LENG Gan-xiao,FENG Xiao-long,QIU Huan-guang

1 School of Agricultural Economics and Rural Development,Renmin University of China,Beijing 100872,P.R.China 2 College of Economics and Management,China Agricultural University,Beijing 100083,P.R.China

Abstract This paper analyses the impact of the poverty alleviation relocation (PAR) program on rural household income and evaluates the heterogeneous income effects of various relocation modes,based on a panel dataset of relocated households from 16 counties in eight Chinese provinces. The results show that participation in the PAR increases the income of both rural and urban resettlers. More specifically,it has a significant positive effect on agricultural and wage income for rural and urban resettlers,respectively. Further analyses show that the income increase for rural resettlers was mainly due to agricultural technology training and that the income increase for town resettlers was attributed to medical security. For the village resettlers,policies should focus on strengthening the development of local industries and training of agricultural technologies. For the urban resettlers,non-agricultural employment and public services in the urban resettlement areas should be promoted.

Keywords:targeted poverty alleviation,poverty alleviation relocation,relocation mode,income effects

1.Introduction

Poverty is a global problem,and poverty alleviation attracts worldwide attention,especially in developing countries(Bapna 2012; Griggset al.2013; Maniet al.2013). “End poverty in all its forms everywhere” is the first goal on the list of the Sustainable Development Goals of the United Nations for the period through 2030 (Liu Met al.2019).Several studies have found that poverty reduction is highly related to economic growth (Ravallion and Jalan 1999) and diversified poverty alleviation measures such as government expenditure and transfer (Fanet al.2000),infrastructure construction (Agénoret al.2008),and the improvement of agricultural technologies (Smith 2004; Ali and Abdulai 2010).

Over the past 40 years,China has made great efforts on poverty reduction. China’s rural poor population dropped from 250 million in 1978 to 16.6 million in 2018 (Liu Met al.2019). The targeted poverty alleviation strategy has been implemented since 2013 and significantly influences poor households’ production and lifestyles (Liet al.2016;Zhouet al.2018). The poverty alleviation relocation (PAR)program is one of the “Five-pronged Poverty Alleviation Measures” Project in China and is a key to solving extreme poverty in areas with poor living conditions. The PAR program’s goal is to move 10 million poor people living in deep mountains or earthquake-prone areas to suitable areas to fundamentally change their living environments and improve their living standards and self-development abilities by 2020. By the end of September 2020,almost 9.61 million poor people had been relocated,accounting for the finish work for relocation.

Resettlement has been finished by the end of 2020.Short-term effects have appeared,and potential problems have also arisen. The PAR program has primarily changed the lifestyle and living environment of the resettlers. It will also improve household income,which is essential for sustainable livelihood (Khatiwadaet al.2017). The purpose of this study is to examine the effects of the PAR program on income and to identify the potential mechanisms through which the program affects household income.

Several studies have analyzed the effects of the PAR program on farmers’ livelihoods and income growth. It was demonstrated that relocation significantly reduced the number of poor people (Wanget al.2018; Liu Wet al.2019),changed livelihood strategies and habits (Khatiwadaet al.2017),and improved household livelihoods (Liuet al.2018;Guo and Li 2019). Case studies and qualitative analysis were used to explore the effects and challenges of the PAR program (Xueet al.2013; Wilmsen and Wang 2015; Zouet al.2019). Scholars have also studied similar poverty alleviation relocation projects in developed countries,for example,the Housing Opportunities for People Everywhere(HOPE VI) project and the Moving to Opportunity (MTO)project in the U.S.in the mid-1990s. Besides other socialeconomic effects,they found that those programs improved young people’s employment and increased their income,but the effect on older people’s income was insignificant (Klinget al.2007; Ludwiget al.2013; Chettyet al.2016).

Few studies,however,analyzed the heterogeneous effect on the income of different relocation modes and explored how these effects were achieved using largescale field survey data. This study has the following three contributions. First,the study uses a large-scale panel dataset of rural households. Unlike previous studies that have examined the PAR program’s effects through case studies and qualitative analysis based on small sample sizes,we collected a three-year panel dataset covering 8 provinces and 80% of the relocated population. Second,although the existing studies have evaluated the effect of relocation policy,an in-depth study of the impact mechanism lacks. We classify household income and relocation mode,analyze the change in several income categories,and explore the potential mechanisms for the income change through agricultural production and public infrastructure and services. Third,because the participation in the other relocation projects previously analyzed was voluntary,households endogenously decide where to reside based on economic considerations. The relationship between relocation and income is often hindered by selection bias.However,such problems do not exist in our study because the households that participated in the PAR program had not been given detailed information on the relocation (Lo and Wang 2018). Thus,based on this unique exogenous policy experiment,we are able to disentangle the causal relationships linking relocation and income.

