Climate extremes, political participation and migration intentions of - - PDF document

1

Climate extremes, political participation and migration intentions of farmers: Case study in western China

Yan Tan, Xuchun Liu

Department of Geography, Environment and Population The University of Adelaide Adelaide, SA 5005, Australia Phone: (+61) 08 83133976 E-mail: yan.tan@adelaide.edu.au; xuchun.liu@adelaide.edu.au

Paper prepared for the XXVIII IUSSP International Population Conference. 29 October–4 November. Cape Town, South Africa.

2 Abstract: Understanding migration intentions helps predict actual migration behaviour and provides evidence for developing effective strategies for adapting to climate extremes. Households’ migration intentions are usually not motivated directly by climate extremes but shaped by their experience of climate impacts and adaptive capacity. One important determinant of households’ adaptive capacity is their political participation: position in the local political structure and participation in decision-making process. Incorporating the ‘goal pursuit theory’ and the conceptualised multi-staged decision making about adaptation, this study quantifies how rural households’ experience of climate extremes and their political participation influenced their migration intentions in terms of strength of migration intention and target destinations. Regression analysis is used to analyse survey data collected from 436 households in Minqin county of Gansu province, western China in 2012. The results show that some drought induced stresses—land loss and deteriorated water quality, as experienced by rural households in the study area—significantly increase their intentions to migrate and stimulated their tendency to undertake long-distance and rural-to-urban migration. Reduced agricultural production and declined water quantity due to climate impact can reduce farmers’ intention to migrate. A high level of household political participation in the local community is also positively correlated with strong intention to migrate and undertake long-distance and rural-to-urban migration. This is compared to a negative association with migration intention of those households with a high level of socio-economic status. The findings have implications for making effective migration and planned relocation policy and programs for affected groups and their communities.

3

• 1. Introduction

Goal pursuit theories used in psychology research suggest that intention is a significant predictor of ‘intention realisation’ (Gollwitzer and Sheeran 2006:71). Intention has two forms: ‘goal intentions’ refer to what one ‘intends to achieve’ and ‘implementation intentions’ that specify ‘when, where, and how one intends to achieve it’ (Gollwitzer and Sheeran 2006:82). Broad migration research has substantiated that people having intentions to move in an earlier time are more likely to migrate later on than those showing no intentions (Gordon and Molho 1995, Böheim and Taylor 2002). Li et al. (2014) points out that understanding migration intentions induced by environmental change is useful to predict people’s actual migration behaviour in the future and thus helps policymakers and practitioners to develop effective strategies for climate adaptation. However, how climate extremes influence migration intentions remains understudied. While a growing number of studies make an important contribution to our knowledge about people’s intention to migrate or not migrate in the context

• f climate variation or change (e.g., Mortreux and Barnett 2009, Kuruppu and Liverman 2011,

Abu et al. 2014), they are limited by having small non-random samples and not assessing the mechanisms for how experience of climate impacts and their political participation impact on migration intentions. Further, very little is known about the factors conducive to ‘implementation intentions’ that relate to migration, such as where to migrate, and a major impediment to our understanding is the lack of suitable data that enables such analysis. Conceptual frameworks aiming to unravel the complex relationship between climate change and adaptation have suggested that the adaptation decision-making process is multi-staged in most circumstances (Black et al. 2011). Climate change does not directly lead to adaptation (including migration), but is linked indirectly through its impacts on the economic, social and

4 political factors of adaptation (Foresight 2011). Adaptation is rooted in people’s actual experience of climate impacts (Tan et al. 2015). Grothmann and Patt (2005) suggests that the severity of threat experienced by people plays an essential role in motivating adaptation intentions through shaping risk experience. Yet there is scant empirical research addressing the relationship between experience of climate impacts and adaptation intentions in terms of whether to migrate or not and where to migrate. Studying the relationship between climate extremes and migration intention needs to be placed in a broad context that involves multiple factors that influence people’s adaptive capacity. Even under the same environmental, social, economic and political situations, people’s experience

• f climate extremes and adaptive capacity can vary significantly between households and

consequently differentiate people’s migration intentions. Among various household factors, political participation is proven essential for promoting adaptation to climate change (Bouriaud et al. 2015; Dutra et al. 2015). Political participation comprises various types ranging from holding party membership, voting, protest and political action to civic engagement and community participation (van Deth 2014:361). However, how different dimensions of political participation influence adaptation to climate extremes is scarcely studied. Even less is known about the role of political participation, such as holding party membership, involvement in political institutions and participation in decision-making, in affecting households’ adaptation intention to climate impact. Such political participation factors are particularly important in influencing human mobility in areas where climate change and associated environmental degradation have induced self-motivated migration and government-led relocation. Minqin county of Gansu province, northwest China (hereafter Minqin), the case study area of this paper, provides an example in point.

5 Minqin is an ecologically fragile area that has significant implications for ecological sustainability of northwest China due to its geographical location—the downstream area of the Shiyang River (the third largest inland river in China) (Zhang et al. 2004). In 2007, the Chinese central government issued a comprehensive Governing Plan for Focal Issues in the Shiyang River Basin (NDRC and MWR 2007) that includes a set of ecological regeneration, climate adaptation, and human resettlement programs in Minqin. Adaptation and resettlement schemes in Minqin, guided by the Governing Plan, are therefore deeply associated with the political process in the region. At the household level, people’s position in the power structure of local communities, access to public resources, and participation in decision-making processes become important explanatory factors of this paper. A research gap exists in the climate change–mobility evidence base, especially when this relates to China. There is a clear need for robust investigations of migration intentions of rural residents affected by climate extremes and the mechanisms that lead to their migration intentions to improve adaptation policy and practices of adaption (including migration) for the benefit of affected people and their communities. Grounded within the goal pursuit theory, this study aims to fill in part of this knowledge gap by quantifying how rural households’ migration intention is influenced by their experience of climate extremes and by household political participation in the study area. We address two categories of migration intentions: ‘goal intentions’ (representing the strength of migration intention) and ‘implementation intentions’, especially target destinations to which people intended to move. The paper begins with an overview of the scant literature on migration intention in the context

• f climate change. This is followed by a discussion of methodology that involves three parts:

an overview of goal pursuit theories, a multi-staged decision-making process of climate

6 adaptation and migration, and a conceptual framework that this study constructed for unravelling the nexus between experience of climate impacts, political participation and migration intentions of rural households in Minqin, China. This is followed by a description of the research setting and data collection. An analysis of the survey data, employing the experience of climate impacts–migration intention framework and using the two-stage regression models, follows. Following a discussion of the findings of the study, the paper concludes with policy implications for adaptation and migration policy and practices in China.

• 2. Literature review: Experience of climate extremes, political participation and

migration intention 2.1 Experience of climate extremes and migration intention Experience of climate extremes refers to self-assessed severity of climate extremes that people have experienced in the past in this paper. Such experience is a fundamental factor shaping people’s responses to climate change and extremes (Grothmann and Patt 2005, Frank et al. 2011). Climate extremes do not impinge on migration directly in many instances but through impact

• n other drivers of migration including economic, social and political conditions (Black et al.

2011, Foresight 2011). Even though people experienced the same climate risks, their perception

• f, and subsequent responses to, the risks could be quite different. Experience of actual threat

was conceptualised by Grothmann and Patt (2005) as an important determinant of risk perception of climate change and adaptation intentions in their MPPACC model. Some researchers (e.g., Frank et al. 2011) used people’s experience of particular climate hazards as

7 an indicator of risk perception—experience-based perception of risk to predict adaptation. Our study employs this indicator as an explanatory variable of the analysis. Farmers’ adaptation means were particularly intervened by their experience of climate impact

• n natural resources (e.g. water and land) and agricultural production (Ezra and Kiros 2001,

Feng et al. 2010, Massey et al. 2010). These studies showed that declines in agricultural production, crop failure and food shortages due to climate extremes significantly influenced adaptation strategies across different rural settings. Personal experience of climate extremes has a positive association with risk perception (Grothmann and Patt 2005, O'Connor et al. 2005, Siegrist and Gutscher 2006, Akerlof et al. 2013). For that reason, risk perception, which ‘assesses the probability and severity of a hypothetical threat in the future’ (Grothmann and Patt 2005:205), was identified to be a significant contributor to increased adaptation intention

• f people when facing environmental hazards (O'Connor et al. 1999, Sheeran et al. 2014,

Yazdanpanah et al. 2014). However, some studies substantiated that experience of particular climate risks and associated risk perception alone are inadequate to motivate adaptation (Tucker et al. 2010, Frank et al. 2011). Methodologically, some researchers (e.g. Mortreux and Barnett 2009; Kuruppu and Liverman 2011) used descriptive statistics of data collected in their case study areas to analyse the relationship between climate variability (or change) and migration intention. Others (e.g. Abu et al. 2014, Li et al. 2014) employed logistic regression models for analyses

• f their survey data. These studies addressed the first aspect of migration intention—goal to

migrate or not. Neither addressed the strength of people’s desire for that goal, nor the aspect of implementation intentions such as when, where, and how to migrate.

