2.1 Literature Review

Community disaster resilience was the core independent variable investigated in this study. It refers to (1) adaptation to a disaster-prone environment over a long period of time, (2) having good prediction and early warning systems, and (3) having a coordinated response capability that does not greatly rely on external rescue; rather, the environment and social structure can be restored to a pre-disaster state in a self-sufficient manner. ^{Reference Cimellaro, Reinhorn and Bruneau20–Reference Zhou, Wang, Wan and Jia24} “Resilience” was introduced into disaster risk management research and widely accepted by the academic community in the 2005 World Conference on Disaster Risk Reduction. ^{Reference Cui, Han and Wang14} Its research objects include individuals, families, communities, and even larger regions. ^{Reference Ostadtaghizadeh, Ardalan and Paton11} As the smallest social organization unit at the grass-roots level, community plays an important role in the occurrence of disasters. Therefore, after the concept of resilience was proposed, many disaster risk reduction projects and studies focused on the resilience capacity building and improvement of communities ^{Reference Cutter, Barnes, Berry and Burton25–27} and believing that community disaster resilience is the foundation of post-disaster reconstruction and recovery. ^{Reference Pfefferbaum, Pfefferbaum and Van Horn28} Since then, more and more scholars began to pay attention to the measure of community disaster resilience and its application in disaster risk reduction, and proposed the measure system of community disaster resilience from different angles. For example, Kafle ^{Reference Kafle29} used process and outcome indicators to measure the community disaster resilience of the Indonesian community; Orencio and Fujii ^{Reference Orencio and Fujii30} measured the community disaster resilience of the Filipino community from the environmental and natural resource management, health and well-being, sustainable livelihoods, social protection, financial instruments, physical protection, and planning regimes 7 dimensions; Mayunga ^{Reference Mayunga31} measured the community disaster resilience of the American community from social, economic, human and physical these 4 dimensions; Ostadtaghizadeh et al. ^{Reference Ostadtaghizadeh, Ardalan and Paton11} introduced the models and tools for community disaster resilience assessment in the review system of community disaster resilience and further pointed out that the measurement of community disaster resilience should include social, economic, institutional, physical, and natural 5 domains. These studies provide useful inspirations for the research of community disaster resilience and disaster risk reduction. However, Chinese research on community resilience and disaster prevention capacity building has focused on post-disaster recovery and reconstruction. For example, from the perspective of architecture, some scholars have proposed strategies for ensuring community security by means of technological progress (eg, Li and Xu ^{Reference Li and Xu32} ). From the perspective of social relations, some scholars believe that the communities should change from having government administration to having coordinated and bottom-up grass-roots participation (eg, Jin and Lu ^{Reference Jin and Lu33} ). Generally speaking, there is no short-term recovery after a disaster, and communities need to rely on their own resource endowment (eg, savings) and outside resources (eg, outside financial and human assistance) for post-disaster reconstruction. ^{Reference Cui, Han and Wang14,Reference Shang and Liu34} It can be seen from the situations of several recent major earthquakes in Sichuan Province that after an earthquake, many communities do not have the resilience and disaster prevention capacity to complete post-disaster reconstruction. Instead, they need to make extensive use of external forces. ^{Reference Yang35} However, for disasters as serious as earthquakes, 72 hours is the optimal time-limit for rescue, but the intervention of external forces will often take longer than this, resulting in further losses of life and property. ^{Reference Peng, Tan, Lin and Xu36} Therefore, knowing how to strengthen the resilience of communities to disasters is crucial. However, at present, the building of disaster resilience in Chinese communities is still at the initial stage, and much research is needed. At the same time, there are only a few studies on community disaster resilience in China, although Cui et al. ^{Reference Cui, Han and Wang14} made some useful explorations on the basis of Joerin et al., ^{Reference Joerin, Shaw, Takeuchi and Krishnamurthy37} Orencio and Fujii, ^{Reference Orencio and Fujii30} Ostadtaghizadeh et al., ^{Reference Ostadtaghizadeh, Ardalan and Paton11} measuring community disaster resilience in China’s earthquake-stricken areas from 5 dimensions: connection and caring, resources, transformative potential, disaster management and information and communication, and exploring the relationship between community disaster resilience and disaster risk reduction. However, the research area only focuses on 1 county, and whether the community disaster resilience measurement index is still feasible for the vast earthquake-stricken areas in China still needs to be further verified. In addition, the correlation between community disaster resilience and residents’ disaster risk perception has not been explored in China. As individuals in the community, residents are the direct victims of the disaster and the direct participants in the post-disaster reconstruction. The strength of community disaster resilience is bound to affect the residents’ disaster risk perception level, and then affect their decision of disaster prevention and reduction. Therefore, it is necessary and extremely important to explore the correlation between community disaster resilience and residents’ disaster risk perception, and relevant studies are urgently needed.

