1Vulnerability to water-related issues disproportionately affects poor urban settlements and their residents across the Global South. These communities frequently face a triad of challenges: scarcity, pollution and contamination of water sources and flooding (De Risi et al., 2013). Rapid urbanisation pushes marginalized groups into hazard-prone peripheries, where substandard housing and high population densities amplify exposure. By 2050, more than three billion people are expected to live in such settlements (UN-Habitat, 2024), with vulnerability projected to intensify under accelerating climate change (Pörtner et al., 2022).
2Flooding already affects 587 million people living in impoverished conditions worldwide (Rentschler, Klaiber and Vun, 2021). Poverty not only amplifies vulnerability but also constrains resilience capacities (Williams et al., 2019). In Indonesia, this convergence of exposure and marginalisation is particularly acute. The country faces multiple hydrological threats – seasonal pluvial flooding from intense rainfall, fluvial flooding from river overflow, and coastal inundation driven by tides, storm surges, tropical cyclones, and sea level rise – that disrupt millions of lives each year (Mizutori and Guha-Sapir, 2020). With 76 million Indonesians living in high-risk flood zones (Rentschler, Klaiber and Vun, 2021), and widespread land subsidence compounding risks, the need for robust resilience strategies is pressing.
3Most flood-affected Indonesians live in poverty (subsisting on less than $5.50 daily) or extreme poverty (less than $1.90 daily), with a significant proportion residing in slums (Rentschler and Salhab, 2020). This combination of physical exposure and socioeconomic marginalisation creates compounding vulnerabilities that demand urgent research attention and intervention in the area of flood resilience. For several decades, scholarship has examined flood resilience, including in Global South contexts. Increasingly, attention has turned to the role of wellbeing, often conceptualised in material and subjective terms.
4Yet, the relational dimension of wellbeing – defined as the quality of people’s relationships with others, institutions, and environments – remains underexplored, particularly in informal settlements or slums. Relational wellbeing, developed by White and Jha (2015; 2016; 2018; 2020; 2021; 2023) through studies in Bangladesh, India and Zambia, offers an alternative to individualist wellbeing frameworks by emphasising social bonds, contextual sensitivity, and structural inequities (White, 2017). However, applying this more culturally attuned framework in Indonesia requires careful engagement with local worldviews and understandings of community, resilience, and wellbeing.
5In slums, where residents often face material deprivation, assessments of wellbeing must be contextually grounded. Examining how relational wellbeing shapes resilience in flood-prone areas is essential for understanding why people remain in precarious living conditions and for designing more effective flood risk mitigation and adaptation strategies. At the same time, such analysis must recognise the cultural specificity and potential limitations of these conceptual categories.
6Our research examines the relationship between relational wellbeing and household resilience based on empirical data collected in three Indonesian cities: Bima, Manado and Pontianak. We analyse survey data from the Resilient Indonesian Slums Envisioned (RISE) project conducted in 2021 (Setiawan et al., 2022), focusing on water management and relational wellbeing. This research was funded by The Netherlands Organization for Scientific Research (NWO,file number 482.20.50L Merian Fund Ristek BRIN), NWO-WOTRO: Merian Fund Cooperation Indonesia –The Netherlands Regional Planning and Sustainable Urbanization
7We acknowledge that the term slum is contested, carrying negative colonial connotations and often failing to reflect local understandings of space and community. In Indonesia, the term urban kampung (‘village’) is often preferred (Octifanny & Norvyani, 2021), Nonetheless, we use the term slum for several reasons. Both formal and informal housing are present in our study contexts, making “informal settlements” insufficient. Besides, slum-like conditions are frequently found in neighbourhoods across Indonesia that are not necessarily designated as such and vice versa. While the Indonesian government designated slum areas through decree as kawasan kumuh (literally ‘dirty area’), in our opinion a derogatory term, we opt for using the term slum, which should no longer bear negative connotations. In fact, slums across the world have been praised for their ability to adopt new practices, as examples of sustainability and for their communal support (Angelo & Wachsmuth, 2020).
8Our investigation is guided by the research question: to what extent does relational wellbeing relate to household flood resilience in Bima, Pontianak and Manado? The remainder of this article proceeds as follows: Section 2 develops our theoretical framework on relational wellbeing, resilience, and their intersection; Section 3 outlines our methodology and analytical approach; Section 4 presents our empirical findings; and Section 5 offers discussion and conclusion.
9Over recent decades, the concept of resilience has received considerable academic attention, extending into policymaking on disaster risk management. Originally emerging from ecological science, resilience referred to the ability of ecological systems affected by disturbances to absorb changes and to recover (Holling, 1973). In recent years, the concept has been widely applied across disciplines, giving rise to terms such as individual, organisational and community resilience (Kendra, Clay & Bill, 2018), psychological resilience (Schwarz, 2018) and household resilience (Nguyen & James, 2013).
10Specifically, flood resilience is commonly defined as ‘the ability of the system to recover from floods’ (Bruijn, 2004: 199). Despite this proliferation, few studies have empirically examined the relationship between resilience and wellbeing, particularly in the context of floods. One notable exception is the work of Walker-Springett, Butler and Adger (2017), who explore the process through which wellbeing is affected in the aftermath of flooding. Their findings highlight how community and institutional responses can jointly support wellbeing and resilience, emphasizing the centrality of social relations and collective action in proactive resilience building. However, their study does not formally test the association between defined measures of wellbeing and flood resilience, nor does it engage with non-Western conceptualizations of these phenomena. Moreover, the study is based on a UK dataset, limiting its applicability to contexts with different social and cultural structures.