8 2.2 Political participation, adaptation and migration intention Political participation in terms of participation in public decision-making process is claimed to enhance adaptive capacity and institutional ability to cope with climate impact (Bouriaud et al. 2015; Dutra et al. 2015). Unequal participation in decision-making processes and unequal access to governmental assistance influenced people’s adaptive capacity and hence differentiated their experience of climate impacts and the effectiveness of adaptation (Thomas and Twyman 2005, Paavola and Adger 2006). In the rural Chinese context, the accessibility of rural residents to natural and public resources and local power was usually measured in two indicators: whether a person is a member of the Chinese Communist Party (CCP); and whether a person is a cadre serving the government (Nee and Lian 1994, Bian 2002). Party members and village cadres had a stronger adaptive capacity than the ordinary people because the former group had privileges to access valuable political and economic information, to establish favourable personal linkages with the outside parties, and to access job opportunities or natural resources (Morduch and Sicular 2001). Tan et al. (2015) identified that a higher degree of participation in decision-making is positively associated with adaptation to climate stressors in the urban setting of China. Li et al. (2014) analysed the relationship between the factor of membership of CCP and the outcome factor of migration intention induced by environmental stressors, but no strong association between the two factors was found. 2.3 Other determinants of migration intention Literature on broad migration intention suggests that migration intention was influenced by macro-economic factors (e.g. employment rates, GDP per capita, and income differentials between regions) and household and individual factors (e.g. remittance, education, community

9 attachment, social network, life satisfaction, and experience of migration) (Otrachshenko and Popova 2014). A study in Turkey (Van Dalen et al. 2005) noted that remittances sent back by migrants had a positive effect on the intention of the household members left behind to emigrate. A high level of educational attainment significantly increased migration intention among rural youths in the Netherlands (Thissen et al. 2010). Social networks in the cities of Germany significantly facilitated migration intention (Kaplan et al. 2016). Other researchers (e.g., Thissen et al. 2010, Ivlevs and King 2012) claimed that the young generation whose parents had migration experience weremore prone to migrate than their counterparts whose parents did not have ant migratory experience. Migration intention was halted by strong place attachment

• f people (Oh 2003, Kaplan et al. 2016).

Perceived future prosperity in target destinations also influenced people’s tendency to migrate. Specifically, perceived job opportunities and higher incomes (Bjarnason and Thorlindsson 2006, He et al. 2016), better working conditions (Gouda et al. 2015), more opportunities for access to education (Dako-Gyeke 2016), and better-off lifestyle (Gouda et al. 2015, He et al. 2016) in target destinations stimulated people’s intention to migrate. On the contrary, migration intentions would decrease if people were satisfied with employment opportunities in their local communities (Thissen et al. 2010), or if they recognised various risks involved in migration processes including accidents and smugglers (Wissink et al. 2013) and discrimination in the destination (Becerra 2012). These factors identified in general migration research provide useful insights into research into climate change related migration intention. Existing literature has also paid attention to the effects of social, economic and demographic factors on migration intention. In their study in Funafuti (Tuvalu), Mortreux and Barnett (2009) showed migration intention is not directly driven by perception of climate impacts but driven

10 by people’s pursuit of lifestyle. Moreover, these islanders’ migration intention was inhibited by people’s religion, social attachment to local community and some demographic and economic characteristics such as young age and low income. Kuruppu and Liverman (2011) noted that most residents in Kiribati (an island nation in the central Pacific Ocean) had no intention to migrate even though they had experienced increases in extreme temperatures and sea-level rise. A primary reason was that the islanders had strong cultural affinity towards their

• land. In the forest-savannah transition zone of Ghana, a study by Abu et al. (2014) showed that

migration intentions of residents were not influenced directly by climate stressors they experienced but driven by social and demographic factors (e.g. young ages, a small household size, and having migrant members in a family). This study, using new data, examines ‘goal’ and ‘implementation’ intentions of migration from rural households’ perspective and focuses on two sets of factors: household’s ‘experience’ of climate extremes, and ‘political participation’. On the basis of the above discussion we postulate two hypotheses: H1: Experience of climate extremes is positively associated with both ‘goal intentions’ (H1-1) and ‘implementation intentions’—choice of target destination, especially long-distance migration and rural-to-urban migration (H1-2); and H2: Political participation at household level is positively associated with both goal intentions (H2-1) and implementation intentions—choice of target destination, especially long- distance migration and rural-to-urban migration (H2-2).

11 3．Methodology 3.1 Research framework: Experience of climate extremes, political participation and migration intention at the household level 3.1.1 Goal intention and implementation intention Goal intention was considered as ‘the most immediate and important predictor of attainment’ in psychological theories of goal pursuit (Gollwitzer and Sheeran 2006:71). Goal intention refers to particular outcomes somebody desires and the strength of those desires (Sheeran et al. 2005). Goal intention has a significant association with future behaviour, but it does not guarantee goal achievement. The gap between goal intention and actual behaviour could be bridged by implementation intention that takes specific actions to facilitate goal intention (Gollwitzer and Sheeran 2006). A measurement of implementation intention, as posited by Sheeran and Orbell (2000), is to ask people (e.g. patients) about when, where and how they would take an action (e.g. making an appointment with a clinic) to achieve the goal (e.g. a medical examination). Implementation intentions significantly increase the likelihood of the goal realisation through enhancing ‘people’s ability to initiate, maintain, disengage from, and undertake further goal striving’ (Gollwitzer and Sheeran 2006:82). However, implementation intention per se is insufficient to ensure goal achievement, rather, it can benefit goal achievement only when its specific plans are underpinned by strong goal intentions (Sheeran et al. 2005). This means both goal intention and implementation intention should be jointly examined when applying these concepts to climate–migration research. Based on the socio-psychological concepts of ‘goal intention’ and ‘implementation intention’, migration intention should not be simply treated as a binary choice in nature between showing intention to move and showing no intention to move. Goal intention of migration should go

12 beyond the binary choice to capture the strength of the goal of migration, which is better treated as a sliding scale along the continuum from the weakest end to the strongest end. Further, identifying implementation intention of migration is to understand people’s specific plans of when, where, who and how to initiate actual migration behaviour. 3.1.2 Multi-staged decision-making process of climate adaptation and migration Frameworks that conceptualise the relationship between climate change and adaptation have acknowledged that decision-making process of adaptation and migration, except for that under a scenario of sudden disasters, is a multi-staged one being influenced by environmental, social, economic and political factors (McLeman and Smit 2006, Perch-Nielsen et al. 2008, Black et

• al. 2011). The process involves a stage of actual experience of climate change and extremes

before subsequent stages of adaption intention and ultimately actual adaption (including migration) (Sinden and King 1990, Vedwan and Rhoades 2001, Thomas et al. 2007, Tan et al. 2015). In the Model of Private Proactive Adaptation to Climate Change (MPPACC), Grothmann and Patt (2005) explicitly pointed out that adaption intention is the immediate antecedent to actual adaptation. 3.1.3 Conceptual framework of the relationship between experience of climate extremes and migration intention Incorporating goal pursuit theories with the conceptual frameworks of multi-staged climate– adaptation nexus (Black et al. 2011) and considering the essential role of adaptation ‘intention’ in the multi-staged nexus (Grothmann and Patt 2005), this study constructs a framework to

13 increase our understanding of the climate–adaptation nexus. This framework unravels the mechanisms for how climate extremes influence livelihood sources of rural households and subsequently how the experience of climate impacts influences the household’s migration intention.

• Fig. 1 Conceptual framework: Experience of climate extremes, political participation and

migration intentions of rural households As outlined in Fig. 1, this framework seeks to investigate why households develop different migration intentions even under very similar environmental and socio-political conditions in a rural community. This framework considers people’s experience of climate impact on access to natural resources (especially water and farmland) and on agricultural production to be the most evident and serious experienced impact observed in rural areas because farmers’ livelihood is land-based and thus heavily relies on agricultural products. Migration intention is

Actual response Intended response Migration intention Goal intention: strength of migration intention Implementation intention Destination: intra-provincial VS. inter- provincial; rural VS. urban Experience

• f impacts
• f climate

extremes:

• n agriculture,

water and land...

