Integrating rural livelihood resilience and sustainability for post-disaster community relocation: a theoretical framework and empirical study

The concepts of resilience and sustainability are becoming popular in disaster planning and management. However, there is an absence of mutual considerations of the two concepts from an integrated perspective to promote post-disaster livelihood, particularly in the relocated rural communities. To fill the research gap, this paper explores the factors and relationships of resilient rural livelihoods (RRL) and sustainable rural livelihoods (SRL) in resettlement communities after a major disaster. Specifically, we construct a theoretical model that integrates resources, strategies, and outcomes of RRL and SRL in the recovery phase. We use a dataset for household-level recovery after the Jiangsu Tornado (2016) in China to validate the theoretical model with a structural equation modeling (SEM) approach. Results show that government-driven, recovery-concentrated policies of “resilience” may not promote the long-term sustainability of rural household livelihoods because financial capital and institutional participation are negatively associated with the community’s self-reliance for future disaster recovery. The tangible assets are likely to make people more confident in disaster preparedness, while the intangible assets lack such an impact and even make the relocated households concerned about future disasters. For RRL outcomes, financial capital and socio-political capital can enhance the recovery, while human capital and livelihood strategies have negative effects. In addition, natural capital and financial capital positively affect household recovery, but the human capital remains a negative contributor. These findings clearly reveal the general patterns of rural livelihoods in relocated communities after a disaster and provide insights into potential measures to promote sustainable and resilient rural livelihoods.


Introduction
Sustainability and resilience are two different but interrelated concepts that are dominating research trends and practical interests in the environmental transformation process of rural livelihoods . Both are building blocks of environment management and life quality improvement (Lew et al. 2016). They can provide a holistic understanding of the livelihood recovery from disastrous disturbances, including the mechanisms of adaptation, improvement, diversification, and transformation in face of disasters. Meanwhile, full knowledge of livelihood resilience to disasters is also critical to the sustainable development of rural communities in disaster-affected areas (Mayunga 2007;Sina et al. 2019a).
Although some disaster-related resettlement strategies are popular in practice, more and more scholars started to challenge their effectiveness because these strategies often overlook livelihood sustainability, such as natural capital and a feeling of security (Christoplos et al. 2001;Guo et al. 2019). Government-led disaster recovery projects can bring many external resources to the survivors of disasters promptly, but it should be just viewed as a temporary and passive strategy. This is because such strategies emphasize a rapid reconstruction of the built environment with an enormous investment of financial and policy resources, but they rarely consider the long-term well-being of the community. These reconstruction activities could even cause a series of negative effects, such as depletion of public resources and reduced capacity of residential buildings. Therefore, the "successful recovery" in the built environment can drain out local and household adaptability and mobility, and, eventually, worsen their livelihood in the long run (De Haan 2000;Helmore and Singh 2001).
To address this dilemma, this study calls for a sustainable livelihood mindset for disaster recovery policies, which encourages a community's self-mobilization and capacity building rather than simply relying on external intervention. It should be a gradual but permanent transformation process to rebuild the livelihood during and after the disaster recovery phase. Furthermore, the notion of sustainable livelihood concerns an all-hazard capacity of long-term adaptation to different types of disturbances rather than the one-time, superficial resilience to a single shock. Therefore, the strategy of seeking sustainable livelihoods can be very different and more holistic than that of a rapid recovery from disastrous disturbances (Pain and Levine 2012). With this consideration, there is an urgent need to conduct the synchronization study of resilient rural livelihood (RRL) and sustainable rural livelihood (SRL) among the communities relocated by governments in response to disastrous disturbances, which is seldom involved in the literature. In addition, a long-term disaster relief strategy is not on the radar of government authorities yet. Therefore, it is urgent and meaningful to integrate RRL with SRL in a common theoretical framework, which can help us disentangle the mechanism between responsive recovery of post-disaster communities and sustainable development of livelihoods in the long run.
Post-disaster resettlement has been a popular recovery strategy of Chinese governments in helping the community recover from natural disasters, such as the 2008 Wenchuan earthquake (Chen et al. 2017) and the 2016 Jiangsu Tornado (the case in this study). Intending to satisfy the housing demand of survivors rapidly, these government-led projects consume enormous resources from other jurisdictions or upper administrative hierarchies rather than the community itself. Weldegebriel and Amphune (2017) emphasize that livelihood recovery is critical to rural livelihood resilience against disasters. When public resources are allocated among households during the livelihood recovery, the survivor households can obtain new, permanent housing units as tangible livelihood capitals, which can improve their livelihood resilience due to the good design of these new housing units. However, this housing policy does not promote self-reliance for sustainable development on both the household and community levels (Daly et al. 2020). With the empirical research in communities relocated by governmental authorities, this study verifies the loop of resilient livelihood capitals, strategies, and existing resilience outcomes and future sustainable livelihood outcomes within the institutional contexts of disaster recovery.
Here, we select Jiangsu Tornado (catastrophe) as a case study to investigate the recovery process of rural neighborhoods against disaster disturbances and evaluate the efficiency of corresponding recovery strategies in promoting sustainable rural livelihoods in the long run. These post-disaster reconstruction strategies are named "counterpart support" (Wu and Guo 2021) and initiated by the Chinese government promptly and in line with the principles of "built back better." The 2016 Jiangsu Tornado was ranked Enhanced Fujita Scale 4 (EF4) and produced detrimental damages to multiple rural communities in affected areas of Jiangsu Province, China. Specifically, the disaster caused a total direct economic loss of US $720 million (4983 million yuan), 35% of which was property loss. The disaster affected 45,509 residents, including 99 deaths, 875 injuries, and 30,303 relocated (Ni et al. 2019). To help residents recover from the disaster, the Chinese government provided $93 million (USD; 650 million yuan) for the post-disaster resettlements. The whole recovery process for rural livelihoods started from the date of the disaster strike (June 23, 2016) and lasted until the delivery of all the new housing units. It is noteworthy that the housing resettlement was not free of charge. The relocated households received government aid and societal donations to cover the costs of housing construction. Additionally, many households implemented a variety of strategies to cope with the livelihood disturbances on top of the governmental aid from the government.
However, the literature remains unclear about how different livelihood capitals and recovery strategies affect existing livelihood resilience outcomes and future sustainable livelihood outcomes following the post-disaster institutional process. To fill the research gap, this study develops a theoretical framework to examine the relationships between different livelihood capitals/strategies and the fulfillment of resilient and sustainable livelihoods. Specifically, the theoretical model integrates the concepts of resilient and sustainable recovery to capture the changes in post-disaster livelihood. Then, an empirical study tests and validates the theoretical model using the archive data from local documentation and survey. The findings can help us reevaluate and optimize the design and implementation of government-led programs toward responsive recovery and long-term sustainability of rural livelihood after a major disaster.