11Building on these perspectives, this article examines the interplay between relational wellbeing and household flood resilience in Indonesia, while remaining attentive to the epistemological challenges inherent in these concepts and their cultural relevance. Understanding resilience in this context requires moving beyond generic, Western-centric models to consider the social relations, community norms, and local governance structures that shape how Indonesian households experience, respond to, and recover from floods.
12In this study, we argue that floods are likely to impact all of a family unit and, therefore, we opt for household resilience as the dependent variable, rather than individual resilience. This choice aligns with Indonesian cultural emphases on collective rather than individual identity. Namely, in Indonesian society, household structure holds particular significance, with members depending heavily on mutual cooperation (Mangundjaya, 2013). Traditional family structures are held in high regard, wherein individuals rely and often depend on each other’s cooperation. Therefore, we do not hold only individual family members accountable as the sole source of household resilience but also include their functioning as a household. Since the household constitutes the foundational unit within social systems, understanding how residents perceive their household's recovery capacity becomes crucial (Takwin et al. 2023).
13Floods impact households in various ways. Not all impacts manifest directly, immediately, or tangibly, making comprehensive assessment challenging (Parker, 2000). In the literature, household resilience is generally measured objectively through aspects such as material wealth and assets as the percentage of households having access to clean water and distance to health centres (Jones and Samman, 2016). This objective measurement overlooks the nuanced effects of floods on wellbeing, particularly among those already living in precarious circumstances, such as slums.
14In response, researchers including Nguyen and James (2013) and Sekulova and van den Bergh (2016) have begun examining floods from a more subjective (psychological and emotional) perspective, broadening our understanding of resilience beyond material recovery. While an objective approach to wellbeing looks at physical and economic indicators to determine a household’s ability to adapt in disasters, it does not include evaluations of the individuals involved as to their household’s ability to function when exposed to hazards. Generally, people have a good understanding of the factors that may contribute to their household’s ability to cope with and recover from floods. This subjective approach to resilience, therefore, assesses an individual’s cognitive and affective self-evaluation when responding to risk. In this study, we take a more subjective approach in evaluating household resilience to floods. In addition, considering subjective household resilience allows individuals to think about their immediate inter-personal context, that is their own household (Jones & Tanner, 2015).
15While household resilience captures collective recovery processes, it is essential to recognize that gender and social position strongly mediate how resilience is experienced and enacted within families and communities. Research on gendered disaster experiences has shown that women and men often face distinct vulnerabilities and capacities in flood contexts due to differentiated access to resources, social capital, and decision-making (Enarson & Chakrabarti, 2009; Fordham, 2013). Moreover, intersectional analyses reveal that these differences are further shaped by class, age, and ethnicity, which together influence both exposure to risk and pathways to recovery (Kaijser and Kronsell, 2014). In Indonesia, gender norms and household hierarchies may therefore influence how relational wellbeing contributes to flood resilience. Attending to these intersectional dynamics enables a more inclusive understanding of resilience as socially differentiated rather than uniformly distributed.
16Nguyen and James (2013) have been among the few to operationalize and investigate the relationship between wellbeing and household resilience, employing a subjective wellbeing approach to measure households’ resilience to floods in Vietnam’s Mekong River Delta. The authors demonstrated this bidirectional relationship: floods affect both the material- and subjective dimensions of wellbeing, which in turn shape households’ flood resilience. They developed measurements that relate to a household’s capacity to cope with, adapt to, and benefit from floods and carried out a factor analysis to explain different properties of households’ resilience to floods.
17Three dimensions were identified: (1) households’ confidence in securing food, income, health, and evacuation during floods and recovery after floods; (2) households’ confidence in securing their homes and not being affected by a large flood event; (3) households’ interests in learning and practicing new flood-based farming practices that are fully adapted to floods for improving household income during the flood season (Nguyen and James, 2013). While their work represents an important contribution, it remains grounded in Western psychological frameworks that may not fully capture the relational dimensions of wellbeing emphasized in Southeast Asian cultures.
18The concept of wellbeing has gained popularity in the field of international development as a positive alternative to deficit-focused approaches that emphasize poverty, vulnerability and risk. By shifting attention from shortcomings to strengths, wellbeing frameworks challenge the stigmatization of vulnerable groups and counter the “othering” often reinforced through policy labels. Moreover, the wellbeing approach adopts a holistic outlook, at once personal and policy-oriented, placing individuals, their priorities and their perspectives at the core (White, 2008). However, dominant wellbeing frameworks often reflect Western, individualistic assumptions that may not fully resonate with Indonesian worldviews. In this study, we define wellbeing in simple terms as a person’s self-assessment of their life conditions (Setiawan et al., 2022).
19Wellbeing is commonly understood to encompass two interrelated dimensions: material- and subjective (i.e. psychological) wellbeing. Material wellbeing refers to people’s access to income, wealth, housing and work conditions sufficient to meet basic needs and enable valued capabilities. It is typically measured objectively through household income, consumption and wealth indicators. Subjective wellbeing, by contrast, concerns individuals’ evaluations and experiences of their lives, often measured through life satisfaction, affect balance, and a sense of meaning or purpose (OECD, 2020).
