Habitability and psychosocial vulnerability to floods in Escuinapa and Rosario

Habitability, Housing and Flood Risk

The habitability of housing exposed to flooding depends on the interaction between the hazard, the physical characteristics of the housing and its surroundings, and the social capacities of households. In settlements with insufficient infrastructure, the combination of precarious housing, limited basic services, and location in low-lying areas produces direct and indirect threats to the population (Hussainzad & Gou, 2024). The review by Bhanye (2025) on community adaptation to flooding in informal settlements of the Global South documented structural measures, such as drainage and plot reorganization, together with non-structural measures, such as early warning systems and community mapping, and concluded that their long-term effectiveness depends on their integration into formal risk management frameworks and on an equitable allocation of resources.

In the Sinaloan context, the available studies emphasize the urban and territorial component of exposure. Ibarra and Salazar Yanes (2024) indicated that floods in the state are related to deficient urban planning and drainage failures, and that household losses include the decrease in housing value, so they proposed a socio-environmental justice approach. Mancera-González and Rivas-Ochoa (2025), using an ethnographic approach in vulnerable neighborhoods of Culiacán affected by two tropical cyclones, showed that the population's permanence in flood-prone areas responds to economic and sociocultural factors. The Resistance of housing to rainfall dimension of the instrument studied captures precisely this facet: whether the housing resists rainfall (R6), whether water enters it (R7, inverse), whether it is located on high ground (R8), whether it offers a safe place for the family (R9), and whether it protects belongings from moisture (R10).

It is worth noting that these items combine attributes of a different nature. The location of the dwelling on high ground is a condition of the site, whereas water ingress or the protection of belongings describe the performance of the construction, and the existence of a safe place alludes to an appropriation of domestic space in the face of the threat. This heterogeneity is not in itself a defect, but it anticipates that the dimension could behave as an index of conditions rather than as a reflective scale of a single latent attribute, a distinction that has consequences for the interpretation of internal consistency.

Psychosocial dimension: concern and distress in the face of floods

The literature on flood risk perception has distinguished cognitive components, such as perceived probability and severity, and affective components, such as worry. The classic review by Bubeck et al. (2012), cited here as the most widely used reference in this field, showed that risk perception alone is weakly related to the adoption of mitigation measures, and that the assessment of coping capacity plays a more consistent role. Recent studies with exposed rural populations have treated worry as an explicit dimension of risk perception, along with preparedness, knowledge, and experience, and have found that housing quality and exposure are related to these dimensions (Savari et al., 2025).

In the area of mental health, Fothergill et al. (2021) documented, through interviews with residents of an area of persistent risk, anticipatory anxiety about future floods, low self-efficacy, and feelings of helplessness, with greater anxiety among those who did not accept the risk. Twiddy et al. (2022) found that residents affected by a flood remained more concerned about future floods than unaffected residents more than a decade later, and that many experienced anxiety when storms were forecast. In low-income contexts, Rahman et al. (2025) associated temporary housing and low perceived safety regarding floods with higher levels of anxiety, and Nöthling et al. (2024) observed that loss of access to basic services accompanied increases in emotional distress and anxiety after a flood. On the northern border of Mexico, Freimund et al. (2022) found that previous experience with floods corresponded with the degree of concern about future floods and that women felt less prepared than men.

The Flood Worry dimension of the instrument integrates these contents: worry that the home will flood during the season (R1), difficulty sleeping when it rains heavily (R2), perception of danger to the family (R3), calmness at the announcement of a storm (R4, reverse), and distress when recalling past floods (R5). The inclusion of an item referring to the memory of previous events introduces an experiential component that, according to the reviewed literature, is associated with current worry, but that is only applicable to those who have experienced a flood.

Social capital and neighborhood support in the face of flooding

Social capital has become established as a central component of community resilience to disasters. A recent systematic review identified five mechanisms through which social capital fosters such resilience: social learning, collective action, preparedness, information communication, and civic responsibility (Zhao et al., 2025). The distinction between bonding, bridging, and linking capital is useful for the present study: in flood-prone rural areas of Bangladesh, Azad and Pritchard (2023) observed that bonding capital predominates at the onset of flooding to protect people and property, bridging capital becomes relevant during the crisis, and linking capital with institutions is decisive in recovery, with all three forms mutually reinforcing each other.

Empirical evidence also shows that the relationship between social capital and resilience depends on the residential context. In flood-prone coastal communities in northern Iran, Pazhuhan et al. (2023) found that bonding capital is notably related to resilience capacity and that respect among neighbors and the availability of public spaces are among the most relevant criteria. In seven flood-prone communities in Nanjing, Chen et al. (2024) identified social cohesion and collective efficacy as the most representative components of social capital, with differences according to the type of housing complex. In the Chinese-language literature, Li et al. (2024) evaluated the social resilience of an urban community affected by the 2021 floods in Henan Province based on resources, organizational management, and culture, and recommended strengthening residents' self-organization and community governance.

The dimension Support among neighbors in the face of floods of the instrument combines these forms of social capital: expected help from neighbors (R11) and mutual warning before storms (R12) correspond to bonding capital; the perception that each family faces floods alone (R13, reverse) expresses its absence; knowledge of where to go in case of evacuation (R14) is related to preparedness, and attention from authorities to the neighborhood (R15) corresponds to linking capital with institutions. This mixed composition, as in the housing dimension, anticipates that the dimension may bring together contents that do not covary in the same way.