Based on a large-scale rural household survey repeatedly conducted for three years,this study aims to empirically investigate the effects on income of the participation in the PAR program and different types of relocation modes.Several important implications for poverty alleviation and rural development are provided. The remainder of this paper is organized as follows. Section 2 introduces the policy background and theoretical framework. Section 3 introduces the data sources,model framework,and variable descriptions.Section 4 reports the empirical results and Section 5 presents conclusions and policy recommendations.

2.Policy background and theoretical framework

2.1.The PAR program

Poverty alleviation relocation program started in China in the 1970s. To solve drought and water shortages in some areas,poor populations were moved to areas with more abundant water resources. Irrigation projects were also built in relocation areas to improve the poor people’s agricultural production. By the end of 1982,the relocation plans were carried out in the Three West Areas (Hexi and Dingxi in Gansu Province and Xihaigu in Ningxia Hui Autonomous Region). The construction of agricultural facilities was also strengthened in these areas. In 1994,the “Seven-Year Priority Poverty Alleviation Program” initiated a voluntary poverty alleviation migration program,in addition to the conversion of farmland to forest and grass programs. The next relocation scheme was carried out in Inner Mongolia Autonomous Region,Guizhou Province,Yunnan Province,and Ningxia Hui Autonomous Region in 2011,and then proliferated to 17 provinces throughout the country. The National Development and Reform Commission also launched an auxiliary investment program to support the PAR. A total of 6.8 million people were relocated between 2001 and 2015.

Since 2016,the Chinese government has invested almost 946 billion CNY on the PAR and aimed to move 10 million people out of the areas with poor environment conditions.This project was a nationwide PAR plan proposed in the 13th Five-year Plan,which is the largest of its kind in the past 70 years in terms of the investment scale. Seventy percent of the relocated populations are concentrated in the western region,of which 59% are in Guizhou,Sichuan,Yunnan,Hubei and Shaanxi provinces. By September 2020,a total of 9.61 million people have moved to the relocation areas. And almost 7 million people from western China and 2.4 million from central China have moved,accounting for 73 and 25% of the national total,respectively. Of these relocated populations,47% moved from villages to towns or cities (urban relocation) and 53% to nearby villages(rural relocation). In the case of urban relocation,resettlers are provided with houses and medical and educational resources. For rural relocation,resettlers are provided with houses and farming opportunities.

2.2.Theoretical framework

Different modes of PAR have different effects on resettlers’income. Because of the changes in their livelihood capital and physical environment,households’ livelihood patterns will also change. Below we present a theoretical framework for different types of resettlers and propose two hypotheses.

First,the village resettlers obtain better-quality land and usually participate in government-led farming projects,thus gaining agricultural income. Because these resettlers are far from their original land in mountains or barren areas,they rent new land in the resettlement areas or obtain cheaper seeds and fertilizer in their agricultural production.Because the land quality in the resettlement areas is much better than that in the original mountainous areas,the factor markets of agricultural products in the resettlement area are more developed,and the price of inputs is lower,and farm production and income are much higher. On the other hand,industrial development for poverty alleviation mainly concentrates on agricultural operations and is implemented in villages. Thus,it promotes the development of the agricultural production in the resettlement areas for the village resettlers. By participating in agricultural production and joining agricultural cooperatives,relocated poor households can enhance their agricultural productivity and improve their farm income (Ma and Abdulai 2016). We propose the following hypothesis:

Due to agricultural development and land resource endowment,village resettlers’ agricultural income will increase after resettlement.

Second,urban resettlers can acquire more employment opportunities and better public services,which contribute to the improvement of their wage income. Owing to improved spatial and geographical location of urban areas,the population is concentrated and industrialization is more robust,which gives young resettlers more employment opportunities. The higher employment rate can lead to increased wage income of relocated households. Public services,such as education and medical care,are much better than those in rural areas (Goldman 2001). On the one hand,good public services promote a sense of security for urban resettlers. On the other hand,good educational resources and healthcare provide development space for urban resettlers (Guo and Li 2019). Besides,convenient transportation infrastructure reduces the travel costs of migrant workers and can promote wage income. Thus the second hypothesis is proposed as follows:

The wage income of the relocated households will increase because the urban resettlers can access better education and healthcare as well as more convenient transportation conditions.