Literature review: the relationship between SRL and RRL
The literature review identifies how sustainability and resilience are jointly implemented to guide livelihood development. Current models of sustainable livelihood operationalize the buzzword "resilience" for both slow-onset and fast-onset disturbances. For example, Sina et al. (2019a, b) argue that livelihood recovery is a top priority in sustainable rural community redevelopment. In other words, resilience and sustainability are two important and relevant constructs in tackling post-disaster interventions and assisting the survivor to recover rapidly or build better. However, there is a lack of theoretical discussion differentiating RRL from SRL or suggesting a means of integrating them into one framework. Therefore, our theorization attempts to clarify such a relationship by considering the 1 3 "recovery performance" as a measure of RRL and, in turn, the RRL as a stepping-stone to SRL among communities. Table 1 summarizes different definitions and theoretical frameworks of SRL, as well as their contributions and deficiencies to the knowledge domain. It is noticeable from many publications that SRL cannot stay away from resilient considerations. For example, SRL requires the ability of livelihood to recover from "stress and shocks" and maintain or enhance "capabilities and assets" into the future (Chambers and Conway 1992). Followed by this, the diversification of livelihood strategies is also added to the resilient aspect of SRL (Davies 1996). Specifically, the conceptual framework proposed by Davies (1996) considers that resilience is useful in analyzing changes in levels of vulnerability to food insecurity within diversified livelihood systems. In addition, the commonly cited theoretical framework of SRL (Scoones 1998) specifies that improving resilience and adaptability is a central part of sustainable livelihood outcomes. Scoones (1998) further argues that resilience contains both temporary adjustments in response to change and adaptive capacity through long-term shifts in livelihood strategies. Since then, SRL has been always serving as the principle of rural development for some international organizations, such as the former Department for International Development (DFID) in the United Kingdoms, the United Nations Development Program (UNDP), and the Canadian International Development Agency (CIDA). These SRL principles focus more on the efficient livelihood strategies that cope with economic poverty (Ashley and Carney 1999), ecological health (Connell 2006), and even local risks (Carney 1999). Then, other scholars keep adding more resource factors into the frameworks of SRL, such as power and politics (Nyamu-Musembi and Cornwall 2004) and technology uses (Mengistu et al. 2015;Donkor et al. 2019). At the same period, Scoones (2009) emphasizes again that the frameworks of SRL conceptually enhance their interactions with the concept of livelihood resilience by incorporating the issues of knowledge, politics, scales, and dynamics.

