Does poverty-alleviation-based industry development improve farmers’ livelihood capital?

LIU Ming-yue,FENG Xiao-long,WANG San-gui,ZHONG Yu

1 Institute of Agricultural Economics and Development,Chinese Academy of Agricultural Sciences,Beijing 100081,P.R.China 2 College of Economics and Management,China Agricultural University,Beijing 100083,P.R.China 3 School of Agricultural Economics and Rural Development,Renmin University of China,Beijing 100872,P.R.China

Abstract Targeted poverty alleviation is a unique approach adopted in China to help achieve the vision of a moderately prosperous society in all aspects and the “Chinese Dream”. Industrial development as a means of poverty alleviation is an integral part of the “Five-pronged Poverty Alleviation Measures” Project of targeted poverty alleviation,and a critical foundation for other poverty alleviation measures. In this study,a comprehensive evaluation method was used to measure farmers’livelihood based on the framework of sustainable livelihood. Specifically,the effects of industrial development on farmers’livelihood capital were estimated by employing the propensity score matching combined with the difference-in-differences(PSM-DID) approach. Findings suggest that industrial development had a significantly positive effect on the livelihood capital of farmers. Industrial development can significantly enhance farmers’ human,social and financial capital,while it cannot significantly affect the natural and physical capital. Industrial development had heterogeneous effects on farmers’livelihood capital,more efficiently impacting on the non-poor than the poor. The effects on farmers’ livelihood capital varied across regions,with Guizhou experiencing a larger effect than Sichuan. However,the effect was insignificant for farmers in Gansu. To improve farmers’ livelihood capital,it is necessary to take measures to strengthen their human capital,promote the innovation of financial products,and make good use of their social capital; it is also essential to strengthen the support of industrial development to the poor.

Keywords:targeted poverty alleviation,poverty alleviation by industrial development,the poor,livelihood capital,PSM-DID approach

1.Introduction

Among the 17 sustainable development goals of the United Nations for the period up to 2030,“End poverty in all its forms everywhere” ranks first on the list (Tollefson 2015;Liet al.2016). Being the largest developing country in the world,China never stops its efforts in poverty alleviation.Over the past decades,China has implemented diverse anti-poverty programs,including relief-type poverty alleviation,development-oriented poverty alleviation,and targeted poverty alleviation,which have already formed a poverty alleviation system that adapts to China’s path of development. Therefore,a solid foundation has been set to eliminate absolute poverty (Liu Yet al.2017; Liuet al.2020). Especially since the 18th National Congress of the Communist Party of China,China has started the“targeted poverty alleviation” strategy,which significantly reduced the amount of poor in rural areas and made a significant contribution to global poverty reduction. The poor population dropped from 98.99 million in 2012 to 5.51 million in 2019,and the corresponding poverty headcount ratio reduced from 10.2% in 2012 to 0.6% in 2019 in China(NBSC 2020). Poverty alleviation by industrial development(PAID),an essential component of the “Five-pronged Poverty Alleviation Measures” Project,aims to build the internal driving force of poor regions and the poverty-stricken population to promote development,which is crucial in achieving the country’s poverty alleviation goals by 2020 (Liuet al.2020). Meanwhile,PAID is an essential foundation for making poverty alleviation measures efficient,including poverty alleviation relocation,ecological protection,and education development (Liu M Yet al.2019).

Anti-poverty practices across the world,either in developed or developing countries,all attach great attention to PAID,with a particular focus on the role of the agricultural industry and tourist industry in increasing family income and improving the living conditions of the poor (Manyara and Jones 2007; Spenceley and Meyer 2012; Chenet al.2018).In China,previous studies about PAID mainly focus on three aspects:its concepts and theoretical basis (Xu and Liu 2011;Wang and Zheng 2017),operating mode and mechanism(Xu and Liu 2011; Liu J Set al.2017; Liu M Yet al.2019; You and Wei 2020),and implementation difficulties and optimal solutions (Li and Zuo 2016; Xu and Li 2017; Linet al.2018).However,there are few studies evaluating its effects. One possible reason is that the return from agricultural industry development often requires multiple years or even longer time periods,which makes it challenging to collect panel data needed to assess the effect of agricultural industry development on farmer livelihood. The existing studies only focus on the effect of PAID on the family income of farmers. The results showed that PAID could improve the participation of the poor in agricultural planting and livestock breeding activities by 12 and 14%,respectively (Huet al.2018),which significantly increased their agricultural income(Liu W Bet al.2019; Wuet al.2019; Liet al.2020).