In the framework presented in Fig.1,we highlight the paths of the effects of the PAR program on household income for different groups. According to the above analysis,we expect that the agricultural income of the village resettlers will increase,and the wage income of urban resettlers will increase. As wage income and agricultural income are essential parts of the total household income,the poor who participate in the PAR program can increase their total income and improve their lives compared to the households that do not relocate.

Fig.1 Theoretical framework of the effect of the poverty alleviation relocation (PAR) program on resettlers’ income.

3.Data and methods

3.1.Data

The data used in this study are from the annual survey of resettlers conducted by the Institute of Poverty Alleviation of the Renmin University of China. The survey was conducted mainly in five provinces and one autonomous region in the western region (i.e.,Sichuan,Yunnan,Guizhou,Shaanxi,and Gansu provinces and Guangxi Zhuang Autonomous Region) and two provinces in the central region (i.e.,Hubei and Hunan provinces),covering the main implementation areas of the PAR program in China.

The survey of poor households was conducted for three consecutive years. The first survey took place between July and August 2016,and approximately 90 enumerators conducted the survey. The survey date was selected because most people had not yet moved by the time of the survey. In the first survey,we collected 1 971 valid samples of poor households. The second follow-up survey took place between July and August 2017; we tracked the same people and obtained 1 758 valid household samples.In April and May 2019 (i.e. three years after the baseline),a repeated survey obtained approximately 1 408 valid household samples. Among them,we found that 343 families had decided not to relocate or had chosen to access different poverty alleviation policies. These samples were dropped. Also,129 household samples were invalid and do not have wage income and agricultural income in the family. Therefore,we use a balanced panel for the years 2016,2017,and 2019. The final dataset includes 2 808 observations of 936 poor households. The survey included basic household characteristics of the population,basic living conditions,income,consumption,employment,social integration,and relocation modes.

The PAR program is almost finished. According to our survey in 2019,94% of the new houses have already been set up,the keys of 87% of the houses have been given to resettlers,and 84% of the families have already moved in. Among the relocated,35% moved to small towns or industrial parks,i.e.,“moving to towns (MTT)”. The rest 65% of them moved into a nearby village,namely “moving into the village (MIV)”. The relocation types vary from county to county. We found that village resettlement is the most common type of relocation in Yunnan,Sichuan,Hubei,Gansu,and Shaanxi provinces. In Guangxi Zhuang Autonomous Region,Hunan,and Guizhou provinces,urban resettlement is the main relocation mode. The distribution of relocation types is presented in Fig.2.

Fig.3 shows the proportion of different relocation types by survey round. Most new entries into the PAR program in our sample were from the second and third rounds. Most households in the PAR program moved to the village (23% in 2017 and 55% in 2019),while the rest moved to towns or did not move. Since the type of relocation in the PAR program is not the result of household choices and participating in different types of relocation has different effects on the poor,we focus on two binary variables,namelyMIVi,t,andMTTi,t.If householdimoved into the village by timet(t=1,2,3),MIVi,ttakes a value of 1. If householdimoved to a town by timet,MTTi,ttakes a value of 1.

3.2.Econometric method

Fig.2 Sample geographical distribution of relocation types in 2019. Source:Authors’ calculation based on survey data.

Fig.3 The proportion of different relocated types of the sample in three years. Source:Authors’ survey data.

We employ a fixed-effect econometric model to explore the PAR program’s effect on household income using a panel dataset. The following econometric models have been estimated. First,we estimated the total effects of relocation on household income,which includes the direct effects and time effects. Second,we measured the effects of different relocation modes on different household income. Finally,we explored the impact mechanism of different relocation modes on household income. We begin with the following household-level fixed effect model to control for baseline differences and some factors that do not change over time at the household level.

whereYi,tis an outcome of interest for householdiin the survey at timet,including total income,agricultural income,and wage income.μiis the fixed effect of timeinvariant characteristics of householdi,andδtare survey round dummies.Ri,tis a dummy variable which equals 1 if householdireceived treatment and relocated in yeart,and 0 otherwise. The coefficient ofγ1represents the average treatment effect of participating in the PAR program. We include seven additional time-varying controls in this regression (i.e.,Zi,t,j),such as household size,the average age of household members,dependency ratio,the proportion of healthy family members,education,asset ownership,and loan condition. Standard errors are clustered at the household level at timet.