Operationalizing RRL with sustainable considerations
Compared with SRL, RRL as a combination term of "resilience" and "rural livelihood" focuses on the design and implementation of adaptation strategies to help rural neighborhoods efficiently resist and rapidly recover from stresses and shock. It has been widely promoted in many fields, such as engineering, psychology, ecology, and sociology. For instance, a humanitarian organization (the International Federation of Red Cross and Red Crescent, IFRC) adopts this concept of resilience when analyzing vulnerabilities. Nyamwanza (2012) develops a stand-alone construct of livelihood resilience as an insightful tool to understand the dynamic environment. Tanner et al. (2015) conceptualize livelihood resilience closely with Chamber and Conway's (1992) SRL, both of which focus on capability and assets across generations. Tanner et al. (2015) also highlight the social aspect of RRL, such as the roles of adaptation policy, human agency, and individual and collective capacities in responding to stressors, thus engaging broader livelihood dynamics.
A sustainable livelihood approach (SLA) with capital has been widely utilized to measure livelihood resilience in a handful of studies (Thulstrup 2015;Quandt et al. 2017;Rudiarto et al. 2019). Such studies pay more attention to the factors that affect rural households' livelihoods, including stock assets and social networks. It is observable that a Maintain and enhance its capabilities and assets Provide sustainable livelihood opportunities for the next generation The first one clarifies the concept of SRL but lacks a detailed analysis process Chambers and Conway (1992) Food security Encompass a broader range of factors than house food systems and security Explain how and why producers pursue mixed strategies to confront insecurity Expand SRL to livelihood security Davies (1996) Rural poverty Obtain a range of livelihood resources (natural, economic, human, and social capital) Pursue different livelihood strategies (agricultural intensification or extensification, livelihood diversification, and migration) Provide an analysis framework, but 'Political capital' is excluded Scoones (1998) Rural poverty People-centered, responsive and participatory, multilevel (micro and macro) Conduct partnerships with private and public sectors Sustainable (economic, institutional, social, and environmental), and dynamic (changing) Add institutional capital into the SRL theoretical framework; Provide operationalization principle Ashley and Carney (1999) Local risks

Participatory assessment
Multiple levels of analysis Complement indigenous technologies Integration in real-time Propose science and technology as promoting elements in the operationalization of SRL; Move toward a rights-based program Carney (1999) Rural poverty Integrate human rights into the frameworks of SRL Address power relationships directly, but does not specify the changed process into the concept of SRL Nyarnu- Musembi and Cornwall (2004) Poverty and ecological health Place the household and its assets within the context of complex systems Scale up the SRL approach with complex adaptive systems theory and ecosystem health methods (2006) capital-oriented approach to studying disaster resilience becomes a key criterion in measuring the success of recovery and reconstruction following a large-scale disaster (Mayunga 2007). Correspondingly, the resources contributing to resilient livelihoods can be grouped into six types of capital, namely human, social, physical, natural, financial, and political capital when rural communities face slow-onset pressures, such as poverty and environmental degradation. However, the indicators of capital assets may not be adequate for the evaluation of livelihood resilience in the cases of post-disaster relocations. A more comprehensive assessment needs to be further developed with additional considerations of the adaptation and adjustment behaviors of rural residents under external shocks. In terms of post-disaster relocation settings, some studies (Speranza et al. 2014;Weldegebriel and Amphune 2017;Smith and Frankenberger 2018;Sina et al. 2019b;Liu et al. 2020;Zhou et al. 2021) use a capacity-based framework to model livelihood resilience, including buffer capacity, selforganization, learning capacity, and disaster prevention and mitigation capacity. These studies also consider capital-oriented measures when characterizing the buffer capacity. They commonly argue that rural residents can adjust their behaviors based on social networks and social learning per the changes in the socio-ecological environment.

Connellz
The capacity-based perspective emphasizes more on social capital while neglecting government-led capacity-building and the institutional process in SRL. As a result, their findings are not fully consistent with the Chinese experiences where the government-led programs dominate and advocate quick recovery from severely damaged areas by disasters. Furthermore, little research is done to articulate the centralized policies of relocations, but these policies are very common in China and are often guided by post-disaster reconstruction planning with the full support of cross-hierarchy assistance at all levels of Chinese governments (Tables 2 and 3).