20More recently, scholars, such as Atkinson (2013; 2020), Calestani (2013), Eckersly (2009), Fischer (2014), Jackson (2011) and Thin (2012), emphasize the need to include relational wellbeing as the third dimension. Eckersley (2009), for example, argued that the material dimension of wellbeing alone is an imperfect measure of human wellbeing. Further, the subjective dimension of wellbeing by itself is a dubious, comparative measurement at best, because cultural differences and other attributes and attitudes also need to be taken into consideration. This development represents a significant step toward recognizing the social embeddedness of wellbeing.
21Notably, White (2015; 2017) has developed and refined the concept of relational wellbeing over the years, empirically grounded on research in Bangladesh, India and Zambia. Together with Global South scholars, including Jha, White has co-authored several papers on how relational wellbeing emerges from an interplay between personal agency, social obligations, moral responsibilities, and the environmental context. Their work moves from grounded Global South case studies to a systematic theoretical framework that highlights this interdependence (White and Jha 2016; 2018; 2021; 2023). Together, White and Jha founded the Relational Wellbeing Collaborative in 2020.
22White (2015) argues that research approaches solely focusing on material and subjective dimensions are epistemologically positivist, treating people as objects rather than subjects within networks of relationships. Building on this critique, White (2017; 2018), as well as White and Jha (2016; 2021) advocate for a move away from individualistic approaches toward relationship-grounded understandings. According to White (2018, p. 9), wellbeing is inherently relational, because people ‘identify wellbeing in collective terms’. In other words, wellbeing is best understood as a shared, social, and collective experience, not just an individual one. White and Jha demonstrate the importance of adopting a relational approach to wellbeing to achieve a more encompassing and culturally sensitive understanding of wellbeing, particularly in non-Western societies like Indonesia, while still requiring careful adaptation to local realities.
23We contend that wellbeing should not be reduced to wealth or happiness alone, nor to their simple combination. Instead, wellbeing is better understood as a personal attainment embedded within the relationships a person inhabits. White and Pettit (2004) propose a relational approach to wellbeing, viewing it not as a fixed outcome but as an ongoing process shaped by personal, environmental and societal dimensions. Building on this, White (2017) emphasizes that relationships are not merely external supports or constraints, but constitutive elements of wellbeing itself. In operationalizing this perspective, White (2010) proposes three dimensions of relational wellbeing: the material (e.g., income and assets), the subjective (e.g., personal growth), and the relational or social (e.g., family and community involvement).
24While this framework provides a foundation for measuring relational wellbeing, we question its contextual appropriateness for Indonesian contexts. The distinctive value of relational wellbeing emerges from its sensitivity to specific contexts, differentiating it from individualistic wellbeing approaches (White, 2015). Examining our Indonesian research sites requires consideration of three critical contextual factors. First, religiosity has reemerged as a dominant influence in Indonesian society over recent decades (Hadiz, 2017) and must therefore be incorporated within our analytical framework. Second, recognising that slums often develop through chain economic migration along family or ethnic ties (Chaudhuri, 2015), we have included aspects of neighbourhood security and lack of economic competition in our assessment. This addition acknowledges that diverse social groups cohabiting in resource constrained environments may (or may not) experience competitive dynamics, particularly where employment opportunities are limited (Pertiwi & Faturochman., 2023). Third, due to context-embedded identity, self-concept was necessary to be added to the subjective dimension. Self-concept identifies how individuals perceive and evaluate themselves, which is key in explaining personal aspirations in relational wellbeing (White, 2015).
25In relation to wellbeing, household resilience is assumed to be linearly associated. In detail, individuals with a positive perception of their abilities are likely to rise to the challenge when floods occur. They are likely to put trust in their ability to secure the essentials for their family and, most importantly, can cooperate with other family members. Further, those who have a positive opinion of their physical infrastructure and public facilities are likely to show a higher level of household resilience. Based on their past experiences, they feel able to trust that their environment will help them recover from floods. Finally, those who have positive relationships with others are likely to show reliance through their social network and vice versa. Even during hardships, these people can share belongings with the whole community, as White (2015) observed in Bangladeshi communities. Thus, relational wellbeing is a necessary factor in determining individuals’ household resilience.
26Based on the explanation above, we aim to investigate to what extent the relational wellbeing of people living in Bima, Manado and Pontianak relates to their household’s flood resilience while remaining attentive to the cultural specificity of these relationships. We build on White’s three dimensions of relational wellbeing but operationalize them in six aspects adapted for Indonesian contexts. Family and community participation and security and lack of competition represent the relational dimension. Material satisfaction and social and facilities satisfaction represent the material dimension. Religiosity and self-concept represent the subjective dimension (see Riasnugrahani et al., 2024).
27We expect that these adapted relational wellbeing aspects are positively associated with household resilience. As such, we hypothesize that family and community participation (H1a), security and lack of competition (H2a), material satisfaction (H3a), social and facilities satisfaction (H4a), religiosity (H5a) and individuals’ self-concept (H6a) are positively correlated to confidence in ability to secure food, income, health and safe evacuation as components of household resilience to floods. Similarly, we hypothesize that all six relational wellbeing aspects are positively associated with households’ confidence in securing their homes during a large flood event (H1b-6b) and their interest in learning and implementing flood-based livelihoods (H1c-6c).