Climate extremes

droughts, sandstorms … Actual migration

Other factors: perceived outcomes of migration, experience of migration

Household adaptive capacity

Political participation: position in power structure, participation in decision-making Socio-economic and demographic factors: income, housing, social networks, education, knowledge, information channels, demographic characteristics (e.g. age, gender, health) Current adaptive means: Adaptive means adopted by households, government support in face of climate risks

14 mediated by farmers’ experience of climate extremes, measured as the severity of previous and current climate impacts on their livelihood resources (e.g. water and land) and on agricultural

• production. The framework also shows that household political participation, especially a

household’s position in the local power structure (e.g. being a CCP member or government cadre) and participation in decision-making process, could have different impacts on people’s migration intention. Other important determinants of migration intentions, as discussed in literature review, including socio-economic factors, demographic characteristics, existing adaptive means, migratory history and perceived outcomes of migration are also included in this framework. Importantly, this framework goes beyond considering migration intention as a dichotomous dependent variable, and seeks to uncover significant forces driving complex migration intentions— both goal and target destinations of intended migration. 3.2. Research setting Minqin county (with a land area of 159.1 km2) is located at the lower reaches of the Shiyang River and sandwiched by Badain Jaran Desert and Tengger Desert (Fig. 2). This region is extremely dry, with annual rainfall being less than 150mm across the majority of Minqin county and less than 50mm in the lake area, with evapo-transpiration as much as 2,000– 2,600mm per annum (NDRC and MWR 2007). The water resource of Minqin was around per capita 500m3 in 2006, falling into the category of ‘extreme water scarcity’ (per capita 500m3 per year) as defined by the United Nations (2014). Water scarcity in this area has three major

• causes. Firstly, arid climate and decreased precipitation over past centuries has dramatically

exacerbated land erosion and salinisation, and an increase in the aridity, and consequently constrained water supply in the entire Shiyang River Basin (SRB) (Zhang et al. 2008). Secondly, rapid population growth and industrialisation in the upper and middle reaches have

15 reduced the river water discharge to the downstream of the SRB over the same time period. Finally, water scarcity was intensified by poor water management (Shi 2000) and poorly regulated human activities including over-grazing and over-cropping (Ma et al. 2007). Fig 2. Location of Minqin county in northwest China, and surveyed townships and villages Farmers have exploited groundwater for agricultural production to cope with drought and shortage of surface water since the late 1950s. Excessive extraction of groundwater has caused a severe groundwater budget deficit (Zhang et al. 2004) and consequently has led to natural vegetation degradation and soil salinisation, reinforcing the process of land desertification in the oasis (Lee and Zhang 2005). As a result, the Minqin area has become one of the four major dust storms across China (Kang et al. 2008). Persistent drought and environmental deterioration has caused 26,500 farmers to abandon farmland and their homes in the area surrounding Lake Qingtu to resettle elsewhere over the 10-year period to 2007 (NDRC and MWR 2007). Since 2002, the Chinese central government has repeatedly ranked the environmental issues of Minqin oasis (where Minqin county encompasses) among the most crucial ecological challenges in the country and the top policy concern of the provincial

16 government of Gansu (CMA 2013). To alleviate water scarcity in the SRB and prevent the Minqin oasis from further desertification, the Chinese government at all levels (central, provincial, county, and township) has developed a spectrum of policies for this region. Various programs involving water and land uses, ecosystem protection, economic development, development of renewable energy, and human resettlement have been encapsulated in the Governing Plan for Focal Issues in the Shiyang River Basin (NDRC and MWR 2007) and implemented since 2007. According to 2010 China Census, a majority (86.6%) of the total population (274,349) whose household status (hukou) was registered in Minqin county are dependent on agricultural activities for a living. Minqin has been a main source of migrant workers in both numerical and relative terms in Gansu province. In 2010, over 37,000 persons (or 13.5% of its total population) migrated out of Minqin and resided in places beyond the county boundary for six months or more. Due to out-migration, the total resident population (241,251) declined dramatically by 14.3% on the 2000 level (281,450) during the first decade of the 21st century (Minqin Statistical Bureau 2011). Old people aged 65 years or over accounted for 9.6% of its total resident population, much higher than the national average (8.9%) in 2010. 3.3 Data collection This study collected primary data from 436 households in 15 villages within five marginal townships located at the terminus of the Shiyang River in Minqin in 2012 (Fig. 2). The total sample size was more than the statistically valid sample size (N=384), accepting a margin of error of 5%, with a 95% level of confidence, with a total resident population of 241,251 in Minqin in 2010. The townships were selected on the basis of the following criteria: high

17 frequency of climate extremes (e.g. drought, sandstorm) over a 5-year period to 2012; their biophysical vulnerability to climate variability and climate-related environmental problems, especially water scarcity, desertification and soil salinisation; and the scale of migration of both government-led resettlement and spontaneous out-migration, which they experienced over the 2008–2012 period. The five townships selected, namely Hongshaliang, Xiqu, Donghu, Shoucheng, and Shuangcike, have experienced the most frequent climate extremes and had the most vulnerable ecosystems, according to our field observations, in-depth interviews with

• fficials from the Minqin county government and leading researchers in the field of natural

resources and human geography at Lanzhou University. These townships have been the largest sources of migrants in Minqin, accounting for one third of the total number of out-migrants from the county according to the population census in 2010. The survey used two levels of sampling units to represent villages and rural households. First, the total sample size (N=436 households) was proportionately distributed to each of the five selected townships in terms of population size in August–September 2012. The survey then used a Probability Proportionate to Size (PPS) sampling method to select three villages within each township. As a result, 15 villages were selected as the survey locations. The sample size for each township was proportionately distributed to the three selected villages, based on the population size in each village. A structured questionnaire survey was conducted through face- to-face interviews. The household head responded to the questionnaire; where unavailable, their spouse or the household member who best understood the household’s situation answered the questions.

18

• 4. Analysis: A modelling approach to assess the relationship between experience of

climate extremes, political participation and migration intentions Based on the conceptual framework constructed in Fig. 1 which draws on the climate risk and migration intention framework, a two-step regression procedure was adopted to analyse the questionnaire survey data. As per Fig. 1, a core focus for understanding the questionnaire survey data is in relation to impact and response, which in the case of this study is climate impact and household response, specifically migration intention. The two-step regression procedure was used to address the unobserved variables such as contextual dimensions as well as household socio-economic and demographic characteristics (Fig. 1) that influence, directly and indirectly, both the spheres of climate impacts and migration intentions. In particular, the regression procedure was used to address some of the problems of sample-induced endogeneity, that is, households that intend to relocate would be more likely to claim adverse climate impacts. For the purposes of carrying out the two-step regression procedure, it was necessary to create and distinguish between dependent variables and independent variables, as explained below. 4.1 Dependent variables 4.1.1 Impacts of climate extremes on livelihoods Households’ experience of climate impact on rural livelihoods, is considered as a dependant variable when we aim to investigate how climate extremes influence households’ experience

• f climate extremes. The questionnaire asked, first: ‘to what extent do you think climate

extremes (e.g., drought, sandstorm and hail) in 2011 affected any aspect of your household’s livelihood in 2011? And second, ‘what are the main aspects affected by climate extremes?’ Answers to the first question were recoded as 0 (‘never affected’) and 1 (‘much affected’). The

19 survey data shows that perceived climate extremes imposed large effects on four aspects of rural households’ livelihood resources: water quantity (decreased), water quality (decreased), agricultural production (decreased), and land loss (i.e. less land available for farming as farmland is abandoned due to drought and wind or soil erosion). 4.1.2 Migration intention Migration intention, in this study, is defined as an intention of the entire household to move

• ut of the original township for six months or longer. The time frame for intended migration

was set as the next two years (2013–2014) from the time point of our survey (September 2012). This study investigates three sets of migration intention: (1) goal intention; (2) implementation intention in terms of distance of migration: intra-provincial vs. inter-provincial movement; and (3) implementation intention in terms of rural or urban settings: intending to stay in countryside, move to township or county centres, or move to medium- and large-scaled cities. The definition and summary statistics of the three sets of dependant variables are presented in Table 1. Table 1. Definitions, means, and standard deviations (SD) of independent variables

Variables Definitions Impact Response Mean SD Migration intentions goal intention 1 if a household had at least one member who intended to migrate in next two years (n=161— 36.9%); 0 otherwise (n=275—63.1%) x 0.37 0.48 intended destination_1 0 if the whole household did not have intention to move in next two years (n=240—55.1%); 1 if the whole household intended to take intra- provincial migration, moving to other places within Gansu in next two years (n=132—30.3%); 2 if the whole household intended to take inter- provincial migration, moving to other provinces beyond Gansu in next two years (n=64—14.5%). x 0.60 0.73