Since the late 20th century,poverty alleviation theories have been considering the capacity building of the poor as an indicator of the effectiveness of poverty alleviation approaches. According to the capability-based theory proposed by Amartya Sen (1999),the internal reason for poverty is the shortage of social existence,adaptability,and development ability. Therefore,it is necessary to rebuild the individual ability to avoid and eliminate poverty (Sen 1999).Sustainable livelihood is an anti-poverty concept aiming to build the poor’s capacity to get rid of poverty. It is a means of living based on the capabilities,assets (stores,resources,claims,and access) and activities,which can cope with and recover from stresses and shocks,maintain or enhance its capabilities and assets both now and in the future (Ellis 2000; Krantz 2001). PAID is an important measure taken by “blood-making” poverty alleviation to motivate the selfdevelopment ability of the poor. However,past studies on the impact evaluation of PAID mainly focus on family income of farmers,neglecting whether farmers’ ability has been improved or whether their livelihood becomes more sustainable. Therefore,this study uses the propensity score matching combined with the difference-in-differences (PSMDID) approach to evaluate the effect of PAID on farmers’livelihood based on the sustainable livelihood framework.

2.Background,data,and methods

2.1.Background of PAID

PAID mainly takes the resource endowment of poor areas as the foundation and the market demand as the guidance and utilizes specialized cooperative organizations of farmers to establish the agricultural industrial chain model that includes production,processing,transportation,and marketing and sales. PAID in this study refers to a comprehensive project of poverty alleviation and development jointly implemented by the central government of China and the World Bank in rural areas in three provinces in 2014. The PAID project involved 27 counties -six in Sichuan,five in Guizhou,and 16 in Gansu. Its major stakeholders included the World Bank,Chinese governments at all levels,cooperatives organizations,technical advisory groups,cooperative counselors,and local farmers.

The World Bank provided financial,technical,and regulatory services. The governments provided supporting policies such as financial and infrastructure construction and coordinated the project implementation. Cooperatives used the financial and technical support to achieve selfdevelopment of agricultural industries,which included improving local farmers’ agricultural production skills through technical and information services,assisting them with the purchase of agricultural production factors (e.g.,fertilizer,pesticide,seeds,and livestock),sales,storage,processing,and transportation of agricultural products.The technical advisory groups mainly provided advisory services for the implementation of the project. Cooperative counselors offered guidance and services for the operation of cooperatives. Farmers used their productive factors,such as labor force,land,and capital,to develop the agricultural industry under the drive of cooperatives,to increase household income and improve the ability of selfdevelopment. Participation was voluntary for both poor and non-poor farmers in the project counties.

In addition,the selection of the agricultural industries was mainly based on the existing foundation of industrial development and resource endowment,combined with the current market demand in the local areas. The major agricultural industries selected by the project included cash crops (e.g.,sorghum,lily,and apple) and livestock breeding(e.g.,sheep and beef).

2.2.Data

In this study,multi-stage sampling techniques were used to select the sample for the baseline survey in 2015. First,eight counties were selected randomly from 27 counties in three provinces,including Longxi,Yongjing,Huanxian,Jingning in Gansu,Chishui and Xishui in Guizhou,and Xuyong and Meigu in Sichuan. Second,nine villages were selected randomly in each chosen county. Finally,all households were ranked according to their household income. About 20 households were randomly selected by stratified equidistant sampling in each village,which resulted in a sample size of 1 440 farmer households. The followup survey was conducted in 2018,resulting in 868 valid households recorded in both surveys. Sample attrition in the follow-up survey was due to the following; first,Longxi was no longer a member county of the project in 2018 due to a project adjustment in Gansu,and therefore all the samples in this county were not present in 2018; second,some households,especially those from Yongjing and Meigu,could not be contacted because all their household members left the original residence to engage in off-farm employment. In general,sample attrition in the follow-up survey occurred naturally without human interference.Table 1 presents the sample distribution. There were two different questionnaires,the village questionnaire,and the farmer questionnaire. The village questionnaire covered village information including the village’s basic characteristics,village infrastructure,and public service in 2014 and 2017. The farmer questionnaire covered information,including basic household characteristics,project participation details,and income source in 2014 and 2017,respectively. There were 370 respondents in Gansu,229 in Sichuan,and 269 in Guizhou,accounting for 42.63,26.38,and 30.99%,respectively,of the total sample.