28This study primarily used a dataset collected in an extensive survey in 2021 as part of the RISE project (2021-2022) on water management and relational wellbeing. The documentation associated with this data collection has been published in Data Archiving and Networked Services (DANS) and can be found publicly online (see Setiawan et al., 2022 for a more detailed description). Here, we briefly explain the data collection procedure and the measures employed in our analysis.
29We purposely selected the three Indonesian cities of Pontianak, Bima and Manado based on two criteria. First, an extensive research between Indonesian National Development Planning Agency, Swiss Confederation and the World Bank on ‘Integrated Urban Flood Risk Management’ in 2018 (internal report) identified these cities as facing a heightened risk of flooding and containing both formal and informal settlements lacking proper sanitation and flood protection. These cities represent a variety of geographic locations and flood risk characteristics, urban growth, climate change impact and hazard types and levels. Second, the three cities have experienced major floods in recent years but have been largely overlooked by researchers studying flood impacts (Rentschler, Klaiber and Vun, 2021).
30Subsequently, we then purposely selected districts and sub-districts in each city based on rapid field assessments. We started selecting districts by consulting mayoral decrees identifying areas considered to be slums in their corresponding cities (The Mayor’s decree of Pontianak; The Mayor’s decree of Bima; and the Mayor's decree of Manado). Within those sub-districts, we selected two or three neighbourhood communities to be part of our random sampling selection of households. Table 1 below lists all the sub-districts included in the three cities:
Table 1. Selected (sub-)districts.
|
City
|
District
|
Sub-district
|
|
Pontianak
|
Pontianak Timur
|
Tambelan Sampit
|
|
Pontianak Barat
|
Sungai Jawi Luar
|
|
Pontianak Tengah
|
Tengah Pontianak Kota
|
|
Bima
|
Rasanae Barat
|
Paruga
|
|
|
Sarae
|
|
Manado
|
Sario
|
Titiwungen Utara
|
|
Singkil
|
Wawonasa
|
31The data collection in Indonesia was conducted between November 2021 and February 2022. We obtained permission to conduct data collection from the Directorate General of Politics and General Administration, Ministry of Home Affairs of the Republic of Indonesia in 2021 (470.02/7428/Polpum) and the ethical approval was granted by the Research Ethics Committee of Universitas Indonesia (No.011/FPsi.Komite Etik/PDP.04.00).
32We selected households using a random walk method, a systematic sampling approach that applies probabilistic selection at regular intervals, to obtain a random sample of adults (aged 18 and above) who had lived in the area for a minimum of three years. When there was more than one adult in a household, we simply asked who had the nearest birthdate to the survey date as a way to randomly select a respondent among the qualified adults. In conducting the survey, we carefully followed ethical considerations by informing the respondents, before their participation, about the study and giving them the option to voluntarily participate in the study or refuse. People who agreed to participate were asked for their written consent. After the study, the participants were given Rp. 50,000 (approximately €2.60/$3) as a small token of appreciation.
33In total, the survey involved 700 respondents from the 920 approached. There were 262 males and 438 females, with an average age of 43 (standard deviation SD is 11.8). Of the people involved, 300 respondents were from Pontianak (150 males and 150 females), 200 from Bima (46 males and 154 females) and 200 from Manado (66 males and 134 females). Table 2 provides descriptive statistics for all the measures.
Table 2. Descriptive statistics of all variables
|
Predictors
|
Range
|
Bima
|
Pontianak
|
Manado
|
F test
|
|
M
|
SD
|
M
|
SD
|
M
|
SD
|
|
|
Confidence in ability to secure food, income, health and safe evacuation
|
1-6
|
3.84
|
0.72
|
3.20
|
0.53
|
3.54
|
1.08
|
F(2,697)=41.80
|
|
Interest in learning and implementing flood-based livelihoods
|
1-6
|
4.23
|
0.70
|
4.24
|
0.60
|
3.83
|
1.03
|
F(2,697)=20.14
|
|
Family and community participation
|
1-6
|
4.73
|
0.56
|
5.08
|
0.21
|
5.07
|
0.71
|
F(2,697)=34.67
|
|
Security and Competition
|
1-6
|
2.86
|
1.06
|
2.47
|
0.43
|
2.67
|
1.71
|
F(2,697)=7.18
|
|
Material satisfaction
|
1-6
|
3.95
|
0.66
|
3.38
|
0.62
|
4.27
|
0.86
|
F(2,697)=101.21
|
|
Social and facility satisfaction
|
1-6
|
4.40
|
0.67
|
4.70
|
0.55
|
4.86
|
0.78
|
F(2,697)=24.96
|
|
Religiosity
|
1-6
|
4.46
|
0.73
|
4.78
|
.43
|
4.96
|
.80
|
F(2,697)=29.57
|
|
Self-concept
|
1-6
|
4.08
|
0.60
|
4.25
|
.47
|
4.80
|
.71
|
F(2,697)=84.23
|
|
Individual characteristics
|
|
|
|
|
Age
|
18-76
|
45.54
|
12.71
|
40.92
|
10.63
|
43.60
|
11.99
|
F(2,697)=9.80
|
|
Gender (female as reference =0)
|
0 or 1
|
.23
|
.42
|
.50
|
.50
|
.33
|
.47
|
-
|
|
Educational level
|
1-10
|
4.12
|
1.76
|
3.73
|
1.74
|
3.71
|
1.23
|
F(2,697)=4.30
|
|
Individual income
|
1-10
|
2.31
|
1.68
|
3.79
|
2.11
|
7.99
|
18.56
|
F(2,697)=17.49
|
Note: Bold indicates significance at the p<.05 level.