20

intended destination_2 1 if the whole household intended to settle in countryside in the future (n=223—56.5%); 2 if the whole household intended to settle in a township seat or county seat in the future (n=80— 20.3%); 3 if the whole household intended to settle in a medium or large city in the future (n=92—23.3%). x 1.67 0.83 INDEPENDENT VARIABLES Climate impact frequency of climate extremes perceived frequency of climate extremes: [0, 10] = [very rare, very frequent] x 8.40 1.62 probability of impact on agricultural production predicted probability [0, 1], calculated from the first stage model x 0.57 0.23 probability of impact on water quality predicted probability [0, 1], calculated from the first stage model x 0.60 0.16 probability of impact on water quantity predicted probability [0, 1], calculated from the first stage model x 0.62 0.19 probability of impact on land loss predicted probability [0, 1], calculated from the first stage model x 0.44 0.17 Political participation Access to power party member 1 if the household has any member of the Communist Party of China (CPC); 0=otherwise x x 0.26 0.44 government official 1 if a household has any member or relative working in the government; 0 otherwise x x 0.28 0.45 Participation in decision-making involvement willingness willingness of the household to participate in policy-making process: [0, 10]=[very unwilling, very willing] x x 6.47 2.59 involvement satisfaction satisfaction level of participating in policy-making process: [0, 10]=[very unsatisfied, very satisfied] x x 4.53 2.42 Economic factors income per capita annual household income per capita ('000 yuan) x x 7.25 5.11 housing satisfaction level of satisfaction with housing conditions: [0, 10]=[very unsatisfied, very satisfied] x x 4.84 2.60 Social factors connection with relatives 1 if a household has any relatives or friends who live beyond Gansu province; 0 otherwise x x 0.43 0.50 relationship with cadres relationship between the household and local cadres: [0, 10]=[very bad, very good] x x 7.12 2.26 information channel number of information channels through which the household receives information on climate change and its impacts: [0, 9] x x 4.14 1.41 knowledge on climate extent of information on weather extremes that a family received: [0, 10]=[very little, very much] x x 5.04 3.32 Demographic factors male proportion proportion of males against the total number of the household members: [0, 1] x x 0.50 0.15

21

disability proportion proportion of those disabled or with chronic medical conditions against total number of household members: [0,1]. Disability includes any intellectual problem, physical disability, mental problem, and psychological disorder. x x 0.25 0.30 household size number of the total household members x x 4.51 1.41 working age proportion proportion of the number of members of working age (15-65 years) against total number of household members: [0, 1] x x 0.80 0.22 schooling average number of years of education of household members x x 6.83 2.65 Other factors migration history 1 if a household has any member who migrated in 2008–12; 0 otherwise x x 0.44 0.50 migration role migration is perceived as an effective way to deal with adverse climate impact: [0, 10]=[disagree, agree] x x 5.32 2.92 private adaptive means types of adaptation means adopted by families (number): [0, 10] x x 6.32 1.52 public adaptive means governmental support for adaptation: [0, 10]=[very little, very much] x x 2.99 2.72

The questionnaire asked a direct question: ‘to what extent does your whole household intend to migrate in the next two years (2013–14) because of experienced impact of climate extremes in the previous year (2011)?’ Answers to the question, which represent the strength of migration intention, were coded according to 11 levels on a Likert scale ranging from 0 (‘no intention at all’) to 10 (‘very strong intention’). The survey data shows that responses to the question were essentially distributed at both ends of the spectrum. Therefore, responses were recoded as 1 ‘planning to migrate’ (for responses coded 5–10) and 0 ‘no migration intention’ (for responses coded 0–4). Accordingly, the households in the sample were categorised into two groups: one group having an intention to migrate, and the other having no intention. It is noted again that migration intention should not be treated as a binary choice, but rather as existing on a sliding scale along a continuum of options. Therefore recoding migration intention as a binary variable is quite a strong assumption, but reasonable as a first

• approximation. Treating migration intention as a binary variable, and accordingly deploying

binary logistic or logit regression models, which are similar to the Probit model used in the present paper (e.g., Yang 2000; De Jong 2000; Li et al. 2014; Bednaříková et al. 2016), has

22 been common practice in empirical studies examining the determinants of migration intentions. The questionnaire further asked the respondents to indicate their intended destinations of

• migration. The answers were grouped into ‘inter-provincial’ or ‘intra-provincial’ according to

the distance between the study area and the reported destinations, and into ‘countryside’, ‘township or county seat’ or ‘city’ according to the rural/urban settings of the reported destinations. 4.2 Independent variables This study is particularly interested in two sets of explanatory variables: (1) experience of climate extremes: perceived climate extremes and their impacts on livelihood resources (land and water) and agricultural production; and (2) household’s political participation. The questionnaire asked people about their perceived frequency of climate extremes (e.g. droughts, dry and hot winds, sandstorms, and hails) in 2011 compared to their perceived frequency of such extremes in the previous 5 years. Their answers were coded to 11 levels on a Likert scale from 0 (‘very rare’) to 10 (‘very frequent’). The experience of climate impacts is used as exploratory variable to investigate how these experience influence migration intentions. The mean scores for probability of decreased water quantity, decreased water quality, decreased agricultural production and land loss are 0.62, 0.60, 0.57, and 0.44, respectively. It is worth noting that these four aspects of perceived severity of livelihood resources to be impacted by climate extremes could also be mediated by other factors, according to the in-depth interviews with village leaders and some respondents. For example, loss of arable land could be a result

• f salinisation, poor farming practice and land acquisition by the government for infrastructure
• projects. Nonetheless, this study focuses on the respondents’ experience of the livelihood

impacts of climate extremes. In our analysis, to predict each household’s migration intentions,

23 the probabilities of these four aspects are considered as dependent variables for the impact

• utcome, whilst the predicted probability of each of the four aspects of impact is considered as

an independent variable for the subsequent responses—three migration intention outcomes. Through a two-stage regression procedure developed by Tan et al. (2015), the specific mechanisms by which climate extremes influence, indirectly, household’s intention to migrate within next two years could be analysed. Informed by adaptation studies focusing on fair access to power and equal participation in decision-making (Thomas and Twyman 2005, Paavola and Adger 2006) and Chinese studies

• n political participation of rural residents (Nee and Su 1998, Bian 2002), political

participation in this study are therefore measured from two aspects: first, position in the local power structure (being a member of the CCP and/or a government official); and second, the level of participation in the decision-making process of climate (environmental) adaption and

• ther matters of local communities. For example, we asked respondents: ‘to what extent were

you satisfied with your participation in any policy-making process in your local community?’ Their responses were also coded to 11 levels on a Likert scale ranging from 0 (‘very unsatisfied’) to 10 (‘very satisfied’). The sample had a mean score of 4.53, suggesting that their satisfaction level was also low. This study measures economic conditions of rural households by especially considering the perceived levels of people’s satisfaction with their housing conditions and per capita household income, as informed by the literature that has identified some significant social and economic determinants of migration intention (e.g., Van Dalen et al. 2005, Thissen et al. 2010, Ivlevs and King 2012, Anniste and Tammaru 2014). Housing property is the most important asset of farmers in rural China (McKinley and Wang 1992). Respondents were asked to rate the level

24

• f the satisfaction with their housing conditions according to 11 levels on a Likert scale ranging

from 0 (‘very unsatisfied’) to 10 (‘very satisfied’). The mean score of sample was 4.84, again indicating a low level of satisfaction with their current living conditions. Social factors are measured by social capital and the average years of schooling of all members in the household. Social capital is usually termed as social networks that link this group of people to the other groups and organisations, and through which information and resources are transferred (Putnam 2000). This study measures two aspects of social capital: household’s social network (e.g. connection with relatives living within (or beyond) the same province, and relationship between the household and local cadres) and information received by the household—number of information channels and the extent of information on climate extremes. A few household demographic characteristics (e.g. age structure, gender composition, and the household size) significantly influenced migration intention induced by climate change in other countries (Mortreux and Barnett 2009; Abu et al. 2014). Such demographic factors are included in the analysis as control variables. Importantly, disability of household members has not been addressed in research on migration intention, and thus it is included in this study. Respondents were asked to give a self-report on any disabled condition (e.g. intellectual problems, physical disability, mental problems, and psychological disorder) of the household members. The proportion of the disabled persons against the total number of the household members in the sample was 0.25, indicating a high proportion of the sample had various disabilities. Further, existing studies (Thissen et al. 2010, Wissink et al. 2013) found that migration history and perceived migration outcomes played an important role in shaping migration intention. These two factors are therefore included as another set of control variables. We asked the respondents: ‘to what extent do you agree with the statement that migration is an effective way to cope with