2.3.Model specifications

Livelihood capital index for farmersThis study uses a comprehensive indicator framework method to calculate farmers’ livelihood capital index that consists of five components based on the sustainable livelihood framework:natural,physical,human,financial,and social capital. For each component,there are several contributing indicators.These indicators for each component,measured by numerical values,represent the household’s characteristics in relation to each component. The livelihood capital index can be used to evaluate the influence of PAID on farmers’livelihood. According to the past studies (Ellis 2000;Nelsonet al.2010; Xu and Yue 2012; Zhuet al.2016;Ning 2017; Xu and Shen 2018),15 indicators form the livelihood capital index (Table 2). Specifically,they include land area,cultivated land area,and woodland as well as grassland area that are considered as natural capital;productive assets,durable consumer goods and house value as physical capital; labor force,education level and health condition as human capital; net income per capita,money lent to others and formal as well as informal loan amounts as a financial capital; and expenditure on social interaction,annual communication expense and township cadres as social capital.

Following Ahsan and Warner (2014),there are four steps to calculate farmers’ livelihood capital index. First,the original indicators are standardized by:

whereiis theith indicator;jis thejth farmer;Pijis the standardized variable for indicatoriand farmerj;xijis the original variable for indicatoriand farmerj;xi,maxandxi,minare the maximum and minimum respectively of the original variable fori. After standardization,the numerical value of each indicator ranges from zero to one.

Second,based on the standardized value of eachindicator and its weight,the value of each capital can be obtained by:

Table 1 Sample distribution

Table 2 Indicators of livelihood capital indexes and their respective weights

whereWiis the weight of theith indicator,PijWiis the score of theith indicator,ais the number of indicators belonging to each capital,andCPis the estimated score of each capital.

Third,the estimated scores for the five capitals can be obtained by:

whereCPZ,CPW,CPR,CPJ,CPSare scores of natural,physical,human,financial,and social capital,respectively,of farmers.

Finally,according to the scores calculated above,the livelihood capital index of farmers can be calculated by:

whereCPZis the livelihood capital index for farmers and ranges between zero and one,where zero stands for the weakest sustainable livelihood while one suggests the strongest sustainable livelihood.

The entropy weight method was adopted to estimate the weight (Wi) of each indicator. The concept of information entropy,first introduced by Shannon (1948),seeks to define the weight of indicators according to the degree of indicator variation. It can measure the amount of useful information with data (Işık and Adalı 2017). Generally,when the difference of values among the evaluating objects on the same indicator is high while the entropy is low,it suggests that this indicator provides more useful information. Thus,the weight of this indicator should be set high correspondingly.On the contrary,a low difference value and high entropy indicate a relatively small weight (Zhanget al.2014; Işık and Adalı 2017). Hence,the entropy weight method can eliminate the subjectivity of each indicator as much as possible,resulting in an evaluation more consistent with the objective reality (Zouet al.2006; Işık and Adalı 2017).Followed by Zhanget al.(2014) and Işık and Adalı (2017),this study used the entropy weight method to calculateWiof each indicator in eq.(2). If the method was used to calculateWiof livelihood capital index in 2014 and 2017,the estimated index will not be comparable. We,therefore,follow the technique proposed by Xionget al.(2015) to merge the data of two periods for the purpose of weight calculation.The details of the indicators of livelihood capital index and their estimated weights are shown in Table 2.

Impact of PAID on livelihood capitalThis study used the difference-in-differences method to evaluate the impact of PAID on the livelihood capital of farmers. In a natural experiment setting,the difference-in-differences method can validly control the differences of research objects before the project implementation and eliminate the fixed effects,which can obtain the real effect of the project (Ravallion 2005;Ali and Abdulai 2010). This method has been widely used to evaluate the impact of policy and technology adoption(Ravallion 2005; Ali and Abdulai 2010).