34In our validity test, we conducted Confirmatory Factor Analysis (CFA) to rigorously assess the structure of our measures, particularly focusing on the relational wellbeing construct. We relied on previous studies for the configuration of relational wellbeing dimensions and their items (Riasnugrahani et al., 2024). We performed CFA following specific criteria: (1) maximum likelihood (ML) estimation was used to verify the relationships among factors and the arrangement of the measured indicators or items; (2) we assessed the model's goodness of fit using a combination of fit indices, namely the Comparative Fit Index (CFI) and the Standardized Root Mean Squared Residual (SRMR). CFI compares the fit of a proposed model to a baseline model (where no relationships among variables are assumed), whereas SRMR assesses the standardized difference between the predicted and actual correlations in a proposed model (Hu & Bentler, 1999); (3) we established a threshold for factor loading coefficients at a minimum of 0.40. We performed our analyses in RStudio using a lavaan package.
35Our dependent variables are dimensions of household resilience to floods, adapted from the measure developed by Nguyen and James (2013) for Vietnamese communities in flood-prone areas of the Mekong River Delta. This measure consists of nine items across three dimensions: confidence in securing food, income, health and safe evacuation during floods; confidence in maintaining flood-resistant homes; and interest in learning and implementing flood-based livelihoods. Participants rated statements such as “I can repair my house quickly” and “I have been carrying out activities to overcome a disaster” on a six-point Likert scale, with higher scores indicating greater agreement.
36The CFA shows that the original 3-dimension scale is better fit for a 2-dimension scale, χ2 = 224.04, p < .001, CFI = .94 and SRMR = .04. Although the chi-square test is significant, which is common with large samples, we can still assume a good fit for the 2-dimension model. The CFI value shows that the proposed model is a better fit compared to the baseline model, and the low value of SRMR (=.04) indicates that our model’s predictions reflect a strong agreement between the model and the actual observations. The dimensions of confidence in securing food, income, health and safe evacuation during floods, and confidence in securing flood-resistant homes are merged into one with a total of five items.
37The second dimension, interest in learning and implementing flood-based livelihoods, consists of two items. In total, we end up with seven items. The convergent validity of each dimension is at an acceptable level (see Table 3), and the items’ factor loading range from .57 to .90. The composite reliability is at .92 for the first dimension and .58 for the second one, suggesting good consistency among items in each dimension. The latter's internal consistency is slightly lower than the proposed level, but this is reasonable considering that there are only two items. When looking at the Pearson correlation between the two items, to replace Cronbach's alpha, we have a medium level of correlation among them (r=.41).
38Further, based on Table 3, there are moderate to strong positive correlations interesting to discuss. Confidence to secure food, income and safe evacuation is shown to be positively correlated with subjective material wellbeing (r=.56), similar to a positive correlation between family and community participation with age (r=.48). The findings are theoretically plausible, as being satisfied with one’s financial condition is required to be confident in one’s ability to secure food and basic necessities when flood strikes. Whereas the former follows a general pattern described in developmental psychology; as individuals age, they tend to be more eager to contribute to society and engage in communal activities (Santrock, 2018). Further, there are also negative correlations found in the analysis. For instance, family and community participation is shown to be negatively related to security and lack of competition (r=-.20), implying that there might be a trade-off between community activities and perceived job security among participants.
39Following White's (2010) theoretical framework, Riasnugrahani et al. (2024) developed the original scale of relational wellbeing consisting of 58 items. The scale is based on three primary components proposed by White, namely subjective, material and relational or social. It assesses individuals' evaluations towards their life across these dimensions. Specifically, the material component involves dimensions such as subjective evaluation towards their current financial situation (material wellbeing) and satisfaction with public facilities. The subjective dimension involves dimensions like religiosity and self-concept, while the relational dimension addresses dimensions such as family and community involvement, along with feelings of security and lack of competition (White, 2010; Riasnugrahani et al., 2024). In this scale, we asked participants to rate themselves on a six-point Likert scale on statements such as “I am able to prioritize my family without trying too hard” and “I often feel uneasy while living in this neighbourhood”, with a higher score indicating a stronger agreement with the statement.
40Based on the CFA results, the items were reduced to 33 across six dimensions: family and community involvement (8 items), security and lack of competition (6 items), subjective material wellbeing (5 items), social and public facility satisfaction (5 items), religiosity (4 items), and self-concept (5 items). The six-dimensional model demonstrated a good fit, χ² = 2755.69, p < .001, CFI = .89, and SRMR = .07. The CFI is slightly below the recommended threshold, but the SRMR value is still lower than the threshold (.08) and therefore, we can still assume that the proposed model is a better fit than the baseline or null model and that the difference between the model’s predictions and the actual observations are very small (Hu & Bentler, 1999).