Second,to explore the time effect of the PAR program on household income,we added the relocation time variable.The equation is as follows:

whereRTi,tequals the relocation duration of the householdiat timetand is a continuous variable.RTi,tequals 0 if the householdihave not relocated at yeart.γ2represents the average treatment in household income per additional month after the relocation. The other variables are defined as in eq.(1).

Next,the different effects of relocation modes on household income have been estimated by the following equation:

whereμi,δt,andZi,t,jare the same as above.MIVi,tindicates whether householdihas relocated to the village at timetand equals 1 if the household relocated to the village.MTTi,tindicates whether householdihas moved to a town at timetand equals 1 if the household moved.β1arepresent the direct income effect of village resettlement,and the coefficient ofβ2arepresents the income effect of urban resettlement,compared to un-relocated households.And the time effects of different modes of relocation are estimated as the following equation:

where the variableRTi,trefers to the relocation duration.β1bandβ2bare the average time effects of different relocation modes for differing households participating in the PAR program.β1brepresents the average time effects of households that relocated to the village on household income.β2bindicates the average time effect of town resettlement. We expect that the longer the time of residence in the new location,the more income increase.Standard errorsεi,tare clustered at the household level and at timet,which means that this is a time-individual fixed-effect model.

After examining the overall effect and the time effect,we estimate a fixed-effects model to explore the influencing mechanism of the PAR program’s effects on household income. The analysis follows three steps. First,we focus on the effects of the PAR program on the infrastructure and the household living conditions,which might influence household income. Second,we examine the effects of infrastructure and household living conditions on household income. Third,we explore both direct and indirect effects of relocation on income by adding the intermediate variables and relocation. The three-step equation is as follows:

whereXi,trepresents the intermediate variables,i.e.,the factors that affect agricultural income (e.g.,technology training and farm size) and factors that influence wage income (e.g.,the distance to road,school,or hospital).γbandβbin model (5) show how the relocation project has affected intermediate variables,including infrastructure and household living conditions.in model (6) indicates the effect of the infrastructure and household living conditions on household income. According to the middle-effect method described by Wenet al.(2004),the third regression estimates the total effects of relocation,which is measured asfor village resettlers,andfor urban resettlers.γaandβaare direct effects of relocation,anddenote the average indirect treatment effects of relocation on household income through intermediate variables.

3.3.Variables description and summary statistics

The main outcome variable is household income,which includes total income,agricultural income,and wage income. The total household income is the sum of agricultural income,wage income,property income,and transfer income. We focus on two main sources of household income:agricultural income and wage income.Agricultural income comes mainly from crop cultivation,and wage income primarily comes from manual work and other hired jobs. Most poor households in rural areas rely on agricultural income and wage income,which are more sustainable than transfer income and property income.Thus,it is crucial to analyze improvement in the two types of income after relocation and find the influencing factors.The household per capita income has been deflated by rural and urban consumer price index (CPI) at 2016 prices,according to theChina Statistical Yearbook(2019).

The average household total income per capita is 3 663 CNY,which is slightly higher than the standard poverty line of 2 952 CNY in 2016. In 2016,the average total income per capita was 2 483 CNY,which is less than the standard poverty line. In 2019,the average total income increased to 4 924 CNY -almost double the baseline level.The average wage income of the sample is 1 374 CNY per capita. And the wage income has increased from 801 CNY in 2016 to 2 048 CNY in 2019. The average agricultural income of the sample is 1 210 CNY per capita in three years. The agricultural income has also increased from 961 to 1 448 CNY per capita. Overall,household income has increased between 2016 and 2019. Also,the household income per capita refers to the disposable income of the permanent residents in a household.