Disaster risk perception refers to residents’ subjective evaluations and judgments of the risk of events, as well as their associated attitudes and decision-making tendencies. It covers the whole process of perception, understanding, memory, evaluation, and response to risks. ^{Reference Terpstra38,Reference Xie and Xu39} At present, the development of community disaster resilience is insufficient. One important reason is residents perceive disaster risk to be low. ^{Reference Birkholz, Muro, Jeffrey and Smith40,Reference Lo and Chan41} The disaster risk perception of residents in disaster-threatened areas and its driving mechanisms have been a focus of academic research. Existing studies have mostly examined residents’ disaster risk perception in terms of personal characteristics (eg, gender, age, level of education), disaster experience (eg, experience of disasters, disaster severity), hazard proximity and hazard education programs, ^{Reference Brody, Highfield and Wilson42–Reference Rohrmann47} and household socioeconomic characteristics (eg, income, population, whether to have children, older people, building structure). ^{Reference Bubeck, Botzen and Aerts48–Reference Xu, Yong and Deng54} This provides a basis for us to understand the disaster risk perception of residents in disaster-threatened areas. However, few studies have focused on the impacts of community disaster resilience on residents’ disaster risk perception.

2.2 Research Hypotheses

Community disaster resilience and residents’ disaster risk perception and its influences have been a focus of academic research. Researchers believe that the relationship between community disaster resilience and disaster occurrence is complex and involves many social, economic, political and physical factors, and is characterized by temporal and spatial changes. ^{27,Reference Yang35,Reference Adger55–Reference Tobin57} For communities with high overall community disaster resilience, residents may have “survivor bias,” ^{Reference Xu, Yong and Deng54,Reference Lo and Cheung58,Reference Pfefferbaum, Pfefferbaum and Zhao59} and believe that their community has strong disaster resilience in all aspects and, thus, the residents have a relatively low level of disaster risk perception. For example, residents generally believe that disasters are unlikely to happen and not be too severe, so they are not very concerned about them. Based on this, the study proposes research hypothesis H1.

The stronger the community connection and caring, the greater residents’ sense of belonging. Residents are full of hope for the future development of the community. In daily life, mutual helping and a harmonious community environment can further reduce residents’ disaster threat perception and worry. ^{Reference Cui, Han and Wang14} High-magnitude earthquake disasters have low frequency, so residents generally think their possibility of occurrence is relatively low. For residents living in communities where residents help each other and neighbors live close to each other, disaster-related information will be timely and accurate, which is conducive to a better understanding of earthquake disasters. As a result, the more connected the community, the less likely are residents to think that a disaster is likely to occur. Based on this, we propose research hypothesis H2.

The better the resource endowment of the community, the better the community is able to solve the problems faced by development. There are special support measures for families, and resource-rich communities can reduce residents’ awareness of the threat and worry of disasters. ^{Reference Aldrich, Aldrich, Oum and Sawada60,Reference Norris, Stevens and Pfefferbaum61} For communities with effective leadership and good welfare policies, even in the event of a natural disaster, residents can unite to deal with the losses caused by the disaster. As a result, residents consider disasters less likely. Based on this, we propose research hypothesis H3.

Strong community transformative potential, planning for community development, and the ability of residents to negotiate and solve problems can further reduce residents’ perception of the possibility and threat of disasters. ^{Reference Cassar, Healy and von Kessler62} With the continuous improvement of community infrastructure, residents believe that the threat of disasters is relatively low. Communities with a consultative democracy that can pool the wisdom and efforts of all residents have the ability to guarantee the safety of life and property after a disaster. Therefore, the higher the transformative potential of community, the lower the worry of disaster. Based on this, we propose research hypothesis H4.

The better a community’s disaster management system, the better the group monitoring and mass prevention system (disaster evacuation and relocation plan). Communities with positive disaster preparedness measures can reduce residents’ perception of the possibility of disaster and their worry. ^{Reference Zhang, Yi and Zhao63} With the advancement of science and technology, communities can receive early warning within tens of seconds before the onset of a major earthquake, and residents believe that their threat is relatively low. Therefore, the more comprehensive the community’s disaster management system, the lower the residents’ perceived threat of disaster. Based on this, we propose research hypothesis H5.

Community information and communication are smooth, and residents are told relevant disaster prevention information and scientific disaster prevention measures by means of media. Moreover, residents believe that the information released by the community is authoritative, which further reduces their disaster awareness and worry (Tierney et al. 2006). ^{Reference Tierney, Bevc and Kuligowski64} Unimpeded communication of information to communities is conducive to residents’ comprehensive and scientific understanding of disasters. As a result, residents of communities with better access to information may perceive disasters to be less likely. Based on this, we propose research hypothesis H6.

3.1 Data Sources

The objective of this study is to explore the correlation between residents’ disaster risk perception and community disaster resilience. Considering the typicality and representativeness of the Longmen mountain seismic belt, the research group conducted a questionnaire-based survey and interviews in the 4 counties of the seismic belt in July 2019. The surveys and interviews were conducted on one-on-one basis in households, and the average length of each questionnaire was about 1.5 hours. The questionnaires and interviews investigated farming families’ situations, residents’ disaster risk perception, and community disaster resilience. The sample selection was mainly determined by stratified equal probability random sampling. The specific sampling process is detailed in Xu et al. ^{Reference Xu, Qing and Deng3–Reference Xu, Zhou and Deng5,Reference Xu, Liu and Deng18,Reference Xu, Yong and Deng54} After data cleaning, 327 valid questionnaires were obtained—an effective recovery rate of 97%. See Figure 1 for maps of the sample county and town locations.