41Hence, utilizing a combination of fit indices provides a more comprehensive assessment as each explains unique information of different aspects of model-data relations. Additionally, the average variance extracted (AVE) analysis in Table 3 indicates acceptable convergent validity level for each dimension. Factor loadings ranged from .51 to .96, and composite reliability for each dimension was high, ranging from .80 to .94.
42Demographic data collected included age, gender, educational level, and individual income. Age and gender were assessed via straightforward questions. Educational level was measured using a 10-point scale ranging from "Did not go to school" (1) to "Doctoral degree" (10), indicating the highest completed level. Individual income was assessed using a 10-point scale representing estimated average monthly income, ranging from "Less than Rp. 1,000,000 (€52 EU/$61)" (1) to "More than Rp. 10,000,000 (€520/$610)" (10).
43Similar to running CFA, we performed our path analyses in an open-source tool that provides an integrated development environment (IDE) named RStudio that enables us to write statistical codes and view the results using a proper available package, namely lavaan package. To ensure that we ran the analyses appropriately, we previously performed the following tests: Normality, linearity, and variance inflation factor (VIF). These preliminary tests are required to perform CFA that uses maximum-likelihood estimation. Normality and linearity ensure that we have normal distribution and the relations between variables of interest are linear, allowing for unbiased estimates and correct standard errors (Sladekova et al., 2024). Meanwhile, VIF ensures that there are no overlapped variables (two variables measuring similar constructs) which can make the estimation unstable or the measurement model become unidentified (Field, 2009).
44First, we checked for normality based on skewness and kurtosis measures. For all the predictors and dependent variables, the obtained values of skewness ranged from -1.53 to 0.95 and for kurtosis from -1.92 to 5.14, indicating that the data was normally distributed with no overly deviated variation. Second, we ran linearity tests on the dependent variables. We found that most of the predictors were linearly related to the dependent variables. Finally, we calculated the variance inflation factor (VIF) to confirm that there were no multicollinearity issues in the scales. The tolerance values of all our predictors were within the range of 0.59 to 0.78 and their VIF values ranged from 1.33 to 1.71, suggesting there were no multicollinearity concerns. Further, these assertions were also reinforced by the values of the bivariate correlations between predictors and the dependent variables shown in Table 3; no overly high (i.e. > .80) correlations between the predictors.
45Subsequently, we ran path analyses on both dependent variables to test our set of hypotheses. We created dummy variables to represent the three cities, with Manado being the reference category. The reason for this was that Manado had the highest average income of all the three cities. City-dummy variables were used to represent mean city differences that could be potentially related to the dependent variables.
Table 3. Bivariate correlations among predictors
|
Variables
|
1
|
2
|
3
|
4
|
5
|
6
|
7
|
8
|
9
|
10
|
11
|
12
|
|
1. Confidence to secure food, income and safe evacuation
|
|
.42
|
.05
|
.35
|
.56
|
.05
|
.07
|
.22
|
.01
|
.10
|
.06
|
-.03
|
|
2. Interest in learning
|
|
|
.08
|
.28
|
.16
|
.15
|
.03
|
.07
|
.08
|
.01
|
-.03
|
.01
|
|
3. Family and community participation
|
|
|
|
-.20
|
.25
|
.36
|
.46
|
.39
|
.48
|
-.04
|
.08
|
.05
|
|
4. Security and lack of competition
|
|
|
|
|
.07
|
.11
|
-.13
|
.09
|
.02
|
.09
|
-.17
|
.01
|
|
5. Subjective material wellbeing
|
|
|
|
|
|
.20
|
.19
|
.42
|
.02
|
.11
|
.14
|
.06
|
|
6. Facilities satisfaction
|
|
|
|
|
|
|
.37
|
.38
|
.29
|
.09
|
-.04
|
.12
|
|
7. Religiosity
|
|
|
|
|
|
|
|
.47
|
.43
|
-.04
|
.17
|
.12
|
|
8. Self-concept
|
|
|
|
|
|
|
|
|
.31
|
-.03
|
.07
|
.15
|
|
9. Age
|
|
|
|
|
|
|
|
|
|
|
-.35
|
-.00
|
|
10. Educational level
|
|
|
|
|
|
|
|
|
|
|
|
.05
|
|
11. Individual income
|
|
|
|
|
|
|
|
|
|
|
|
|
|
AVE
|
.69
|
.41
|
.55
|
.82
|
.60
|
.69
|
.66
|
.49
|
|
|
|
|
*. Bold font indicates significance at the p<.05 level
46We first looked at the descriptive statistics for all the variables and their one-way ANOVA to investigate differences among the three cities. We did not conduct a post hoc test of the ANOVA results because this would only provide redundant information for our dummy variables. Table 2, shown earlier, includes the descriptive results and mean differences for all the variables.
47Regarding outcome variables, we found that there was a difference among the three cities in terms of their confidence in being able to secure food, income and safe evacuation, F(2,697)=41.80, p<.05. On average, the participants from Bima (M = 3.84, SD = .72) were the most confident in being able to secure food and income when compared to Pontianak (M = 3.20, SD = .53) and Manado (M = 3.54, SD = 1.08). This level of confidence may indicate greater or lesser resilience when coping with flood events. The result is particularly interesting given that Bima has the lowest individual incomes of the three cities. Further, there are also city differences in terms of their interest in learning and implementing flood-based livelihoods, F(2,697)=20.14, p<.05. That is participants from Bima (M = 4.23, SD = .70) and Pontianak (M = 4.24, SD = .60) were more interested than those from Manado. Similar to confidence, higher interest may reflect more adaptive responses in flood-prone settings, suggesting higher resilience.