The comparison between SRL and RRL
Since the 1990s, SRL in earlier research has been a prevailing concept within conceptual frameworks focusing on poverty alleviation through improving livelihood assets. The goal of SRL is to guarantee the long-term sustainable development of rural neighborhood systems by keeping function operating. Accordingly, the conceptual development of RRL is based on the SRL theoretical framework and meanwhile, it also stimulates lots of new empirical methods and applicable tools. This can be generally attributed to the following three characteristics underlying RRL: (a) good outcomes despite high risk, (b) sustained competence under conditions of threat, and (c) rapid recovery from shocks and stresses (Boyden and Cooper 2007). Therefore, RRL can be a short-term or long-term concept, which is mainly determined by the types of shocks: low-pace stresses or rapid shocks. Different from the local-scale perspective of SRL, livelihood resilience emphasizes the engagement of cross-scale connectivity in the theoretical conceptualization of RRL. This naturally involves a transfer and flow process of resources amongst different entities within the livelihood system (Nyamwanza 2013). In addition, some researchers (Jones and Tanner 2017;Alonso-Tapia et al. 2019) propose the concept of "subjective" resilience to characterize an individual's self-evaluation of their household's capabilities in resisting and recovering from disasters. This can be achieved by the design and delivery of survey questionnaires to quantify the "subjective" resilience and supplement the theoretical framework of SRL that mainly consists of objective socio-economic variables before.

Foundations of the theoretical model
In the context of post-disaster rural community relocation, we define the concept of RRL as a rapid recovery and improvement of rural livelihoods relying on government subsidies and other assistance. In contrast, the goal of the SRL is to promote socio-economic self-sufficiency and self-reliance with limited external support (Barrios 2014;Chen et al. 2017). To further differentiate the two concepts, we consider the RRL as a performance measure of livelihood recovery (household recovery and recovery improvement) after a disaster and measure the SRL as the level of independence from the support of upper hierarchies in the governmental administration (independent from government and confident in the next disaster).
In addition, we assume that the RRL can be the initial strategy for the fulfillment of the long-term SRL. For example, a household can achieve RRL by recovering from the disastrous disturbance and even living a better life with aids outside its community. However, this household may still not achieve SRL without self-confidence in coping with future disasters itself. Similarly, the government-led recovery process, along with its resilient goals, can be achieved after the delivery of the relocated housing units; however, it does not necessarily fulfill the sustainability goal in the reconstruction process of rural livelihoods. SRL should be accomplished with the capabilities of self-adaptation and self-recovery against future disasters. To build and enhance the capabilities of SRL, the capital-based approach is commonly used to assist in integrating livelihood resilience and sustainability in the relocation process of affected communities by disasters, which is also used and further discussed in this study.

Disaster recovery: a window for RRL and SRL
Disaster recovery and reconstruction open a window for promoting RRL and SRL, which are a phase often related to restoring the pre-disaster conditions or built-back-better from disturbances with the aim of sustainable well-being on a material basis. We use the definition of disaster resilience from DFID (2011), "maintaining or transforming living standards in the face of shocks or stresses," to extend such a typical notion to a broader context than material reconstruction. In other words, this study suggests that RRL can be understood as a target of a set of post-disaster rehabilitation practices. In rural areas, government-led recovery can be welcome because of the local social-economic conditions, such as inadequate disaster insurance, low family income, and the aging population. Such a goal of resilience focuses on tangible infrastructure rather than building mutual trust, equity, and community cohesion. However, a resilient livelihood recovery cannot be achieved without incorporating the elements of local livelihood, including empowerment, toughness against adversity, and independence. We challenge that the government's top-down approach can inhibit the grassroots resilience-building efforts and motivations for rural livelihoods. In observation of these issues in practice, this article examines the roles post-disaster rehabilitation plays in building RRL and considers how they can impact SRL in various aspects, including livelihood resources, livelihood strategies, and livelihood outcomes.

Establishment of the theoretical model: from RRL to SRL
As illustrated in Fig. 1, there is an interactive loop among rural livelihood resources, rural livelihood strategies, post-disaster rehabilitation institutions, and livelihood outcomes, which is also the main idea of the integrated theoretical model. Specifically, the livelihood capitals could be reconfigured in the process of post-disaster rehabilitation. Rural households can get new houses, infrastructures, and subsidies from the government; meanwhile, they also consume household savings and other capital to adapt to the disaster, such as the decoration of new houses. Rural households may make responsive decisions against living strategies to get a safer environment or earn more money. The three aspects are interrelated and can co-effect livelihood outcomes.
The theoretical model integrating SRL and RRL is adapted from Scoones's (1998) framework in modeling and conceptualizing the livelihood dynamics throughout the recovery process. The first key factor is the livelihood resources in the forms of different capitals. These resources are defined as the sum of monetized assets, including labor forces, arable lands, household incomes, working days, etc. (Scoones 1998), which correspond to human, natural, and financial capital, respectively. In addition to the three capitals, some   (Forster et al. 2014). However, seldom is done in the literature considering the psychological capital (i.e., self-efficacy, hope, optimism, and resilience) of survivors during the recovery process. To fill the research gap, we attempt to model all five capitals in our theoretical model and explore their relationships with both resilient and sustainable livelihood outcomes.
Another key factor is the livelihood strategies referring to the allocation of assets and selection of livelihood activities to achieve people's livelihood goals (Guo et al. 2019). Livelihood strategies are not static and are constantly evolving due to the changes in the recovery policies, socio-economic environments, and the capital people possess. Previous studies have shown that livelihood strategies exert a great influence on the recovery process from disasters, such as hurricanes (Schramski and Keys 2013). Therefore, livelihood strategies remain a major component of our integrated theoretical model.
In addition, we are particularly interested in the post-disaster rehabilitation institutional process because it plays a unique role in explaining livelihood outcomes under China's post-disaster support programs. Some political relationships can be established between the relocated residents and the authorities through public participation or empowerment procedures in the institutional process of reconstruction. For instance, the residents have opportunities to provide suggestions about resettlement sites, the design of the neighborhoods, floorplans, and additional resources as needed. Such participation activities can also influence the allocation process of resources and, consequently, affect livelihood recovery and sustainability.