The key independent variables are relocation,i.e.,village or urban resettlement. According to the PAR program,approximately half of the poor households had urban resettlement,where they could enjoy better transportation facilities,healthcare,and educational resources. The other poor households in the PAR program were village resettlers,including moving to nearby villages,where their lives mainly depend on agriculture and livestock production. We treat the two relocation types as treatment groups,while the un-relocated households are treated as the control group. Because there are two preconditions,there exists no endogeneity issue. First,because the PAR program was not a typical “voluntary resettlement”program,the relocated households could not choose the relocation type and location. After the housing was built,the poor households living in deep and remote mountainous areas moved mostly involuntarily. And after relocation,the original house will be demolished and reclaimed almost in a year. Thus,participation in the PAR program and the two treatment groups are random. Second,the low-income households lived in deep and remote mountainous areas prior to relocation; the living conditions and total household income were similar at less than 2 952 CNY per capita every year,according to the standard line of poverty (Liu Met al.2019). Thus,we hypothesize that the three groups have the same chances of acquiring information on PAR policy and similar livelihood capital and capacity,meaning that they would all have the same trend of increasing household income if they had not participated in the PAR program.

The mean proportion of relocation status and relocation duration are shown in Table 1. The percentage of resettlers was 28.6% in 2017 and 83.4% in 2019,respectively. The percentage of village resettlers was 22.8% in 2017 and 54.7% in 2019,respectively. The percentage of urban resettlers was 5.9% in 2017 and 28.7% in 2019,respectively.The relocation duration was on average 1.6 months in 2017 and was approximately 14.8 months in 2019.

To better understand the PAR program’s influence on household income,we further explore the indirect effect by including intermediate variables. The variables reflecting the effects on agriculture of the PAR program included agricultural farm size in operation and whether family members attended agricultural technical training. The farm size represents a household’s agricultural ability,and governmental technical training can help farmers improve their skills and find alternative livelihoods (Wanget al.2018). The non-agricultural impact of the PAR program is reflected by the distance to the nearest asphalt or cement road,the distance to the primary school,and the distance to the nearest hospital,which represents the quality of the local infrastructure and public services. The average household operating land area was 6.6 mu (1 ha=15 mu).The percentage of households receiving technical training was 24.3%. The average distance to a road is 1.2 km. The average distance to a primary school is 6.2 km,and the average distance to the nearest hospital is 12.7 km. The distances to the road,to the school,and to the hospital have all decreased from 2016 to 2019.

To isolate the effects of the PAR program on household income,we also control some household characteristics. In the fixed-effect model,some factors that do not change over time are controlled. Time-variant household characteristics that influence income include household size,age,dependency ratio,the proportion of healthy family members,the proportion of family members with more than nine years of education,household assets,and loan status. Table 1 reports the descriptions of all variables and the summary statistics of the total sample and in different years.

3.4.Preliminary analysis

We now discuss the descriptive statistics and preliminary analysis of the effects of relocation on household income.There are three separate groups (i.e.,un-relocated group,village resettlers,and urban resettlers),and we have observed their differences in household income in 2016 and 2019. Besides,we calculate the difference in household income between the two years,and at-test was conducted to compare the changes from 2016 to 2019 for different groups. The results are shown in Table 2.

The increase in mean household income per capita is different among the three groups from 2016 to 2019. In the baseline year of 2016,the un-relocated households were more prosperous than the households in the other two groups (i.e.,village resettlers and urban resettlers). In 2019,the total income per capita of the households that moved into villages and towns after relocation,especially households that moved to villages,was higher than the income of households that were un-relocated. The income increase of urban resettlers is largest among the three groups,at a significant 2 749.3 CNY. Whereas the incomeincrease of un-relocated households is only 1 253.4 CNY.The household income of village resettlers has significantly increased by 2 641.1 CNY. The increase in agricultural income for village resettlers is 747.9 CNY per capita,which is four times greater than that of the other two groups.The increase in wage income for households that moved to towns is approximately 1 812.1 CNY per capita,which is more than that for the other two groups. Thus,it can be concluded preliminarily that the village resettlers have increased their agricultural income and total income,and the urban resettlers mainly increased their wage income and total income. Relocation can improve the total income; the increase of total income after urban resettlement is larger than that of village resettlement and not relocating.

Table 1 Variables description and summary statistics

Table 2 A t-test of difference of the average household income per capita of households of different relocation modes in 2016 and 2019 (CNY)1)

4.Results

Table 3 presents the estimated direct and time effects of relocation on total household income,wage income,and agricultural income. Table 4 shows the results of the income effects of different relocation modes. Tables 5 and 6 present the intermediate effect of village resettlement and urban resettlement on household income through technology training,distance to a hospital,and other factors. Table 7 shows the robustness test results using the difference-indifferences with the propensity score matching (PSM-DID)approach.