Figure 1. Map of sample county and town location.

3.2 Selection of Model Variables

Community disaster resilience is the core independent variable of this study. By referring to Cui et al., ^{Reference Cui, Han and Wang14} Cutter et al., ^{Reference Cutter, Burton and Emrich65} Han et al., ^{Reference Han, Ba and Xin66} Pfefferbaum et al., ^{Reference Pfefferbaum, Pfefferbaum and Nitiéma67} and Ungar, ^{Reference Ungar68} community disaster resilience was divided into the following 5 dimensions in this study: connection and caring, resources, transformative potential, disaster management, and information and communication. To better reflect the situation in the study region, a total of 24 indicators were used (Table 1). The Cronbach’s alpha test values of connection and caring, resources, transformative potential, disaster management, information and communication, and total community disaster resilience are 0.77, 0.68, 0.82, 0.79, 0.72, and 0.92, respectively. It shows that the internal consistency of the community disaster resilience measurement index adopted in the study is good and can be further analyzed.

Table 1. Definition and descriptive statistics of the model variables

Notes:

^a 1 = totally disagree, 2 = disagree, 3 = neutral, 4 = agree, 5 = totally agree

^b SD = standard deviation.

Disaster risk perception was the dependent variable used in this study. As for the measurement of disaster risk perception, referring to the studies of Chandra et al. ^{Reference Chandra, Williams and Plough69} and Xu et al., ^{Reference Xu, Liu, Wang, Tang and Liu2,Reference Xu, Liu and Deng18,Reference Xu, Yong and Deng54,Reference Xu, Peng and Su70} this study divided this variable into 3 dimensions: possibility, threat, and worry. Some terms were designed to measure each dimension (Table 2).

Table 2. Earthquake disaster risk perception measurement

Notes:

^a 1 = totally disagree, 2 = disagree, 3 = neutral, 4 = agree, 5 = totally agree

^b SD = standard deviation.

In order to improve the explanatory power of the model, referring to the studies of Lawrence et al., ^{Reference Lawrence, Quade and Becker71} Lo and Cheung, ^{Reference Lo and Cheung58} Peng et al., ^{Reference Peng, Lin and Liu72,Reference Peng, Xu and Wang73} Xu et al., ^{Reference Xu, Peng and Su70} and Yu et al., ^{Reference Yu, Cruz and Hokugo74} variables that may affect residents’ disaster risk perception were added as control variables. These mainly included the socioeconomic characteristics of respondents and their families, including age (years), gender (0 = male, 1 = female), years of education (years), nationality (0 = else, 1 = Han), occupation (0 = farmer; 1 = else), residence time (years), income (annual cash income of farmers’ families in 2018, Yuan), housing structure (0 = else, 1 = civil structures), and so forth.

3.3 Analytic Strategy

As the dependent variable, disaster risk perception was the result of factor analysis; it is the computation of scale scores by summing over multiple items that produce a symmetric, unimodal distribution that is likely to approximate a normal distribution. Based on the distribution characteristics of the dependent variables, ordinary least square (OLS) was used to explore the correlation between community disaster resilience and residents’ disaster risk perception. The simple expression of the model is as follows:

$${{\rm{Y}}_i} = {\alpha _0} + {\beta _{1i}} \times C{R_i} + {\beta _{2i}} \times Contro{l_i} + {\varepsilon _i}$$

In the formula, ${\rm{\;}}{{\rm{Y}}_i}$ represents residents’ disaster risk perception, which can be divided into 4 indicators: possibility, threat, worry, and overall disaster risk perception; ${\rm{\;}}C{R_i}$ represents community disaster resilience, which can be divided into 6 indicators: connection and caring, resources, transformative potential, disaster management, information and communication, and overall community disaster resilience. ${\rm{\;}}Contro{l_i}$ represents the control variables; ${\rm{\;}}{\alpha _0}$ , ${\beta _{1i}}$ , and $\;{\beta _{2i}}$ represent the model parameters to be estimated; ${\varepsilon _i}$ is the residual term. All data analysis processes use SPSS 23 (IBM Corp, Armonk, NY).

3.4 Data Characteristics

In terms of the personal characteristics of the interviewees, the male to female ratio was about equal (54% male), ages were mainly middle and old (mean = 53.44 years), and the average length of education was low (mean = 6.29 years). In terms of the characteristics of the respondents’ building structure, the dwellings were made of reinforced concrete (48%) or brick and tile (37%), with the remaining 15% being a civil structure, which means a house made of wood and clay. In addition, the average annual cash income of the respondents was 66 191 Yuan.