Hypotheses of the theoretical model
Rural households accumulate resources for livelihoods before any disaster strikes, such as human and natural capital, which can assist their recovery process in face of disasters. Meanwhile, disaster disturbances can also bring changes to the capital during the response and recovery phases. Hence, our theoretical model captures these capital and livelihood factors in the post-disaster settings to opt-out of potential compound effects due to the predisaster livelihood strategies. To achieve it, we have to first obtain a basic understanding of the Chinese practices in post-disaster community relocation. For example, (1) relocated households by disasters receive governmental subsidies and compensations for their original lots submitted; (2) Relocated households pay for new housing units with their savings or loans; (3) Some households can earn salaries by joining the reconstruction projects; (4) The farming activities shift from household responsibilities to land shareholding cooperation, which can change land revenues and free up labors. Therefore, we construct the three hypotheses of the theoretical model as follows: H 1 : The resources (capital) of relocated households are associated with their post-disaster livelihood sustainability.
• H 1A : The resources (capital) of relocated households are associated with their perceived independence from the support of the central government and policies. • H 1B : The resources (capital) of relocated households are associated with their confidence in dealing with future disasters. • H 1C : The resources (capital) of relocated households are associated with their perceived status of disaster resilience.
• H 1D : The resources (capital) of relocated households are associated with their perceived status of disaster recovery.
Since disasters striking rural areas can severely affect the crops and non-harvest areas, the survivor households have to take necessary response strategies to maintain their livelihood. For example, some households can purchase housing units in urban areas; some can choose to work in larger cities and switch to industrial jobs for higher incomes. Those who still stay in rural communities can participate in various socio-political groups to enlarge their social network and exchange information about their lives, such as the reconstruction and resettlement projects and policies.
H 2 : The response strategies of relocated households are associated with their post-disaster livelihood sustainability. Participation in the reconstruction process can suggest the following characteristics of a household: (1) possessing good knowledge of institutional opportunities, (2) being willing to obtain community leadership or political powers, and (3) having local occupations. The households who give up public participation opportunities in the recovery process may have busy schedules or do not believe that their ideas can make any difference. On the other hand, those who care about the reconstruction details or even quit other jobs for such low-salary local jobs might have a better understanding of the potential capital reconfiguration and livelihood changes in the long run.
H 3 : The institutional participation of relocated households is associated with their postdisaster livelihood sustainability.
• H 3A : The institutional participation of relocated households is associated with their perceived independence from the support of the central government and policies. • H 3B : The institutional participation of relocated households is associated with their confidence in dealing with future disasters. • H 3C : The institutional participation of relocated households is associated with their perceived status of disaster resilience. • H 3D : The institutional participation of relocated households is associated with their perceived status of disaster recovery.

Study area and community resettlement policy
The 2016 Jiangsu Tornado affected 9 towns (districts) and 29 villages with 7301 housing units seriously damaged. Those affected households experienced two relocations after the disaster.
(1) Temporary housing: The local governments implemented two housing assistance policies, including a disaster emergency relief fund for US $45 (310 yuan) per person and transitional living assistance for US $87 (600 yuan) per person-month (no more than 3 months).
(2) Permanent housing: The provincial government resettled a total of 5918 rural households in 20 centralized relocation sites by initiating the Chinese housing support programs. Specifically, every household received a subsidy of US $5073 (35,000 yuan) from Jiangsu Province, compensation of US $2900 (20,000 yuan) for old homestead reclamation, concessional loan of US $2900 (20,000 yuan) from county-level governments, and the opportunity to purchase one new housing unit at a price much cheaper than the market. Additionally, the childless senior and lower-income households whose annual income was less than US $300 (2000 yuan) per person received permanent housing free of charge. These characteristics make the Jiangsu Tornado an appropriate case for testing our theoretical model. Since the survivor households were rapidly relocated by the upper levels of government and the reconstruction projects were finished within one year, the empirical study focused more on the establishment of a "resilient" recovery from the standpoint of policymakers.