The difference-in-differences method assumes that the treatment and control groups have no significant difference and similar influence factors before the policy intervention.This hypothesis cannot be satisfied in this study’s scenario because farmers’ participation in the project will result in the selection bias problem arising from observed factors.The PSM-DID approach can correct the selection bias stemming from observed factors (Lei and Lin 2010). The PSM method is used to “balance” the program between the treatment and control groups,and make the groups as similar as possible with respect to those observed baseline characteristics according to the propensity scores.To estimate the propensity scores,a multinomial logistic regression is estimated. The PSM-DID estimator (Smith and Todd 2005; Azam 2018; Duong and Thanh 2019) is presented as:

2.4.Descriptive results

The livelihood capital index,natural capital,physical capital,human capital,social capital,and financial capital were selected as dependent variables to assess the effects of PAID.

The treatment variable used in the analysis referred to farmers’ participation in PAID,with 1 representing participation and 0 otherwise. To calculate farmers’ propensity score,we selected a vector of exogenous variables as the independent variables to predict participation. They include basic characteristics of the head of household (e.g.,gender,age,and education),household (e.g.,poverty status,household size,dependency ratio,and off-farm work),and villages (e.g.,the existence of the cooperative organization,availability of agricultural skill training,organized off-farm work and the distance to the nearest county and market). A logistic regression model was used to generate a propensity score for each farmer in the sample. Depending on the propensity score,the PSM method matched the treatment group (participants) and the control group (non-participants)who were similar in observed characteristics (e.g.,gender,age,education,and household size) to address the potential selectivity bias arising from observed factors. Table 3 presents the descriptive statistics for the variables used in the PSM-DID approach.

The results show that the average value of farmers’livelihood capital index in 2014 was 0.14,which increased to 0.16 in 2017,which means that farmers’ livelihood ability remained at a low level despite a slight increase. The scores for some of the five components of the livelihood capital index had a significant increase between 2014 and 2017.Specifically,the average score of farmers’ human capital reached 0.5 in 2017,which is the highest among the five components. By contrast,the score of physical capital was the lowest -0.03 in 2014 and 0.05 in 2017. The average score of farmers’ social capital increased from 0.09 in 2014 to 0.15 in 2017,which was the largest increase among the five components. The results also show that there were 244 households participating in PAID,accounting for 28.11% of the total sample.

There are no obvious differences in the independent variables between 2014 and 2017,and the following describes the main characteristics in 2017. Most heads of households are males (92.74%),the average age of the heads of households were 53 years,and the mean number of years of schooling was 5.80 years. It seems that heads of households were mainly middle-aged and not well educated. There were 396 poor households,accounting for 45.62% of the total sample. The mean household size was 4.93,of which 0.88 had off-farm employment,and the mean dependency ratio was 0.39. In terms of basic village characteristics,90.44% of the sample villages had farmers’professional cooperatives organizations,84.91% organized agricultural skill training for farmers,and 64.63% organized local farmers to work outside the village. The mean distance from the villages to the nearest county is 57.42 km,and the distance to the nearest market is 8.47 km.

Table 3 Summary statistics of the selected variables in the propensity score matching combined with the difference-in-differences(PSM-DID) approach

3.Results and discussion

3.1.Results of propensity score matching

Table 4 presents the results of the logistic sample selection model for project participation. The results show that heads of households’ age and education level are the most important determinants of farmers’ choice of participation.Age had a negative and significant influence on participation.The possible reason is that the older heads of households,as the decision-maker of agricultural production in the family,usually are more risk-averse and tend to have a “wait-andsee” approach in the face of a new program. Education had a positive and significant impact on farmers to take part in the project. Higher educated farmers tend to have higher learning abilities and are more likely to try new programs.

Turning to the factors at the household level,poverty status,and household size are the most important factors affecting farmers’ participation. Specifically,poverty status had a positive and significant impact on participation. On the one hand,the project specifically targeted the poor,and consequently,poor households were more likely to participate. On the other hand,the non-poor households,who usually have more abundant livelihood capital and diversified livelihood strategies,may find participation less necessary. Household size had a positive and significant impact. Large households usually have more labor resources to be invested in the project.