25 adverse climate impacts?’ Their responses were also coded to 11 levels on a Likert scale ranging from 0 (‘strongly disagree’) to 10 (‘strongly agree’). The sample had a mean score of 5.32, indicating that the level of perceived role of migration as one of climate adaptation means is just above the average. Current adaptive means also shape the households’ adaptive capacity. The survey collected the information about current private adaptive means adopted by rural households and public adaptive means initiated and supported by local government. Governments’ adaptive means include the provision of living subsidies; access to small bank loans; provision of high-yielding seeds of crops and fruits and species of livestock; and on-site training in agricultural techniques and skills. Refer to Table 1 for a summary of the definition, means and standard deviations of independent variables. 4.3 Implementing the two-stage regression models A two-stage regression procedure (Tan et al. 2015) was adopted to investigate, firstly, how climate extremes, households’ political participation and other factors in the study area in 2011 influenced the four main aspects of rural households’ livelihood risks, and secondly, their subsequent intentions of migration in 2013-14 as a response to those experience of climate impacts. In Stage 1, a Multivariate Probit (MProbit) model (Greene 2008:826-831) is used to examine how their perceived frequency of climate extremes, political participation and other factors in 2011 affect the perceived levels of climate impact on four key domains (water quantity, water quality, agricultural production, and land loss) of households’ livelihood resources in that year, as each domain of livelihood is measured as a dichotomous variable. An important purpose of the first stage model is to obtain the predicted probability (or severity) of each impact, and the MProbit model best meets this purpose. The four probabilities of climate impact are predicated

26 probabilities from Stage 1 regression rather than subjective belief, and thus should not be

• endogenous. Using the predicted probabilities is similar to the technique in two-stage least

squares (2SLS) to address endogeneity. These predicted probabilities of climate impact are used as the independent variables in Stage 2 (migration intention) models. In Stage 2, this study uses a Probit model to analyse how the perceived four impacts (i.e. predicted probabilities from the MProbit model in Phase 1) and political participation factors influence households’ goal intention to migrate or not. Greene (2008:772-775) provided the detailed econometric specification for Probit models. Further, two Multinomial Logit (MLogit) models are employed to examine how the four domains of specific climate impacts and other factors influence two types of implementation intentions of migration in terms of where rural households intended to move (or live) in the next two years (2013–2014). Their choices of target destination are distinguished by: firstly, distance (between no intention to migrate, intra- provincial migration, and inter-provincial migration); and secondly, settlement setting (countryside, township/county centres, and medium/large cities). MLogit models are frequently used in situations of multiple choice of dependent variables. Please see Greene (2008:843-845) for technical details.

• 5. Results

5.1 Goal intentions Experience of climate impacts imposes significant and mixed effects on households’ intention to migrate or not (Table 2). Those households which experienced land loss due to climate extremes dramatically increased the probability of migration intention (by 0.52, at the 5% significance level), compared to those that did not experience such impact (Table 3). The

27 probability of intention to migrate for the group that experienced a decline in water supply due to climate extremes would decrease significantly (by 0.14, at the 1% significance level), compared to those who did not suffer that impact. The findings suggest that different aspects

• f climate impact experienced by rural households could differentiate people’s intention to

migrate or not. Experience of climate impacts thus does not necessarily increase migration intention, which partly agrees with our hypothesis H1-1. Table 2. Probit regression results: ‘Goal intention’ model

Variables Coef. Climate impact probability of impact on water quantity

–13.091***

probability of impact on land loss 7.613** Political participation government officials 0.615*** involvement satisfaction 0.297** Economic factors income per capita

–0.097*

housing satisfaction

–0.172***

Social factors information channel 0.403** schooling

–0.259**

Demographic factors male proportion 2.995*** disability proportion 1.651*** working age proportion

–2.656**

Other controls migration history 2.120*** migration role 0.309*** private adaptive means 0.448*** constant 0.658 Observations 428 Wald chi2 71.740 Count R2 0.729

28 Table 3. Predicted marginal effects of factors influencing migration intention of rural households

Marginal effects (change in probability of household’s migration intention) Climate impact climate extremes have no impact on water quantity have an impact

–0.137

climate extremes have no impact on land loss have an impact 0.515 Political participation any family member or relative working in the government: no yes 0.232 satisfaction level of participating in any policy-making process: mean mean plus 1 unit 0.089 Economic factors annual household income per capita: mean mean plus 1 unit

–0.029

level of satisfaction with housing conditions: mean mean plus 1 unit

–0.052

Social factors number of information channels: mean mean plus 1 unit 0.121 schooling year: mean mean plus 1 unit

–0.078

Demographic factors proportion of male members in the household: mean mean plus 0.1 unit 0.115 proportion of disabled members in the household: mean mean plus 0.1 unit 0.062 proportion of working age adults in the household: mean mean plus 0.1 unit

–0.091

Other controls migration experience: do not have have 0.690 extent to which migration was considered as an effective way to cope with climate impacts: mean mean plus 1 unit 0.093 number of public adaptive means adopted by the household: mean mean plus 1 unit 0.134 Note: In the baseline model, the dummy variables are set as 0; and the continuous variables are set at their respective means (which is the usual practice for continuous variables in generating predicted probabilities). The marginal effect measures the change in probability for every one unit change in corresponding independent variables (i.e. change from 0 to 1 for a dummy variable, or increase by a one unit above the mean for most continuous variables except for demographic factors).

As an important finding, political participation has significant and positive association with goal intention of migration. Those households which had family members or relatives to be employed as government officials, or were more satisfied with their participation in policy- making process exhibit greater intention to migrate (at the 1–5% significance level) (Table 2). Particularly, households in which any member or relative worked in government bodies had a

29 higher probability (by 0.23) to migrate out of their current villages than those without this kinship (Table 3). These findings support our hypothesis H1-2. Among the economic and social factors, a higher level of satisfaction with housing conditions and a higher level of educational attainment are significantly, but negatively, correlated with goal intention of migration. The probability of intending to migrate would decrease by 0.05 for an increase of one unit above the mean (at the 1% significance level) for households expressed a higher level of satisfaction with housing (Table 3). On the contrary, households having more channels to access information about climate change and its impacts presented an increased probability of migration intention, by 0.12 for an increase of one unit (i.e. one additional channel) above the mean, assuming other factors remain unchanged. Demographic characteristics that are positively associated with goal intention of migration include: a higher proportion of male members in the household, and a higher proportion of disabled members within the household. By contrast, the probability of migration intention would decrease (by 0.09 for an increase of 0.1 unit above the mean) for households characterised by a higher proportion of working age members in the household. Households showed stronger intention to take migration if their household member(s) had migration experience and/or if they perceived migration as an effective way to deal with climate impacts. The probability of intention to migrate would increase substantially, by 0.69 at the 1% significance level, for the households having prior migration experience. These findings are in line with existing literature.

30 5.2 Implementation intentions: Choice between intra-provincial and inter-provincial destinations As an important finding, two aspects of experience of climate impacts influenced households’ migration intention in terms of target destination. First, if households perceived that deterioration in water quality was produced by climate extremes, they were more likely to undertake either inter-provincial migration or to stay, than to move within the province (Table 4). Second, for those households that experienced greater loss of agricultural production owing to climate impact, the probability of intending to move inter-provincially would decline dramatically (by 0.32) (Table 5). These findings show that experience of climate extremes does not always increase intention to pursue long distance migration. The findings partly agree with

• ur hypothesis H2-1.

Political participation is significantly and positively associated with intention of long distance

• migration. If any of household members was employed as a government official, there was a

greater propensity for such households to have an intention to undertake intra-provincial migration than no intention to move out at all, compared to their counterparts who did not have such kinship. For the households that were more satisfied with participation in decision-making process, they were more likely to have an intention to take inter-provincial migration rather than intro-provincial migration. High degree of participation in decision-making does not encourage people to stay, which reflects the fact that current out-migration from rural China is not driven by political pursuit. High degree of political participation might provide farmers with information about adaptation policies, migration programs and destination conditions, which in turn enhance their confidence to undertake long distance migration. These findings agree with our hypothesis H2-2.