Data collection and description
In 2018, Jiangsu Huaxin Planning and Municipal Design Co. Ltd. (Huaxin) in China administered a survey to document the recovery status of the relocated survivors after the delivery of new relocated houses. The survey tool contained a comprehensive list of questions to measure the social, economic, political, and psychological aspects of community recovery at the household level. Using the clustered sampling method, the investigators of Huaxin received 494 responses from 13 newly relocated communities, making it one of the best candidate datasets for our research objectives. Specifically, these investigators randomly selected 13 communities from a total of 20 resettlement communities and collaborated with the community/village leaders to distribute the survey questionnaires during their village convocations. Among the 494 responses collected by Huaxin, 461 were valid and entered in the descriptive analysis and 429 had all questions completed for the structural equation model. As presented in Table 4, we locate the following questions from this survey to help us operationalize the constructs in our hypotheses, and we also saw the traces of the rural livelihood sustainability framework in the survey design. The numbers in the parentheses are labeled in the column headings in parentheses as well. Most respondents (280 out of 461) reported annual household income within the range of 30k and 100k Yuan (approximately $4.2k-$14k USD). The average household size is 3.89 individuals (standard deviation 1.25), and 1.54 (standard deviation 0.82) household members are currently employed. Regarding their political affiliations, only about one-fourth (111 out of 461) are members of any party, including the Communist Party, the Communist Youth League, and the democratic parties. Inspired by the capital-based approach measuring livelihood development (Zhang and Fang 2020;Berchoux et al. 2020;Garrigos-Simon et al. 2018; Šlaus and Jacobs 2011), we  identify the following questions from the survey for scaling the constructs in the model. We use the cultivated area to measure the natural capital. The family income, economic condition, and expenditure of purchasing the housing unit are the scales of the financial capital. Human capital is represented by household size, education, and employment. The measurement of the socio-political capital includes partisanship, membership of social groups, and frequency of attending group activities. Nine of the psychological questions in the survey match the comprehensive scaling of psychological capital (Lorenz et al. 2016), namely: the confidence in dealing with future traumatic events, communicating with neighbors/friends/family members, actively seeking help when needed (self-efficacy); perceived risks of disasters happening in the community and those affecting jobs (optimism); being able to get economic and emotional help from others (resilience). The livelihood strategies and institutional process contain the activities that are unique in the relocation phase after the tornado. We convert five sets of questions into five dummy variables to document the strategic activities available for the relocated households after the disaster. The five strategies include submitting pre-disaster properties for reallocation, changing jobs/occupations, changing the affiliation of social groups (e.g., the Communist Party), having household members working outside the locality, and purchasing new housing units. For the institutional process, the local and provincial reconstruction offices offer eight opportunities for the relocated households to participate in the decision-making process. Similarly, we convert these questions into binary scales, such as whether participating in the reconstruction process, participating in collecting location preferences/opinions, and participating in the hearings of comprehensive planning, etc.
The endogenous variables include perceived independence from central-government support, confidence in dealing with future disasters, household recovery, and recovery improvement. The survey has one question asking the rural households how necessary the support from the central government and policies are for recovery. We reverse-code the scale of this question to capture the perceived independence from the central government's support of the local recovery. The worrisome of the same disaster in the future is also reverse-coded to measure the confidence construct. The recovery-related questions in the survey are grouped using principal component analysis (PCA) initially, resulting in a component of post-disaster changes and another component with general self-assessment of disaster recovery and life improvement. Upon further examination of the second component, we find that the answers are clustered at the higher ranks in both questions (i.e., almost/totally recovered and some/lots of improvement) and that sizeable households (N = 179) who report "some improvement" or "lots of improvement" also mark "almost recovered." Therefore, we multiply both variables to capture the potential interaction between them.
We use PCA to validate the unidimensionality of these multiple-scale measures, of which the loadings are presented next to the survey questions in Table 4. The reliability tests, including Cronbach's Alpha (α) and Guttman's Lambda-2 (λ 2 ), indicate acceptable to strong internal consistency for socio-political capital (α = 0.560; λ 2 = 0.583), psychological capital (α = 0.752; λ 2 = 0.769), and household recovery (α = 0.751; λ 2 = 0.753) as many measures are exploratory in this study (Callender and Osburn 1979). The measurement of human capital and financial capital have low Alphas, but the average variance extracted (AVE = 0.443 for financial capital and 0.490 for human capital) and composite reliability (CR = 0.697 for financial capital and 0.740 for human capital) confirm the reliability and convergent validity of it as well (Bernardi 1994;Ab Hamid et al. 2017). Hence, we move along to the structural model for testing the theoretical hypotheses with these measures.