Results also show that the availability of the cooperative organization and geographical location of the village are important determinants of participation. The cooperative organization had a positive and significant impact as it can help farmers better understand the industrial development program’s operation details. Distance to the county was negatively and significantly correlated with participation.The village close to the county would provide more nonagricultural employment opportunities for local farmers,reducing their participation incentive. Similarly,distance to the market had a negative and significant impact on participating. It seems that villages close to the market have access to a more developed market economy that providesdiversified livelihood options,reducing farmers’ enthusiasm in participation.

Table 4 Regression result of logistic model for participating in poverty alleviation by industrial development (PAID)

Although the PSM method can eliminate the selection bias stemming from observed factors,it is necessary to check if at each value of the propensity score,the variables have the same distribution for the treatment and control groups (Lee 2013). Table 5 presents the results of the balance test before and after matching. The results show that before matching,the treatment and the control groups were systematically different in terms of certain observed characteristics,while after using the kernel-based matching technique none of the mean differences of the selected variables between the treatment and the control groups were statistically significant,even at the 10% level,suggesting that the balancing property of the covariates was satisfied(Maet al.2018). Meanwhile,to satisfy the common support hypothesis,following Suet al.(2019) and Lianget al.(2020)we excluded individuals that are not on the common support and used samples only on the common support to estimate the effect of PAID on the livelihood capital of farmers. This resulted in 45 households not on the common support being excluded,and 823 households being retained after matching.

3.2.Effects of PAID on farmers’ livelihood capital

Table 6 presents the average treatment effects (ATTPSM-DID)of PAID on farmers’ livelihood capital estimated by the PSM-DID approach.1ATTPSM-DID is estimated using the Stata diff command (Villa 2016).The results show that before project implementation in 2014,the mean livelihood capital index for farmers in the treatment group was 0.145,only 0.003 lower than that in the control group. That means that there was no significant difference in the livelihood capital index between the treatment and the control groups at the beginning. After project implementation in 2017,the mean livelihood capital index for farmers in the treatment group increased to 0.173,and that for farmers in the control group increased to 0.164.The difference between the two groups was 0.008,which was statistically significant at the 5% level,implying that the livelihood capital index for farmers in the treatment group was significantly higher than that in the control group. The treatment effect on farmers’ livelihood capital index was 0.011,which was statistically significant at the 5% level.This finding suggests that PAID improved farmers’ livelihood capital index significantly. When farmers participate in the project,their livelihood capital,such as financial,human,and social capital,were improved.

The results show that PAID played a significant role in increasing farmers’ human,social,and financial capital.Specifically,before implementing the project (2014),the mean score of human capital for farmers in the treatment group was 0.506,0.007 lower than that of the control group,and the difference between them was insignificant. After implementation in 2017,the mean score of human capital for farmers in the treatment group increased to 0.527,which was 0.018 higher than that of the control group and statistically significant at the 10% level. The treatment effect on human capital was 0.025,statistically significant at the 10% level. Participants in the project can receive various technical training in agricultural planting or livestock breeding provided by the cooperatives or the government that enhanced their human capital. The mean score ofsocial capital for farmers in the treatment group was 0.087,which was 0.006 lower than that of the control group. The difference between them was statistically insignificant before project implementation in 2014. After project implementation in 2017,the mean score of social capital in the treatment group increased to 0.163,which was 0.013 higher than that of the control group. The treatment effect on social capital was 0.019,statistically significant at the 5% level. When farmers participated in the project,they built strong personal and social connections with other farmers,managers of the cooperative,members of technical advisory groups,and technical extension workers of the government,and spent more on communication and social contact to obtain technical and market information related to agricultural production,which increased their social capital.