According to the econometric model in Section 3.2,the results of eq.(1) that represent the direct effects of relocation on income are shown in columns 2 to 4(Table 3),and the outcomes of eq.(2) estimating the time effect are given in columns 5 to 7. The heterogeneity analysis eqs.(3) and (4) of different relocation modes are presented in Table 4,and direct effects are shown in columns 2 to 4,and time effects are in columns 5 to 7.Eqs.(5),(6),and (7) that represent the intermediate effect of participating in the PAR program on household income are given in Tables 5 and 6. Next,we show robustness test results on the PAR program’s effect on income using the PSM-DID approach proposed by Heckmanet al.(1997).Table 7 presents the income differences before and after relocation for each of the three groups.

Table 3 Impact of the relocation on household income per capita (CNY)1)

Table 4 Impact of the poverty alleviation relocation (PAR) modes on household income per capita (CNY)1)

4.1.Impact of the PAR program on household income

Using panel data and the fixed-effect model,we have investigated the short-term effect of participating in the PAR program on household income. First,the total effects of relocation on income have been estimated. We find that,compared to un-relocated households,the relocated households have improved their total income after the relocation. The total household income per capita for those relocated was 555.05 CNY more than that of the unrelocated households at the 1% significance level. Next,the agricultural income of relocated households has increased by 239.24 CNY (P＜0.05) compared to that of un-relocated households. Second,the time effects of relocation on income are shown in columns 5 to 7 in Table 3. It can be seen that the wage income increased significantly with relocation duration,but the total income and agricultural income of PAR participates have not increased with relocation duration.It is probably because the increase of total income and agricultural income is a short-term phenomenon,and the influence of resettlement on agricultural income is faster than on wage income. After relocation,the household members would spend time finding jobs to increase the wage income.However,they can quickly get used to farming work in the new setting after relocation such that the agricultural income has grown at a faster pace.

We find that different relocation modes have different effects on income. First,it can be seen that both village and urban resettlements have significant and positive impact on the total household income per capita. Compared to the unrelocated households,the total income of village resettlers increased by 537.25 CNY per capita. The total income per capita of urban resettlers increased by 590.42 CNY,which is higher than the income increase of village resettlers.This is probably because the urban resettlers have more opportunities to access a wider range of job opportunities.This finding is consistent with the results obtained by Liuet al.(2018).

Table 5 Impact of the poverty alleviation relocation (PAR) modes on intermediate variables1)

Table 6 Impact of intermediate variables on household income (CNY)1)

Second,our analysis found that urban resettlers’ wage income significantly increased and that village resettlement had a significant positive impact on their agricultural income. Compared to un-relocated households,the annual wage income per capita for urban resettlers increased by 412.45 CNY,and the agricultural income per capita of village resettlers significantly increased by 360.08 CNY. The reason is that urban resettlers have more opportunities to access employment information for migrant workers and gain wage income in industries or companies. This finding is in line with Guo and Li (2019) and Nguyen and Nordman(2018),who pointed out that urban resettlement helps relocated people obtain more non-agricultural employment opportunities. This positive effect on the wage income of rural-to-urban migrates corroborates the findings of Wilmsenet al.(2011) and Lianget al.(2012) on the participants who have relatively more local wage income in urban. Besides,the positive effect on agricultural income for the village resettlers results from the development of agricultural production in the resettling areas. The positive impact onagricultural income for village resettlers is in accordance with previous studies that have demonstrated that the “whole village relocation”resettlement model in China has been relatively successful and that the land conservation program of China has benefited the poor (Uchidaet al.2007; Zouet al.2019).

Table 7 Robustness test using the difference-in-differences with the propensity score matching (PSM-DID) approach1)

The time effect of participating in different modes of PAR program was examined by adding interacting time variables with the variables representing the two modes. Table 4 shows significant time effects of town resettlement on wage income and agricultural income and village resettlement on agricultural income. The longer the duration after urban relocation,the more wage income they obtain in resettlement and the less agricultural income they acquire. For urban resettlers,wage income significantly increased by 41.42 CNY per capita after one month in the relocation site,while agricultural income decreased by 18.65 CNY. This is because the livelihood mode of urban resettlers changed from agricultural work to non-farm employment in the short run,which is in accordance with the results reported by Liet al.(2018)and Guo and Li (2019). For the village resettlers,the wage income was insignificantly increased by each incremental month of relocation.However,the longer the period of village relocation,the higher the gained agricultural income. The agricultural income of village resettlers increased by 15.8 CNY after each incremental month of relocation. It seems possible that these results are due to the rapid development of agricultural production in the resettlement areas. However,the total income did not significantly increase in the two groups compared to the un-relocated households,which is probably because the total income includes other types of income and cannot change in a short period of relocation time. In general,the results regarding the time effects on household income do not present much difference from the results of the previous analysis.