SEM results and interpretation
The theoretical model suggests potential relationships between seven predictors (five capitals, livelihood strategies, and institutional process) and four outcomes (independence, confidence, household recovery, and recovery improvement), most of which are latent constructs containing three or more measures each. Then, we use the SEM approach and survey data to test the theoretical model by capturing/controlling the measurement relationships, the causal effects between latent variables, and the correlations among the predictors at the same time. SEM can also assess the reliability and validity of the model measures, as well as detect potential error covariance between scale items that the model does not specify but the theory permits (Kang and Ahn 2021). The structural model in the following illustrates the relationship between the exogenous variables (household capital, response strategies, and institutional process) and the endogenous variables (livelihood sustainability and livelihood resilience) with the model fit indexes. The measurement model section specifies its model fit and the error covariance.  the light blue ovals and the endogenous variables on the right side are color-labeled with the paths toward them. The statistically significant paths are the thicker, solid arrows with directional signs (+/−) next to them, while the insignificant ones are the thinner dashes.

Results of the structural model
Five household capitals, including natural, financial, human, social/political, and psychological capitals, capture the resources that can contribute to a household's recovery process. The livelihood strategies/responses and the institutional process/participation document the survivors' activities after the disaster. The structural model is significant as the major fit indexes (Table 5) are within an acceptable range (i.e., root mean square error of approximation/RMSEA < 0.06; standardized root mean square residual/SRMR < 0.08; comparative fit index/CFI > 0.90; incremental fit index/IFI > 0.90). Table 5 details the relationships between exogenous and endogenous variables in the structural model, including the statistical significance (t-tests, sig. in bold), path strength (standardized estimates), and directions.
In particular, two predictors (financial capital and institutional process/participation) are negatively associated with the perceived independence from the central government's support if disasters happen in the future ("Independence"). It is reasonable to believe that the wealthier households and those who participate more often in the institutional recovery projects are closer to the central government and, therefore, see fewer opportunities for local communities to recover without central support.
Regarding confidence in dealing with future disasters ("Confidence"), a household's natural capital and financial capital can help make its members less worried about dealing with the next one of the same kind. However, those with a higher human capital concern more about future tornados on average. These findings suggest that tangible assets are likely making people more confident, while intangible assets lack such an impact and even make the relocated households more concerned about future disasters.
Household recovery is associated with four significant exogenous variables statistically. Specifically, both financial capital and socio-political capital can enhance the recovery at the household level within one year after the disaster. By contrast, the human capital and response strategies negatively predict the levels of the recovery measures. In other words, possessing resources and having some political affiliations could promote household disaster recovery in a variety of aspects. However, investments in future household development can jeopardize or delay the short-term recovery as they might occupy too many resources of a household.
In terms of recovery improvement, natural capital and financial capital are positively predicting recovery improvement, while human capital remains a negative predictor. This indicates that only the tangible assets (natural and financial capitals) can keep promoting the recovery to an upper level, while the socio-political affiliations could only boost the recovery but not necessarily make further improvements in livelihood. The household investment in human capital, measured by the household size, education, and employed members, remains a liability to making the post-disaster living condition better one year after the disaster.
From the predictor's standpoint, financial capital is statistically significant in explaining all the endogenous variables and their relationships are all positive except when explaining the independence from the central government's support. Human capital plays the opposite role in explaining three out of the four outcomes with negative relationships, namely confidence in dealing with future disasters, household recovery, and recovery improvement. The natural and socio-political capitals can significantly predict one outcome each. These findings suggest that our first hypothesis receives the strongest support from the data. The response strategies and institutional process, respectively, explain only the household recovery and perceived independence among the four constructs of livelihood sustainability. That is, the data can provide partial support for our second and third hypotheses. Table 6 reports the measurement model fit and the error covariance added to the complete structural equation model. The measurement model alone, with the same modifications, has fit indexes as good as the complete model consisting of both measurement and structural models. The error covariance represents the modifications suggested by the SEM package (Lisrel 8.70). In particular, we add 13 pairs of error covariance labeled from 1 to 13 shown in the last column and any two variables with the same label have an error covariance connecting to each other. For example, Label 5 is shared between "natural capital/cultivated area" and "household size," meaning an error covariance is added to connect them because the cultivated area is designated based on household size. These error covariance terms have their roots in the policies and norms of rural communities (e.g., household size and ability to get economic help; Label 6), if not between two variables measuring the same constructs in this analysis (e.g., partisanship and number of social groups; Label 3).