Table 5 Results of balance test of the unmatched and matched samples using kernel-based matching

Table 6 Effects of poverty alleviation by industrial development (PAID) on livelihood capital

Before project implementation in 2014,the mean score of financial capital for the treatment group was 0.045,which was 0.001 higher than that of the control group,and the difference between them was insignificant. After project implementation in 2017,the mean score of financial capital for the treatment group increased to 0.056,while that for the control group remained at the same level in 2014. The mean score of financial capital for the treatment group was 0.012 higher than that of the control group,which was statistically significant at the 5% level. The treatment effect on financial capital was 0.011 and was statistically significant at the 10% level. There are two possible reasons. First,participants can increase their family income because their agricultural production skills and efficiency were improved by attending the cooperatives’ skill training activities. Second,governments or cooperatives can provide formal or informal financial loans to farmers to be used for investment in agricultural inputs (e.g.,pesticides and fertilizers,agricultural machinery,and agricultural irrigation facilities),which enhanced agricultural production efficiency,and resulted in increased household income and consequently the improvement of their financial capital.

3.3.Heterogeneous effects of PAID on livelihood capital for poor households and non-poor households

There were 396 poor households and 472 non-poor households in the sample. Due to the differences in their household characteristics,the effects of PAID may vary between the two groups. The PSM-DID approach was used to analyze the treatment effect on the poor and non-poor households,respectively. The results (Table 7) show that when considering the selection bias arising from observed factors,the treatment effect on livelihood capital index for poor households was 0.009 with no statistical significance,while that for non-poor households was 0.014,which was statistically significant at the 5% level. This finding implies that the effects of PAID for poor and non-poor households were heterogeneous,with a more significant and positive influence on non-poor households. The possible explanation is that,on the one hand,compared with nonpoor households,poor households tend to have a relatively weak level of education,health status,and acceptance ability,which results in PAID to have a less significant impact on the livelihood capital of the poor households than that of non-poor households. On the other hand,although the projects of PAID are inclined to the poor households,the precision and strength of supporting measures in human,physical and social capital may be not enough,resulting in a non-significant impact on the livelihood capital of poor households. Therefore,it is necessary to innovate the supporting mode to improve human,physical,and social capital for poor households to enhance their livelihood capital.

The results show that the effects of PAID on poor and non-poor households varied by the dimension of livelihood capital. Specifically,the treatment effect on human capital for non-poor households was 0.037 and statistically significant at the 5% level. The non-poor households with a high level of education tended to take part in technical training in industrial development to improve their skills of agricultural production,which contributed to increasing human capital. The treatment effect on social capital for non-poor households was 0.036,while being statistically significant at the 5% level,indicating that PAID only had a significant impact on social capital for the non-poor households. Based on the stronger communication skills,non-poor households can rapidly expand their social network by participating in PAID,and consequently,increase their social capital. By contrast,the treatment effect on financial capital for poor households was 0.019,which was statistically significant at the 5% level.

PAID had a significant impact on the poor households’financial capital,which may be explained by two possible reasons. First,PAID can increase the operating income,wage income,and property income of poor households,improving their financial capital. For example,PAID may completely change poor households’ traditional agricultural production methods,improve the quality of agricultural products and production efficiency,and bring an increase in operating income. PAID can also provide temporary employment opportunities for family members of the poor,which can increase their wage income; and some poor households can use their land,capital,and other resources to become shareholders in PAID and obtain dividend income,thereby increasing their property income.Second,local governments usually provided direct financial support,such as interest-free loans,industrial funds,and subsidies,to the poor participants. However,the impact of PAID on poor households’ livelihood capital was not significant,mainly because the livelihood capital includes natural,physical and human capital. The improvement of these capitals is less direct than that of financial capital,which needs a long-term cultivation process. Meanwhile,the current measures for the improvement of these capitals may be insufficient,resulting in a slower growth rate of livelihood capital.

3.4.Regional effects of PAID on livelihood capital

There were noticeable differences among the three provinces (Gansu,Sichuan,and Guizhou) in terms of economic development,the structure of the agricultural industry,poverty status,and the progress of project implementation. Thus,the effects on the livelihood capital of farmers may vary among the three provinces. Table 8 shows that PAID significantly and positively impacted farmers’ livelihood capital index in Sichuan and Guizhou,with a treatment effect estimated at 0.019 (significant at the 5% level) in Sichuan and 0.031 (significant at the 1%level) in Guizhou,respectively. Guizhou experienced the largest effect,followed by Sichuan. However,the effect in Gansu was not statistically significant. This result is closely related to the selection of agricultural industry in these three provinces. PAID in Guizhou and Sichuan mainly dependson annual commercial crops (e.g.,sorghum and lily) and livestock breeding (e.g.,sheep and beef) with a one-year production cycle,which can improve the livelihood capital of farmers in the short term. However,PAID in Gansu depends on permanent crops (e.g.,apple) with a production cycle of at least five years,resulting in an insignificant effect between 2014 and 2017.