The significant effects on the household income of some control variables are interpretable. The larger the household size,the less household income per capita the household has. Likewise,the higher average household age,higher proportion of healthy people,higher education conditions,and more household assets are significantly related to a higher household income. Healthy people in the household contributes to wage income; and educated people in the household benefits from agricultural income.

4.2.Intermediate effects on household income

In this part,we focus on the intermediate effects of participating in the PAR program on household income. First,we estimate the impact of village resettlement and urban resettlement on technological training,land operating area,road distance,school distance,and hospital distance. Second,we explore the effects of the five intermediate variables on household income. Next,we add both intermediate variables and relocation into the model to estimate the direct and indirect effects of relocation on income. By analyzing the effects of the PAR program on intermediate variables and the influence of mediators on household income,we can explore the impact mechanism and examine the most important factors impacting income for resettlers.

The increase in household total income and agricultural income is due to the technical training offered in the resettlement areas for village resettlers. Tables 5 and 6 show that the village resettlers have more opportunities to access agricultural technological training after relocation than the un-relocated populations; and technological training has a significant positive effect on total household income and agricultural income per capita. Using a fixed-effect Probit model,the results show that village resettlement has a significant positive effect on the probability of acquiring agricultural technological training. Likewise,the households receiving technological training increase total household income by 481.26 CNY and agricultural income by 367.55 CNY,compared to the non-participating households.The direct effect of village resettlement is 440.84 CNY on total income and 295.59 CNY on agricultural income. The indirect effect of village resettlement is 84.07 CNY on total income and 64.9 CNY on agricultural income through technological training (marginal effect on technical training is 0.19).

The finding of the correlation between technological training and household income is consistent with the results obtained by Wilmsenet al.(2011),who pointed out that the investment in science and technology can increase resettled households’ income. And science and technology development also contributes to achieving sustainable development goals and increasing human welfare (Bapna 2012). However,even though the increase in farmland operating area is related to the increase in total household income and agricultural income per capita,participating in the PAR program does not impact the farmland operating area. Therefore,the farmland operating area is not a possible influencing intermediate path for the positive effect of the PAR program on household income.This may be because of the difficulty of renting land from others in resettling areas. Thus,demand for the land of resettlers decreases (Yinet al.2016). Although involuntary resettlement reduces farmland area,improvements in quality and access to irrigation lead to a substantial increase in both yields and total grain output,which may contribute to the rise in agricultural income (Webber and McDonald 2004).

For urban resettlers,the improvement of healthcare is the main reason for the increase in total household income and wage income. This study indicates that moving to towns leads to a significant decrease in the distance to roads,primary schools,and hospitals. Village resettlement only has a slightly negative effect on the distance to a cement road. Compared to un-relocated households,the distance to cement roads of urban resettlers significantly decreased by 1.52 km (P＜0.01),the distance to primary schools decreased by 5.46 km,and the hospital distance decreased by 10.44 km. The distance to smooth roads for village resettlers is approximately 0.92 km less. These improvements in the living environment for the urban resettlers is in line with the findings of other studies,which show that resettlement can help poor people obtain access to safer,healthier,and more supportive environments with better schools and more responsive services (Dickinson and Webber 2007; Chaskin 2013).

Table 6 shows that the decrease in hospital distance is associated with an increase in total household income and wage income. For a one kilometer decrease in the distance to the nearest hospital,the total household income increases by 21.02 CNY per capita,and the wage income increases by 20.1 CNY per capita. Also,the relocation has no direct effects on income as shown in columns 5–7 in Table 6. The indirect effects on total income can be calculated according to the method mentioned above. The total income has significantly increased by 220.72 CNY and the wage income has increased by 183.73 CNY through the decrease in the distance to the hospital for urban resettlers. This can be explained by the possibility that households living in resettlement areas are close to hospitals,feel safer and healthier than the un-relocated households,and do much more non-agricultural work. This finding corroborates the ideas of Guo and Li (2019) who suggest that healthcare facilities have a positive impact on the non-agricultural employment of relocated households.