Results of the measurement model
In summary, the first set of hypotheses (H 1 ) has at least one predictor (financial capital) consistently explaining all four livelihood sustainability outcomes after a tornado. It also has another predictor (human capital) explaining three out of the four outcomes. We find partial support for the natural and social capital in predicting sustainability outcomes. The data do not provide enough evidence for the relationship between psychological capital and the outcomes. The other two sets of hypotheses (H 2 and H 3 ) are partially supported, in which H 2C and H 3A find statistically significant evidence in the data.

Discussion and conclusion
This study integrates the theories of RRL and SRL to enrich our knowledge of the policy and livelihood outcomes in terms of disaster resilience and sustainability. The policy efforts aiming at short-term resilience do not necessarily build long-term sustainability, which is important for both livelihood recovery and disaster capacity at the local level. In contrast to the short-term recovery outcomes towards livelihood resilience, livelihood sustainability is a long-term process built on the self-motivated development and strategies of rural households and communities. Ineffective household strategies and institutional development can make little improvement in such sustainable goals at best, if not doing any harm to the relocated survivors. The large number of resources allocated by the upper levels of government hierarchies are mainly used for constructing permanent housing in the new locations. However, they rarely focus on the slower processes of capacity and livelihood building, such as promoting human capital, community relations, and psychological recovery. Therefore, the survivors might have no choice but to rely on their private resources if they need to plan for the future. For example, the relocation policy requires the survivor households to submit their homestead lot, along with the disturbance of farming, making the farmland management further concentrated in the village. If the income from the village is uncertain, the rural residents prefer to purchase housing units and look for jobs in urban areas, which could spoil their opportunities for the recovery of rural livelihood.
More contextual evidence can echo our findings and interpretations on a cumulative scale. From our social survey, a community named Danping is a typical "hollow village" that suffered greatly from the tornado strike. More than 2/3 (1823 people) of its relocated residents (544 households, 2726 people) were working outside the region, leaving less than 1000 seniors, juniors, and patients living in the village only. Every household has around 80% asset and property losses during the disaster and has to use personal loans and mortgages for purchasing new homes, which exceed their savings and government subsidies. This indebtedness becomes a major concern or even a stressor for these households. Hence, it is reasonable that those households with a better financial status have better recovery and higher confidence in dealing with future disasters. The relocation policy is accompanied by some forms of landholding concentrations, which could return a new housing unit to each rural household whose annual income is less than US $300 (2000 yuan) on average. In the meantime, the new living style of the semi-urbanized relocated villages incurs many additional costs, such as homeowner association fees and interior design/decoration costs. The rural residents were "mandated" to diversify their incomes from agriculture to the manufacturing and service industries. These transitional strategies are rather attempting to slow down livelihood declines due to policy-oriented disturbances than promoting livelihood during disaster recovery. Additionally, the rapid recovery of the government reconstruction/relocation projects could yield counterproductive consequences. On the one hand, many childcare and healthcare facilities are in an idle state due to the hollow-village effect. On the other hand, these facilities are inadequate for the local population structure, such as public restrooms and fitness grounds. The qualitative answers in the survey also revealed that a lack of communication is a potential cause of these issues. For instance, some respondents reported, "We don't know how the local authority allocates the provincial/central incentives" even if they participate in the reconstruction processes. Others wrote, "We have no idea who can offer help as the government projects are over… so we have no choice but to rely on ourselves to deal with life pressures." This suggests that even the speedy relocation and reconstruction can inform the rural residents of how powerful the government wills are, but still miss proper channels to assess the public demand at the household level. Consequently, the resources from the central and provincial governments do not become local and household sustainability in terms of rural livelihood after a major disaster. Local and provincial authorities can develop special programs to facilitate the communication between government and rural residents and empower rural communities rather than dominating the reconstruction process.
This study has several limitations future works can address. First, the sample is only collected from the relocated villages after a major tornado. Those rural communities rebuilt on the same site or recovered from other disasters may have different results. On the other hand, the authors are confident in their findings that suggest a closer examination of the livelihood consequences of the resettlement policies dominated by upper government hierarchies. Such policies aiming at "resilience" might trigger unexpected issues (e.g., draining resources of rural households and overly relying on support from the central government) in countries and polities with strong central governments. Second, the data are from an existing survey designed for policy and planning purposes. That is, a future study can improve the scaling by designing the questionnaire more specifically for the disaster sustainability and resilience of rural livelihood. For example, the measures of the response strategies can cover asset allocation and business activities separately and the outcome variables can be more comprehensive to include traditional sustainability constructs. Finally, future studies can consider adding semi-structured interviews and focus groups of critical stakeholders, such as local policy-makers and rural community leaders, to fully explore policy alternatives that have the best promise of rural livelihood sustainability.