31 Table 4. Multinomial logit regression results: Target destination model (1): intra-provincial

• VS. inter-provincial migration

Variables No intention VS. intra-provincial migration No intention VS. Inter-provincial migration intention Inter-provincial migration VS. Intra-provincial migration Coef. Coef. Coef. Experience of climate impacts probability of impact on agricultural production

–1.405

3.088

–4.493*

probability of impact on water quality 16.940** 6.736 10.203* Political participation government officials

–0.619* –0.942**

0.323 involvement satisfaction 0.335

–0.149

0.484* Economic factors income per capita

–0.376*** –0.162 –0.213*

Social factors connection with relatives

–1.144* –1.069 –0.075

relationship with cadres

–0.625** –0.053 –0.572**

Demographic factors male proportion 1.799

–2.064

3.863* hh_size

–0.255** –0.235* –0.020

working age proportion 5.037* 2.893 2.143 Other controls private adaptive means 0.112

–0.330

0.441* public adaptive means 0.296** 0.188 0.108 constant

–7.053*** –3.318 –3.735

Observations 428 Wald chi2 115.650 Count R2 0.650 * p<.1; ** p<.05; *** p<.01

32 Table 5. Predicted marginal effects of factors influencing choice of target destination: intra- provincial VS. inter-provincial destination

No intention Intention to migrate within Gansu province Intention to migrate beyond Gansu province Climate impact climate extremes have no impact on agricultural production have an impact 0.137 0.180

–0.317

climate extremes have no impact on water quality have an impact 0.226

–0.302

0.076 Political participation any family member or relative working in the government: no yes

–0.172

0.101 0.071 satisfaction level of participating in any policy-making process: mean mean plus 1 unit 0.036

–0.065

0.029 Economic factors annual household income per capita: mean mean plus 1 unit

–0.057

0.058

–0.001

Social factors any relatives or friends who live beyond Gansu province: do not have have

–0.270

0.212 0.058 extent of relationship between the family and local cadres: mean mean plus 1 unit

–0.084

0.107

–0.023

Demographic factors proportion of male members in the household: mean mean plus 0.1 unit 0.018

–0.044

0.026 number of household members: mean mean plus 1 unit

–0.045

0.034 0.011 proportion of working age adults in the household: mean mean plus 0.1 unit 0.103

–0.092 –0.012

Other controls number of private adaptive means adopted by the household: mean mean plus 1 unit

–0.002 –0.034

0.036 degree of governmental assistance for adaptation: mean mean plus 1 unit 0.048

–0.043 –0.005

Note: In the baseline model, the dummy variables are set as 0; and the continuous variables are set at their respective means (which is the usual practice for continuous variables in generating predicted probabilities). The marginal effect measures the change in probability for every one unit change in corresponding independent variables (i.e. change from 0 to 1 for a dummy variable, or increase by a one unit above the mean for most continuous variables except for some demographic factors).

Households having higher economic and social status are more likely to have the intention to take intra-provincial migration than to stay or undertake inter-provincial migration. Households showed stronger intention to migrate within the province where they currently

33 resided than to stay or to migrate beyond the province of current residence, if they had higher per capita income, and/or maintained closer social connection with local cadres (Table 4). Concerning the demographic factors, large-sized households intended to migrate, either intra-

• r inter-provincially, than to remain in their current hometown (Table 4). While households

having a high proportion of working aged members were less likely to develop an intention of intra-provincial migration than to have an intention of staying. Households characterised by having a higher proportion of male members showed stronger intention of inter-provincial migration rather than intra-provincial migration. In addition, households having obtained more government support for adaptation expressed little intention to undertake intra-provincial migration compared to choose to stay in where they lived at present. 5.3 Implementation intentions: Destination choice between rural and urban settings Experience of climate impact on water does not always motivate farmers to plan to move out

• f countryside and resettle in urban areas. Households that experienced greater levels of climate

impact on water quality and quantity presented stronger intention to move to other rural communities, rather than township/county seats or medium/large cities, compared to their counterparts who perceived few constraints of water (Table 6). The probability of intending to move to other countryside would increase dramatically by 0.44 and 0.32 (at the 5% significance level) for the households which experienced deterioration of water quality and reduction in water supply, respectively (Table 7). Noted that households that experienced greater deterioration in water quality showed stronger intention of moving to medium- or large-sized cities rather than small urban areas (i.e. township or county seats), compared to those who did

34 not experience declines in water quality. These findings also partly support our hypothesis H2- 2. Looking into the political participation factors, Households reporting a higher degree of satisfaction with their participation in policy-making processes expressed a greater tendency to move to a medium or large city than to a township seat or small city. These findings show that

• ur hypothesis H2-2 is warranted.

Table 6. Multinomial logit regression results: Target destination model (2): rural VS. urban settings

Variables township/ county seat VS. countryside Medium/ large city

• VS. countryside

Medium/ large city

• VS. township/

county seat Coef. Coef. Coef. Experience of climate impacts probability of impact on water quality

–12.327**

1.101 13.428** probability of impact on water quantity

–8.115 –10.476** –2.362

Political participation involvement satisfaction

–0.238

0.238 0.476* Economic factors income per capita 0.119

–0.081 –0.200*

Social factors information channel 0.891** 0.548*

–0.343

Demographic factors male proportion 4.575** 3.887**

–0.688

disability proportion 2.076* 1.416

–0.660

schooling

–0.397* –0.141

0.256 Other controls private adaptive means 0.455** 0.438***

–0.017

public adaptive means

–0.283* –0.112

0.171 _cons 5.243*

–2.951 –8.193***

Observations 387 Wald chi2 114.22 Count R2 0.62 * p<.1; ** p<.05; *** p<.01

35 Table 7. Predicted marginal effects of factors influencing choice of target destination: rural

• VS. urban settings

Countryside Township/ county seats Medium/ large cities Experience of climate impacts climate extremes have no impact on water quality  have an impact 0.439 –0.578 0.139 climate extremes have no impact on water quantity  have an impact 0.321 0.098 –0.418 Political participation satisfaction level of participating in any policy-making process: mean mean plus 1 unit –0.004 –0.042 0.046 Economic factors annual household income per capita: mean mean plus 1 unit –0.002 0.019 –0.017 Social factors number of information channels: mean mean plus 1 unit –0.144 0.097 0.047 Demographic factors proportion of male members in the household: mean mean plus 0.1 unit –0.103 0.054 0.049 proportion of disabled members in the household: mean mean plus 0.1 unit –0.041 0.025 0.016 schooling year: mean mean plus 1 unit 0.053 –0.047 –0.005 Other controls number of private adaptive means adopted by the household: mean mean plus 1 unit –0.091 0.043 0.048 degree of governmental assistance for adaptation: mean mean plus 1 unit 0.039 –0.033 –0.005 Note: In the baseline model, the dummy variables are set as 0; and the continuous variables are set at their respective means (which is the usual practice for continuous variables in generating predicted probabilities). The marginal effect measures the change in probability for every one unit change in corresponding independent variables (i.e. change from 0 to 1 for a dummy variable, or increase by a one unit above the mean for most continuous variables except for some demographic factors).

Considering other factors, households presented stronger intention of moving to cities rather than county seats if they had higher income per capita (at a lesser significance level of 10%) (Table 6). Social factors have mixed effects on intention of choosing destination settings. Importantly, households with access to more information channels had greater intention to move out of rural areas and settle in urban areas. The estimated probability for such households to choose to settle in rural areas would decrease significantly (by 0.14 for every addition of one

36 information channel above the mean) (Table 7). Among the demographic factors, those households featured by high proportions of male and/or disabled members particularly intended to settle in township or county centres, or even medium- or large-scaled cities, rather than moving to other rural communities. Moreover, private adaptation means increased households’ intention to move to urban areas while public adaptation means decreased the intention.

• 6. Discussion

A key finding of this study is that farmers’ experience of climate impacts influences migration intentions differently. Some aspects of agriculture- or land-based livelihood affected by climate extremes can foster farmers’ intention to migrate (e.g. due to increased land loss), and nurture more ambitious migration intentions to move to other provinces or to settle in urban areas (e.g. due to deteriorated water quality). Some other aspects affected by climate extremes can hinder the intention of migration (due to decreasing water supply), or only inspire conservative intention to move within the same province as people lived at present (Gansu). Water shortage has been a ubiquitous state in the case study area since the 1950s. The challenge

• f decreased water supply has been addressed by implementing in-situ adaptation schemes,

strictly controlling the provision of irrigation water (including underground water), reducing cropping land (and consequently lowering demand for irrigation water), and planting drought- resistant crops (e.g. cotton, sunflower, fennel and medicinal herbs). Engaging in in-situ adaptation could be a primary reason why the households that reported declines in water supply due to climate extremes (especially drought) had little intention to migrate. Loss (or abandonment) of farmland is a function of rampant droughts and deficiency of irrigation water

• wing to the strict control by the government to regenerate the ecosystem services in the

37 extremely arid and ecologically vulnerable area. Clearly, arable land is the vital natural resource for farmers to produce food to feed themselves and generate household income. Unsurprisingly as a result of land loss, people’s intention to migrate increased significantly and

• substantially. Despite steady endeavours for local governments and rural households to apply

diverse adaptation methods to tackle water shortage in Minqin, no effective means has been adopted in this area to address water quality issues (interview with the Deputy Director of Water Bureau of the Minqin County Government, 28 August 2012). This is why those households that attributed deterioration of water quality to the effect of climate extremes were more prone to develop an ambitious migration intention (moving to other provinces or settling in urban areas) rather than a conservative migration intention reflected by other households that perceived little climate impact on water quality. Nevertheless, experience of decreased agricultural production caused by climate extremes inhibits farmers from developing such ambitious migration intention. This is because long-distance migration and rural-to-urban migration often requires more financial and social resources than short-distance and rural-to- rural migration. Reduction in agricultural production directly reduces the household income, and in turn, it undermines farmers’ affordability for long-distance and rural-to-urban migration and thus inhibits their planning to take such ambitious migration. Diverse experiences of climate impacts exhibit complicated influences on migration intentions. Tucker et al. (2010:27) claimed that ‘there were important differences not only in perception