4.3.Robustness test

To test the robustness of the above results,we use the PSMDID approach to estimate the PAR program’s income effect for different relocation modes. The underlying assumption of difference-in-differences is that households would have exhibited similar trends in the control and treatment groups after solving the differences between the two groups(Heckmanet al.1998; Gibson and McKenzie 2014).Because the households do have different characteristics,we first estimate propensity scores,match the control group with the treatment group following the kernel matching approach,and finally perform the difference-in-differences with the matched samples. The results are shown in Table 7.

We compare the total household income,wage income,and agricultural income before and after relocation in the three groups. We have the following three observations.First,compared to un-relocated households,the village resettlers significantly increased their total income and agricultural income,as can be inferred in columns 2 to 4.Prior to relocation,the agricultural income of the two groups was varied slightly,while the agricultural income of village resettlers was significantly higher than that for the un-relocated households after relocation. This indicates that moving into a village has a purely positive effect on agricultural income. Second,the urban resettlers increased their total household income and wage income compared to the un-relocated households. It can also be suggested that the increase in wage income is due mainly to urban relocation. Finally,in comparing the changes in household income of village and urban resettlers,we show that urban resettlement has a significantly positive effect on wage income and has a negative effect on agricultural income compared to village resettlers. The outcomes presented in this section are consistent with the previous results,proving the robustness of the estimation.

5.Conclusion and policy recommendations

This paper examined the impact of participation in the PAR program on total household income,wage income,and agricultural income for resettlers. The paper explored the factors influencing the income of resettlers using the panel dataset of 936 resettlers from Sichuan,Shaanxi,Gansu,Yunnan,Guizhou,Hunan,and Hubei provinces and Guangxi Autonomous Region from 2016,2017 to 2019. The average household income for resettlers of different types of relocation in 2016 and 2019 were compared,and the impact of the PAR program on different income sources were analyzed. Given that these comparisons do not account for the other factors that affect household income,we further employed time and individual fixed-effect regression model,which takes into account the observed and unobserved factors that do not vary over time,to solve the selection bias problem and address the endogeneity problem. Also,to test the robustness of the results,this study applied the PSM-DID approach.

The empirical results showed that participation in the PAR program could increase the total household income for resettlers. Village resettlement has a positive effect on agricultural income,and urban resettlement is related to an increase in wage income. In particular,compared to the un-relocated households,the total income and agricultural income of resettlers significantly increased by 555.05 CNY and 239.24 CNY per capita. The total income and wage income for urban resettlers significantly increased by 590.42 CNY and 412.45 CNY per capita.The village resettlers increased total household income by 537.25 CNY and agricultural income by 360.08 CNY. This finding suggests that relocation increases income of the poor in China. Agricultural income is important for village resettlers,and wage income greatly contributes to the economic prosperity of urban resettelers. The results also show that the improvement in healthcare of urban resettelers delivers significantly positive effects on the rise of their wage income. The increase in agricultural income for village resettlers mainly depends on the agricultural development and technological training in the resettlement areas.

Three policy recommendations are proposed. First,the PAR program has had remarkable achievements in increasing the income of poor households and eliminating poverty. Therefore,the government should continue to concentrate on poor people who participate in the poverty alleviation relocation program after 2020 and help them to get a better environment. Moreover,the government could ensure that the resettlers and the original residents share equal opportunities to obtain a job or to acquire public services. Second,technological training is also an important factor that influences the increase in agricultural income for village resettlers. Thus,the government should intensify support for training in agricultural technologies and enhance the development of the agricultural industry in resettling villages. Third,the positive and significant impact of access to hospitals suggest that improving healthcare and public services in resettling urban areas would enhance farmers’senses of security,which further increases farmers’ wage income and household welfare. Finally,China has a dual system where there are substantial gaps between urban and rural populations. In the future,government should gradually improve the system for urban–rural integration and provide equal access to public services and social security for the resettling populations to ensure sustainable employment and income.

Acknowledgements

This work was supported by the National Natural Science Foundation of China (71861147002 and 71761147004) and supported by the Fundamental Research Funds for the Central Universities,and the Research Funds of Renmin University of China (20XHN086).

Declaration of competing interest

The authors declare that they have no conflict of interest.