Table 7 Treatment effects of poverty alleviation by industrial development (PAID) on livelihood capital for poor and non-poor households

Table 8 Effects of poverty alleviation by industrial development (PAID) on livelihood capital in different regions

The project had a significant and positive influence on farmers’ human capital in Guizhou,with a treatment effect estimated at 0.064 and significant at the 10% level. This result is consistent with that derived from the whole samples(Table 6). The treatment effects on farmers’ social capital were 0.033 (significant at the 10% level) in Sichuan and 0.049 (significant at the 5% level) in Guizhou. The project had a significant and positive effect on farmers’ financial capital in Guizhou,estimated at 0.025 and statistically significant at the 1% level.

4.Conclusion and policy implications

This study measures the livelihood capital of farmers by a comprehensive evaluation method. It then employs the PSM-DID approach to evaluate the effect of PAID project on farmers’ livelihood capital,using the framework of sustainable livelihood. It was found that first,PAID had a significantly positive influence on farmers’ livelihood capital,estimated at 0.011. Second,PAID can significantly enhance farmers’ human capital,social capital,and financial capital,while its influence on their natural capital and physical capital was insignificant. Third,PAID had heterogeneous effects on farmers’ livelihood capital that varied by poverty status.Specifically,the influence on livelihood capital for the nonpoor was more positive than that for the poor. PAID had a significant influence on human capital and social capital for the non-poor households,while it had a significant influence on financial capital for the poor households. Finally,the impacts of PAID on the livelihood capital varied among the three provinces,with farmers in Guizhou experiencing the largest impacts,followed by those in Sichuan,while those in Gansu were not significantly impacted. The effects of PAID on human,social,and financial capital also varied among these provinces.

Although the project has improved the sustainability of farmers’ livelihood under evaluation,it nevertheless remains at a low level. This study provides the following recommendations. First,the government should continue to enhance poor farmers’ livelihood capital,which can be done through promoting their human capital,including improving personal knowledge,skills,and health,which builds up their self-confidence and enables them to better participate in PAID. The government also needs to promote innovative financial products (e.g.,microfinance and agricultural insurance) to provide diverse financial support to local farmers in developing the agricultural industry. Moreover,the government can play an active role in facilitating farmers to build and expand their social network and accumulate social resources. These measures combine to increase livelihood capital and enhance the development capacity and resilience of farmers. Second,the government should pay more attention to whether and to which extent the poverty-stricken households can benefit from the PAID projects. Findings in this paper suggest that PAID may significantly improve the livelihood capital of non-poor households rather than that of its target group -the poor households. The government may need to provide more incentives (e.g.,credit and infrastructure) to poor households to increase their participation rate.

On the one hand,it is necessary to set up a model of industrial development and hold various propagandizing and educating activities about agricultural production,which can dispel the misgivings of poor households to participate in PAID projects and stimulate their enthusiasm to develop the agricultural industry. On the other hand,targeted measures (e.g.,training in planting and breeding techniques,educating and credit) should be taken for poor households to encourage and help them participate in PAID. In addition,the government should pay more attention to cultivate and develop local cooperatives and improve their operating capacity and profitability. The cooperatives should provide the service and training on agricultural production and products sale to improve the self-development ability of poor households. Poor households should actively participate in projects of local agricultural industries suitable for their circumstances (e.g.,labor,lands and capital) to improve the capacity and efficiency of agricultural production,and thus achieve sustainable household livelihood.

Acknowledgements

This work was supported by the National Natural Science Foundation of China (72003185 and 71661147001). The authors thank to the editors and the two anonymous reviewers for their valuable suggestions.

Declaration of competing interest

The authors declare that they have no conflict of interest.