtained from local activities make up a relatively small proportion. Assets owned by relocated households are more likely to be durable goods. The greater the investment, the more likely they are to settle in the area. With fewer local non-agricultural opportunities, household income relies more on the functioning of ES.

For family demographics, the gender of head of household has a significant negative impact on IDES. Female heads of household, both nominal and actual, have a significant negative effect on IDES. Compared with those having a male head, rural households with a female head decrease their dependency on ES. For the households with actual female heads, outmigration of male labor could return more remittance income than female migrants, which definitely dilutes the proportion of income from ES. For a nominal female head household, female leadership emphasizes security over profit from ES due to physical considerations. After meeting subsistence requirement, they prefer supplementary cash benefits from socio-economic activities. During the survey, we found a number of community workshops had been established in the resettlement community and were welcomed by the relocation families. Among the employees who participated in the production of handcrafts, plush toys as well as electronic components, 90% were women left-behind. Through such workshops, they were better able to coordinate earning money and taking care of children and elder. The square of the age and age of heads of household are brought into the model at the same time to see whether these ages have a “U” or inverted “U” shape with the IDES. The results show that these two variables had opposite effects, which means that the marginal effect of age will increase first and then decrease. Therefore, we can speculate that the relationship between the age of the head of household and IDES is an inverted “U” –shape.

Regarding geographical features, the distance to the town and closeness to an adjacent reserve also have a significant positive impact on IDES. Along with increased distance from a household’s house to the local town, households’ dependency on ES increased. In the mountain area, local village residents are spread widely over the mountain. A town consists of many administrative villages and gradually becomes a local administrative and market center with relative more population and better infrastructure. The closer the location to the town, the more convenient the traffic conditions and the more accessible is the market. We can speculate that the more opportunities for farmers to make a living by other means, such as non-farming activities, the less will be the reliance on ES. In addition, a protected area with abundant natural resources is usually restricted by relevant policies; thus farmers are less likely to engage in farming. However, there is an increased possibility of engaging in agritainment and tourism, which in turn increases the possibility of gaining access to cultural services.

5. Conclusion

As an important measure for poverty alleviation, this relocation program involved nearly 10 million poor people from 22 provinces in the China. It is totally different from other "original site" poverty alleviation measures, and is not a simple spatial population displacement and community reconstruction. It is giant system of engineering, involving the distribution of population, reconstruction of resources and environment, as well as adjustment of the economy and society. As the main participants and stakeholders of the program, rural household’s livelihoods are crucial to resolving the development dilemma and achieving a win-win situation in terms of ecosystem conservation and human wellbeing.

Using data collected from rural households in the mountainous areas of Southern Shaanxi, this study analyzed from a household perspective the impact of the relocation and resettlement program on the ecosystem service dependence of farmers. The results show that the implementation of relocation and resettlement can effectively reduce the dependence of farmers on ES. Participation in the program helps to optimize the farmer’s income structure and reduce provisioning services from ecosystems, as well as increase the proportion of income from social-economic activities. Voluntary relocation, centralized resettlement and relocation in the new stage have played a positive role in reducing farmers’ dependence on ES. Voluntarily relocated households are more willing than non-voluntarily relocated farmers to take the initiative in responding to the negative changes in the outside world and taking new opportunities to change livelihoods, so as to complete more quickly the transition to a non-resource-dependent livelihood model. Centralized resettlement can generate some economies of scale and policy spillover effects. Farmers have the opportunity to obtain more matching follow up support to improve their ability to take advantage of external opportunities and resources and broaden their access to non-resource-based income. South Shaanxi represents the relocation and resettlement policy in the new stage, where the measures and subsidies for the relocation program have been greatly enhanced compared to previous ones. At the same time, it has also strengthened the generation of the relocated peasant household’s self-development ability during the non-agricultural transformation.

In the process of constructing the index of dependency on ecosystem services of farmers, we integrated and quantified the various benefits that farmers obtain from ES at the micro-scale. The index system and analysis in this paper make a useful complement to the research and application of ES and micro-family well-being. It provides a basis for balancing the relationship between ES and well-being, as well as references for the analysis and evaluation of policies related to ecological and anti-poverty development. From the perspective of policy formulation, the effective measurement and decomposition of dependence of relocated households on the ES also helps to identify appropriate policies while enhancing relocated households’ sustainable livelihoods and resilience.

Southern Shaanxi is the birthplace of the anti-poverty relocation and settlement policy during the “13th Five-Year Plan”. To resolve the development dilemma of ecological conservation and human development, “relocation” is only a “means” while poverty alleviation is the goal. To a certain extent, this study affirms previous policies and measures, especially the voluntary principles, centralized resettlement and mode of urbanization, which were emphasized by the government. The active role these policies play in achieving the goal of “ecological protection “has also provided the basis for the subsequent policy improvements. “Getting out of poverty” is the precondition for achieving the goal of “ecological protection.” Otherwise, relocated households will continue their existing livelihoods with high ecological dependence. Our research confirms the positive effect of the relocation program in solving the development dilemma “the environment of a place can’t afford the inhabitants in the place”. Relocation can stimulate the relocated family to reduce their dependence on ES by revising household livelihoods’ strategies. On one hand, family’s main labor can be transferred to non-farm sectors after resettlement, which can increase the socio-economic income, for example, by out-migrating to an urban area. On the other hand, in the original community, households can return their farm land to forest to obtain a subsidy from Grain to Green or transfer the management right of contracting land to obtain land circulation income. Even though the available land is scarce in the new locations, this research contributes by analyzing factors that specifically affect ES dependence and that exploring other non-provisioning services that may reduce such dependence. For this reason, government non-agricultural guidance and follow-up support is particularly crucial. In addition, while emphasizing reduction of the dependence on ES, it is especially crucial to adjust the income structure from ES; for example, to encourage relocated farmers to obtain more cultural services from the ecosystem, such as ecotourism, agritainment, as well as to increase the proportion of income from farming non-timber forest products and other provisioning services, rather than traditional farming and forestry. Only in this way, can we reduce the extraction of resources directly from the ecological environment, and truly achieve a “win-win” environment for sustainable development of both population and ecology.