48The path analysis examined hypotheses regarding household resilience (confidence in securing food, income and safe evacuation, and interest in flood-based livelihoods). Results (Table 4) indicated that security and lack of competition (b = .26, p = .00) and subjective material wellbeing (b = .55, p = .00) were significantly and positively related to confidence in securing food, income and safe evacuation (H2a and H3a supported). This suggests, for example, that when individuals experience an increase of one unit in security and lack of competition, their confidence in securing food and income is likely to increase by .26 units. Although these numbers may seem arbitrary, the finding suggests that there is a positive impact of job security on one’s confidence in securing essential needs when flood occurs.
49Conversely, social and facility satisfaction was negatively related to this household resilience dimension (b = -.11, p = .00). Unlike the positive relation that we found earlier, this finding suggests that higher satisfaction with social and facility aspects tends to render lower confidence in securing essential needs, plausibly due to their overreliance on social networks and available facilities. Consequently, H1a, H4a, H5a, and H6a, which posited positive relationships between other relational wellbeing dimensions (family and community participation, social and facility satisfaction, religiosity, and self-concept) and confidence in securing food, income and safe evacuation, were rejected.
50Next, we found that security and lack of competition (b = .19, p = .00), subjective material wellbeing (b = .24, p = .00), and social and facility satisfaction (b = .13, p = .00) were the only relational wellbeing dimensions positively associated with interest in learning and implementing flood-based livelihoods. In contrast, family and community participation, religiosity, and self-concept showed no significant effects. Consequently, our hypotheses predicting positive relationships between these dimensions and livelihood adaptation (H1b, H5b–H6b) must be rejected. These findings suggest that confidence in personal security, satisfaction with material conditions, and access to social and physical infrastructure provide more immediate incentives for livelihood innovation than broader communal or spiritual dimensions of wellbeing.
Table 4. Path coefficients of relational wellbeing dimensions on household resilience dimensions (standard errors in parentheses)
|
Predictors
|
Confidence in ability to secure food, income, health and safe evacuation
|
Interest in learning and implementing flood-based livelihoods
|
|
Family and community participation
|
.04 (.06)
|
.06 (.06)
|
|
Security and lack of competition
|
.26 (.03)
|
.19 (.03)
|
|
Subjective material wellbeing
|
.55 (.04)
|
.24 (.04)
|
|
Social and facility satisfaction
|
-.11 (.04)
|
.13 (.05)
|
|
Religiosity
|
.04 (.04)
|
-.05 (.06)
|
|
Self-concept
|
.06 (.05)
|
.07 (.05)
|
|
Bima
|
.45 (.07)
|
.54 (.09)
|
|
Pontianak
|
.20 (.07)
|
.74 (.09)
|
|
Age
|
-.00 (.00)
|
-.00 (.00)
|
|
Gender (female as reference)
|
-.01 (.05)
|
-.14 (.06)
|
|
Educational level
|
-.01 (.02)
|
-.02 (.02)
|
|
Household income
|
.00 (.00)
|
-.00 (.01)
|
|
Adjusted R2
|
.46
|
.22
|
|
CFI=.91; SRMR= .03; χ2(8, N=700)=95.68, p<.000.
|
*. Bold indicates significance at the p<.05 level
51We continued by examining differences in household resilience across individual cities. Participants from Bima (b = .545, p = .00) and Pontianak (b = .20, p = .01) reported higher confidence in securing food, income and safe evacuation, despite having lower individual income levels than those in Manado. Additionally, Bima (b = .54, p = .00) and Pontianak (b = .74, p = .00) showed greater interest in learning flood-based livelihoods. Finally, among demographic factors, only household income was positively related to confidence in securing food, income and safe evacuation (b = .01, p = .04), though the effect size was small and may lack practical significance. Furthermore, being female was negatively associated with interest in learning flood-based livelihoods (b = -.14, p = .02).
Figure 1. Path analyses of relational wellbeing dimensions on household resilience dimensions.
Note: Bold indicates significance at the p<.05 level; Demographic variables are omitted for the sake of clarity.
52We conclude that some of our hypotheses on the relationships between dimensions of relational wellbeing and household resilience to floods are confirmed. On the one hand, we show that participants’ level of security and lack of competition (H2a&b) within the community and their subjective material wellbeing (H3a&b) can positively predict their overall level of household resilience to floods. On the other hand, participants’ social and facility satisfaction (H4b) positively predicts their level of interest in learning and implementing flood-based livelihoods.
53Our analysis highlights critical dimensions of household flood resilience, revealing the complex interplay between relational wellbeing, demographic factors and geographic context. Several key findings warrant discussion, both in light of our empirical results and existing scholarship. First, perceptions of security, low economic competition for livelihoods, and subjective material wellbeing emerged as positive predictors of household resilience. In this context, low competition indicates that households perceive abundant livelihood opportunities, reducing economic pressures and enabling them to focus on preparedness and recovery. This lessens economic stress and fosters greater confidence in securing food, income, and safe evacuation during floods. This resonates with research that highlights the importance of livelihood stability for adaptive capacity in disaster contexts (Cutter et al., 2008; Adger, 2010).