• f climate risk more generally, but also the type of hazard experienced’. Our study agrees with

the argument and further identifies that migration intention is not only differentiated by the perceived frequency of various types of climate extremes, but also by various risk experiences

• f climate impacts. To inform effective migration strategies as well as other forms of adaptation

38 means, local governments should clearly understand how specific aspects of climate impact experienced by people differentiate their migration intentions. A high level of political participation of a household promotes migration intention. Two important aspects of household political participation—position in local power structure and the level of participation in policy-making processes—are associated, significantly and positively, with a strong goal intention, and also with more ambitious implementation intentions to migrate (i.e. moving inter-provincially or moving to cities). The finding is consistent with the general adaptation literature that participation in decision-making processes and can strengthen people’s adaptive capacity to climate change (e.g., Thomas and Twyman 2005) and consequently raise adaptation intention (Grothmann and Patt 2005). Interestingly, households that obtained more governmental support for adaptation showed little intention to migrate or to undertake long-distance and rural-to-urban migration. There are at least two possible reasons for this result. Firstly, some households’ resilience to climate impact have been significantly enhanced by government support, so farmers did not consider climate impact as a threat to their livelihood and thus found it unnecessary to move out of their villages. Secondly, other households in the community have relied heavily on governmental assistance, so they were reluctant to take their own responsibility for proactive adaptation (including migration). These findings suggest that local governments need to enhance people’s adaptive capacity by promoting equal access to adaptive means and equal participation in decision- making processes, and that local governments need to provide direct assistance to farmers with caution when the government confirms that the support does not undermine farmers’ proactive adaptation.

39 This study acknowledges that the relationship between experience of climate impacts and migration intention is embedded in more complex socio-economic context. Better-off economic conditions and a high level of educational attainment among household members reduce the probability for rural households to develop a strong goal intention to migrate or an ambitious implementation intention to migrate to distant localities or large cities. This result supports the findings of other researchers (e.g. Mortreux and Barnett 2009) that low income households are more prone to plan migration than wealthy ones. Our finding further asserts that high income can strengthen people’s resilience to climate risks and help avoid mobility (Smith et al. 2006). Literature shows educational attainment plays different roles in climate adaptation and migration. A higher level of schooling can increase in-situ adaptation to climate change in some areas (Deressa et al. 2009) but increase actual migration elsewhere (Henry et

• al. 2004). However, few studies have investigated the influence of education on migration

intention and our study fills this research gap. In a climate change hotspot and ecologically vulnerable area like the case study of this research, planned relocation schemes need to be carefully designed. After years of practicing ecological migration in western China, millions of rural residents have been relocated, or voluntarily out- migrated, from their original villages. Many of the remaining households, like those in our case study area, were not interested in planned relocation, due to either increased in-situ adaptive capacity or lack of socio-economic resources. It would need substantial resources for these people to plan to implement migration (or relocation), especially if the government cannot financially compensate for the loss of affected households produced by relocation. We recommend the local governments to plan cohesive development programs in which local socio-economic development and entitling people equal access to governmental support, local adaptation programs, and participation in decision-making are integral, aiming to facilitate

40 transferring some farmers’ intention to migrate and some others’ intention to stay into effective adaptation and risk management schemes.

• 7. Conclusion

Built on the ‘goal pursuit theory’, a two-stage framework for climate–adaptation nexus, and the conceptualised socio-cognitive attributes of adaptation intention, this study constructed an analytical framework to test two sets of hypotheses for how intended migration is mediated by experience-based climate risks and political participation of rural households in a climate change hotspot, Minqin county of western China. Robust evidence from the case study shows that migration intentions of farmers involve multiple interacting factors. This paper enhances the nuanced understanding of the non-linear linkages between climate extremes and adaptation intention by being the first paper to investigate how climate extremes, through shaping people’s experience of climate impacts, impact on both goal and implementation intentions of migration. Recognising the mixed effects of farmers’ experience of climate impacts and political participation, together with social, economic and demographic factors at the household level,

• n migration intentions, local climate adaptation and migration policies and programs need to

avoid viewing the population as a homogeneous group and to identify specific needs of different groups. Migration can act as an effective adaptation means and can also be regarded as a failure of in-situ adaptation, thus simply encouraging or constraining migration could lead to adverse consequences. Understanding people’s ‘goal’ and ‘implementation’ migration intentions facilitates policymakers and practitioners to build the resilience of vulnerable groups and their communities and empower them to make risk-informed decisions about proactive and planning activities, and contingency planning that affect where and how people migrate.

41 References Abu, M., Codjoe, S. N. A., Sward, J., 2014. Climate change and internal migration intentions in the forest-savannah transition zone of Ghana. Population and Environment, 35(4), 341-364. Adger, W. N., Nelson, D. R., 2010. Fair decision making in a new climate of risk. In: ed. Climate Change, Ethics and Human Security. 83-94. Akerlof, K., Maibach, E. W., Fitzgerald, D., Cedeno, A. Y., Neuman, A., 2013. Do people "personally experience" global warming, and if so how, and does it matter? Global Environmental Change, 23(1), 81-91. Anniste, K., Tammaru, T., 2014. Ethnic differences in integration levels and return migration intentions: A study of Estonian migrants in Finland. Demographic Research, 30, 377. Becerra, D., 2012. The Impact of anti‐immigration policies and perceived discrimination in the United States on migration intentions among Mexican adolescents. International Migration, 50(4), 20-32. Bednaříková, Z., Bavorová, M., & Ponkina, E. V. 2016. Migration motivation of agriculturally educated rural youth: The case of Russian Siberia. Journal of Rural Studies, 45, 99-111. doi: 10.1016/j.jrurstud.2016.03.006 Bian, Y., 2002. Chinese social stratification and social mobility. Annual Review of Sociology, 28, 91-116. Bjarnason, T., Thorlindsson, T., 2006. Should I stay or should I go? Migration expectations among youth in Icelandic fishing and farming communities. Journal of Rural Studies, 22(3), 290-300. Black, R., Adger, W. N., Arnell, N. W., Dercon, S., Geddes, A., Thomas, D., 2011. The effect of environmental change on human migration. Global Environmental Change, 21, S3-S11. Bouriaud, L., Marzano, M., Lexer, M., Nichiforel, L., Reyer, C., Temperli, C., Peltola, H. et al., 2015. Institutional factors and opportunities for adapting European forest management to climate change. Regional Environmental Change, 15(8),1595-1609. Böheim, R., Taylor, M. P., 2002. Tied down or room to move? Investigating the relationships between housing tenure, employment status and residential mobility in Britain. Scottish Journal of Political Economy, 49(4), 369-392.

42 China Meteorological Adminitration, (CMA), 2013. Zhengjiu Minqin (Securing Minqin against Water Scerecity). Beijing: The China Meteorological Aministraiton. Dako-Gyeke, M., 2016. Exploring the migration intentions of Ghanaian youth: A qualitative

• study. Journal of International Migration and Integration, 17(3), 723-744.

De Jong, G. F. (2000). Expectations, gender, and norms in migration decision-making. Population Studies, 54(3), 307-319. Deressa, T. T., Hassan, R. M., Ringler, C., Alemu, T., Yesuf, M., 2009. Determinants of farmers’ choice of adaptation methods to climate change in the Nile Basin of Ethiopia. Global Environmental Change, 19(2), 248-255. Dutra, L.X., Thébaud, O., Boschetti, F., Smith, A.D. and Dichmont, C.M., 2015. Key issues and drivers affecting coastal and marine resource decisions: Participatory management strategy evaluation to support adaptive management. Ocean and Coastal Management, 116, 382-395. Ezra, M., Kiros, G., 2001. Rural out-migration in the drought prone areas of Ethiopia: A multilevel analysis. International Migration Review, 35(3), 749-771. Feng, S., Krueger, A. B., Oppenheimer, M., 2010. Linkages among climate change, crop yields and Mexico-US cross-border migration. Proceedings of the National Academy of Sciences of the United States of America, 107(32), 14257-14262. Foresight, 2011. Migration and global environmental change: Future challenges and

• pportunities. Final project report. London UK: U.K. Government Office for Science

Science. Frank, E., Eakin, H., López-Carr, D., 2011. Social identity, perception and motivation in adaptation to climate risk in the coffee sector of Chiapas, Mexico. Global Environmental Change, 21(1), 66-76. Gollwitzer, P. M., Sheeran, P., 2006. Implementation intentions and goal achievement: A meta‐ analysis of effects and processes. Advances in experimental social psychology, 38, 69- 119. Gordon, I., Molho, I., 1995. Duration dependence in migration behaviour: Cumulative inertia versus stochastic change. Environment and Planning A, 27(12), 1961-1975. Gouda, P., Kitt, K., Evans, D. S., Goggin, D., McGrath, D., Last, J., Hennessy, M., Arnett, R., O’Flynn, S., Dunne, F., 2015. Push and stay factors affecting Irish medical student migration intentions. Irish Journal of Medical Science (1971-), 1-7. Greene, W., 2008. Econometric Analysis. New York: Pearson Education Inc.