54At the same time, relational wellbeing fundamentally depends on perceived quality of social relations. Households that experience both livelihood security and supportive social environments feel more confident in their ability to respond to flood risks. These findings are consistent with Maslow’s hierarchy of needs, which identifies safety and material stability as prerequisites for adaptive behaviour (Mathes, 1981), and with self-determination theory (Ryan & Deci, 2000), which suggests that secure and supportive social environments allow individuals to fulfil psychological needs – autonomy, competence, and relatedness – that in turn foster proactive behaviours.
55Second, our results show that social and facility satisfaction negatively predicted confidence in securing food, income, health, and safe evacuation during floods. While strong social ties and communal infrastructure are often highlighted as positive predictors of resilience (Aldrich & Meyer, 2015), our findings complicate this narrative. They suggest that in certain contexts, reliance on communal facilities and neighbourly relations may foster a sense of complacency, leading households to underinvest in individual preparedness. Complacency here does not stem from cohesion per se, but from the perception that family, neighbours, or public services will reliably meet needs during a flood. This is significant because it challenges the prevailing assumption in the resilience literature that social capital is universally beneficial. It aligns with White’s (2015) assertion that wellbeing extends beyond individual control and can at times generate dependencies that undermine confidence when networks are simultaneously disrupted. This insight provides a critical nuance: while relational wellbeing is central to resilience, it may also contain vulnerabilities when overreliance replaces individual or household-level preparedness.
56Third, family and community participation, together with social and facility satisfaction, were exclusively associated with interest in learning and implementing flood-based livelihoods. These relational dimensions reflect perceived connections with family, access to public facilities, and opportunities for community engagement (White, 2015). Such forms of participation are crucial for transformative adaptation, where substantial shifts in livelihood strategies require collaboration and social support. This finding resonates with the literature on community-based adaptation, which emphasizes the role of collective action and cultural values in shaping adaptive pathways (Ensor & Berger, 2009; Djalante & Thomalla, 2012). However, our data also suggest limits: while these dimensions foster openness to learning new practices between flood events, they do not necessarily translate into confidence in securing basic needs during disruptive events. This indicates a division between everyday adaptive practices and crisis-time resilience, a distinction underexplored in resilience research.
57The geographic context further shaped resilience outcomes. Notably, households in Bima, despite having the lowest average individual incomes, reported higher confidence in securing food, income and safe evacuation than those in Manado and Pontianak. This points to the importance of macro-level variables, such as poverty levels, local governance, and infrastructural investments, which were not included in our models but are likely to exert significant influence (Wisner et al., 2004). It also suggests that resilience in informal settlements cannot be understood solely through individual- or household-level variables but must be situated within broader socio-political and economic contexts.
58A crucial dimension that emerges from this discussion is the role of power relations and local values. Flood resilience in Indonesian slums is not only a matter of resources or perceptions but also reflects hierarchies of power – between residents and authorities, between formal and informal governance systems, and within communities themselves. Research on slum life has shown how values such as gotong royong (mutual aid) and kekeluargaan (familial solidarity) shape everyday practices and responses to crises (Pribadi et al., 2014; Simone, 2018). Our findings confirm the centrality of these relational values, while also pointing to their ambivalence: they can enable collective action and recovery, but they may also reinforce reliance on others and limit individual agency. Moreover, residents of slums often operate under precarious political conditions, where their rights to land and services are insecure, and where power asymmetries vis-à-vis the state or private developers constrain adaptive choices. Any effort to enhance flood resilience must therefore engage with these power dynamics, recognising that adaptation is not only technical but also deeply political.
59This has implications for how resilience and wellbeing are measured. While subjective and relational indicators provide valuable insights, they risk obscuring the structural conditions that shape vulnerability if considered in isolation. Measurements that capture perceptions of security, participation, and satisfaction must therefore be interpreted in relation to broader inequalities, governance arrangements, and cultural norms. Our findings caution against overreliance on universalised resilience frameworks and point to the need for contextually grounded approaches that incorporate local epistemologies, such as adat (customary law) and community authority structures, alongside material and subjective dimensions of wellbeing.
60We acknowledge several limitations. Data collection in Bima was disrupted by the December 2021 flooding (Syarifudin, 2021), which may have affected responses. Follow-up research comparing locations under flood and non-flood conditions could help identify temporal shifts in wellbeing and resilience. Additionally, the exclusion of macro-level variables – such as regional minimum wages, poverty levels, NGO advocacy, and government involvement – constrains our explanatory scope. Future research should integrate these variables to provide multilevel explanations of resilience dynamics.
61Overall, our results contribute to ongoing debates on resilience and wellbeing by offering a nuanced understanding of the relational foundations of household flood resilience in Indonesian slums. They show that perceptions of neighbourhood safety, low competition for livelihoods, and material asset satisfaction correlate with confidence in family security and adoption of adaptive livelihoods. They also demonstrate that family and community participation, together with satisfaction in social relations and facilities, are linked to transformative adaptive behaviours. Yet, these relational assets can both strengthen and undermine resilience, depending on how they intersect with structural inequalities, power relations, and cultural values. This underscores the importance of moving beyond universal frameworks of resilience and wellbeing, towards contextually grounded, relational, and politically aware approaches that reflect the lived realities of households in flood-prone informal settlements.