43 Grothmann, T., Patt, A., 2005. Adaptive capacity and human cognition: The process of individual adaptation to climate change. Global Environmental Change, 15(3), 199-213. He, Z., Zhai, G., Asami, Y., Tsuchida, S., 2016. Migration intentions and their determinants: Comparison of college students in China and Japan. Asian and Pacific Migration Journal, 25(1), 62-84. Henry, S., Schoumaker, B., Beauchemin, C., 2004. The impact of rainfall on the first out- migration: A multi-level event-history analysis in Burkina Faso. Population and Environment, 25(5), 423-460. Ivlevs, A., King, R. M., 2012. Family migration capital and migration intentions. Journal of family and economic issues, 33(1), 118-129. Kang, S., Su, X., Tong, L., Zhang, J., Zhang, L., Davies, 2008. A warning from an ancient

• asis: Intensive human activities are leading to potential ecological and social catastrophe.

The International Journal of Sustainable Development & World Ecology, 15(5), 440-447. Kaplan, S., Grünwald, L., Hirte, G., 2016. The effect of social networks and norms on the inter-regional migration intentions of knowledge-workers: The case of Saxony, Germany. Cities, 55, 61-69. Kuruppu, N., Liverman, D., 2011. Mental preparation for climate adaptation: The role of cognition and culture in enhancing adaptive capacity of water management in Kiribati. Global Environmental Change, 21(2), 657-669. Lee, H. F., Zhang, D. D., 2005. Perceiving land-degrading activities from the lay perspective in northern China. Environmental Management, 36(5), 711-725. Li, Y., López-Carr, D., & Chen, W. 2014. Factors affecting migration intentions in ecological restoration areas and their implications for the sustainability of ecological migration policy in arid Northwest China. Sustainability (Switzerland), 6(12), 8639-8660. doi: 10.3390/su6128639 Ma, Y., Fan, S., Zhou, L., Dong, Z., Zhang, K., Feng, J., 2007. The temporal change of driving factors during the course of land desertification in arid region of north China: The case of Minqin County. Environmental Geology, 51(6), 999-1008. Massey, D. S., Axinn, W. G., Ghimire, D. J., 2010. Environmental change and out-migration: Evidence from Nepal. Population and Environment, 32(2), 109-136. McKinley, T., Wang, L. N., 1992. Housing and wealth in rural China. China Economic Review, 3(2), 195-211. McLeman, R., Smit, B., 2006. Migration as an adaptation to climate change. Climatic Change, 76(1-2), 31-53.

44 Morduch, J., Sicular, T., 2001. Risk and insurance in transition: Perspectives from Zouping County, China. In: A. Masahiko and Y. Hayami, ed. Community and Market in Economic

• Development. Oxford: Oxford University Press.

Mortreux, C., Barnett, J., 2009. Climate change, migration and adaptation in Funafuti, Tuvalu. Global Environmental Change, 19(1), 105-112. National Development and Reform Committee (NDRC), Ministry of Water Resources (MWR),

• 2007. Governing Plan for Focal Issues in the Shiyang River Basin. Beijing: NDRC and

MWR. Nee, V., Lian, P., 1994. Sleeping with the enemy: A dynamic model of declining political commitment in state socialism. Theory and Society, 23(2), 253-296. Nee, V., Su, S., 1998. Institutional foundations of robust economic performance: Public sector industrial growth in China. In: J. Henderson, ed. Industrial Transformation in Eastern Europe in The Light of The East Asian Experience. New York: St. Martin's, 167-187. O'Connor, R. E., Bord, R. J., Fisher, A., 1999. Risk perceptions, general environmental beliefs, and willingness to address climate change. Risk Analysis, 19(3), 461-471. O'Connor, R. E., Yarnal, B., Dow, K., Jocoy, C. L., Carbone, G. J., 2005. Feeling at risk matters: Water managers and the decision to use forecasts. Risk Analysis, 25(5), 1265- 1275. Oh, J.-H., 2003. Social bonds and the migration intentions of elderly urban residents: The mediating effect of residential satisfaction. Population Research and Policy Review, 22(2), 127-146. Otrachshenko, V., Popova, O., 2014. Life (dis) satisfaction and the intention to migrate: Evidence from Central and Eastern Europe. The Journal of Socio-Economics, 48, 40-49. Paavola, J., Adger, W. N., 2006. Fair adaptation to climate change. Ecological Economics, 56(4), 594-609. Perch-Nielsen, S. L., Bättig, M. B., Imboden, D., 2008. Exploring the link between climate change and migration. Climatic Change, 91(3-4), 375-393. Putnam, R. D., 2000. Bowling alone: The collapse and revival of American community. New York: Simon and Schuster. Sheeran, P., Harris, P. R., Epton, T., 2014. Does heightening risk appraisals change people's intentions and behavior? A meta-analysis of experimental studies. Psychological bulletin, 140(2), 511-543. Sheeran, P., Orbell, S., 2000. Using implementation intentions to increase attendance for cervical cancer screening. Health Psychology, 19(3), 283.

45 Shi, J., 2000. A case study of Minqin oasis in China. Water International, 25(3), 418-424. Siegrist, M., Gutscher, H., 2006. Flooding risks: A comparison of lay people's perceptions and expert's assessments in Switzerland. Risk Analysis, 26(4), 971-979. Sinden, J. A., King, D. A., 1990. Adoption of soil conservation measures in Manilla Shire, New South Wales. Review of Marketing and Agricultural Economics, 58(2/3), 179–192. Smith, V. K., Carbone, J. C., Pope, J. C., Hallstrom, D. G., Darden, M. E., 2006. Adjusting to natural disasters. Journal of Risk and Uncertainty, 33(1-2), 37-54. Tan, Y., Liu, X., Hugo, G., 2015. Exploring relationship between social inequality and adaptations to climate change: Evidence from urban household surveys in the Yangtze River delta, China. Population and Environment, 36(4), 400-428. Thissen, F., Fortuijn, J. D., Strijker, D., Haartsen, T., 2010. Migration intentions of rural youth in the Westhoek, Flanders, Belgium and the Veenkoloniën, The Netherlands. Journal of Rural Studies, 26(4), 428-436. Thomas, D. S., Twyman, C., 2005. Equity and justice in climate change adaptation amongst natural-resource-dependent societies. Global Environmental Change, 15(2), 115-124. Thomas, D. S., Twyman, C., Osbahr, H., Hewitson, B., 2007. Adaptation to climate change and variability: Farmer responses to intra-seasonal precipitation trends in South Africa. Climatic Change, 83(3), 301-322. Tucker, C. M., Eakin, H., Castellanos, E. J., 2010. Perceptions of risk and adaptation: Coffee producers, market shocks, and extreme weather in Central America and Mexico. Global Environmental Change, 20(1), 23-32. United Nations, (UN), 2014. International Decade for Action 'Water for Life' 2005-2015. http://www.un.org/waterforlifedecade/scarcity.shtml. Accessed June 2015. Van Dalen, H. P., Groenewold, G., Fokkema, T., 2005. The effect of remittances on emigration intentions in Egypt, Morocco, and Turkey. Population Studies, 59(3), 375-392. Van Deth, J.W., 2014. A conceptual map of political participation. Acta Politica, 49(3), 349- 367. Vedwan, N., Rhoades, R. E., 2001. Climate change in the Western Himalayas of India: A study

• f local perception and response. Climate Research, 19(2), 109-117.

Wissink, M., Düvell, F., van Eerdewijk, A., 2013. Dynamic migration intentions and the impact of socio-institutional environments: A transit migration hub in Turkey. Journal of Ethnic and Migration studies, 39(7), 1087-1105. Yang, X. (2000). Determinants of migration intentions in Hubei province, China: Individual versus family migration. Environment and Planning A, 32(5), 769-787.

46 Yazdanpanah, M., Hayati, D., Hochrainer-Stigler, S., Zamani, G. H., 2014. Understanding farmers' intention and behavior regarding water conservation in the Middle-East and North Africa: A case study in Iran. Journal of Environmental Management, 135, 63-72. Zhang, K., Qu, J., Liu, Q., 2004. Environmental degradation in the Minqin oasis in northwest China during recent 50 years. Journal of Environmental Systems, 31(4), 357-365. Zhang, X., Wang, X., Yan, P., 2008. Re-evaluating the impacts of human activity and environmental change on desertification in the Minqin Oasis, China. Environmental Geology, 55(4), 705-715.