Justice in the Face of Risk and Southern Sinaloa

The justice approach to disasters allows psychosocial vulnerability to be placed in its structural context. Gran Castro (2025) showed that the rights violations associated with disasters in Mexico combine threat, vulnerability, and the action of institutions, so that the authorities' attention to flood-prone neighborhoods is not a fact unrelated to habitability, but one of its conditions. Along these lines, the R15 item introduces the institutional dimension of risk into the scale, and its behavior with respect to the other support items will make it possible to assess whether trust in the authorities is part of the same construct as neighborhood support or constitutes a different facet.

The south of Sinaloa, with Escuinapa and Rosario on the coastal plain, constitutes an appropriate setting for studying these relationships. The proximity of Hurricane Willa to Escuinapa in 2018 (Conagua, 2018) and the documentation of the social construction of risk in other cities of the state (Mancera-González and Rivas-Ochoa, 2025) suggest that flood experience and associated concern are part of the daily life of many households. Measuring that experience validly requires, however, a gradual process of building evidence, which begins with the review of the instrument in a pilot survey such as that reported below.

Design

A quantitative, non-experimental, cross-sectional study was carried out with a descriptive and exploratory scope. The data correspond to the survey LEV-0021, "Southern pilot of Sinaloa · Floods and housing", from the FZ-HAB-0016 project of the Factoriza UAS platform, and to the frozen dataset DST-0011, with a cut-off date of June 9, 2026. The survey constitutes an independent sample and is the only one of the project, so data from other surveys were not combined.

Population, sample, and data cutoff

The survey population was defined as persons aged 18 years or older living in households exposed to flooding in southern Sinaloa. The territory corresponds to Escuinapa and El Rosario, Sinaloa, within an exposure area delimited by the project lead. According to the Factoriza log, 14 persons opened the questionnaire and 6 completed it; no cases were excluded due to quality control, so 6 cases were analyzed (Table 2).

The analyzed cut is preliminary: it brings together 6 cases out of a target of 100, and the survey remained open, in field status, as of the cut-off date. This condition limits the scope of all the results. With 6 cases for 15 items, the ratio is 0.4 cases per item, far below that required by any internal structure analysis; the adequacy of the data and of the sample is the first condition that the methodological literature asks to verify in item factor analysis (Ferrando et al., 2022; Goretzko et al., 2021). Nor was it possible to distinguish in the cut the proportion of participants from each municipality or their sociodemographic characteristics.

Instrument

A 15-item scale with a 5-point response format was applied, from Strongly disagree (1) to Strongly agree (5), organized into 3 dimensions of 5 items each: Concern about flooding (R1 to R5), Resistance of the home to rain (R6 to R10), and Support among neighbors during floods (R11 to R15). 3 items are worded in the reverse direction, R4 ("I feel calm when a storm is announced"), R7 ("Rainwater enters my home") and R13 ("In my neighborhood, each family faces floods alone"), and are recoded in the data matrix. After recoding, a higher score indicates more concern, more perceived resistance of the home, and more neighbor support, respectively. Table 1 presents the code, dimension, text, and direction of each item.

Procedure

Responses were collected through the survey questionnaire managed in Factoriza UAS, which records express online informed consent before the first item and does not collect identifying data. Factoriza froze the DST-0011 dataset with a cryptographic fingerprint, generated the matrix with reverse-scored items recoded, and produced the descriptive and internal consistency analyses (runs RUN-00047 to RUN-00053). Using the data file and the item dictionary exported by the platform, the complementary analyses described below were run.

Data analysis

In a first stage, Factoriza UAS calculated the mean, standard deviation, skewness, kurtosis, and response percentages at the extreme values of each item, as well as the internal consistency of each dimension using Cronbach's alpha (Cronbach, 1951), with the corrected item-total correlation and alpha if the item is deleted. In a second stage, the program analisis_psicometrico.py, version 1.0, was run in Python 3.13.16 (Python Software Foundation, 2026), with NumPy 2.5.3 (Harris et al., 2020), SciPy 1.18.1 (Virtanen et al., 2020), and pandas 3.0.5. The program applies fixed and declared procedures, without manual intervention in the calculations, and excludes from all analyses items without variation.

The program estimated McDonald's total omega (McDonald, 1999/2013) for each dimension from a one-factor model, without inverting reactants, as a complement to alpha, which assumes equivalent charges between reactants. He also attempted exploratory factor analysis of all items with variation, preceded by the Kaiser-Meyer-Olkin sample adequacy measure (Kaiser, 1974), reagent adequacy (Lorenzo-Seva & Ferrando, 2021) and Bartlett's sphericity test (Bartlett, 1950), with number of factors by parallel analysis, extraction by minimal residual, and oblimin rotation; the classical indices are cited in their original source because they are the methodological reference for each procedure. Finally, he estimated the relationship between the mean scores of the dimensions with the Spearman coefficient and its 95% confidence interval, using the Fisher transformation with the standard error of Bonett and Wright (2000).

The reference values for interpreting the indices are treated as guidelines from the literature and not as decision rules (Ferrando et al., 2022). No item was eliminated by the analysis's decision; the program flags problematic items and leaves the decision to the authors. No confirmatory factor analysis was performed, which is proposed as future work with the final wave or with 1 independent sample.

Ethical considerations

Participation was voluntary and anonymous. Factoriza UAS records each participant's express online informed consent and does not collect names, addresses, emails, or other identifying data, which is especially relevant in a population that may have experienced material losses or traumatic experiences.

Declaration on the use of artificial intelligence

During the preparation of this manuscript, the Factoriza UAS platform was used for data management and descriptive and internal consistency analyses; a Python analysis program with fixed and validated procedures for omega, exploratory factor analysis, and correlations between dimensions; and the Claude Opus language model (Anthropic), to run that program, assisted literature search, writing a draft, and assembling the document from the results. The authors reviewed, corrected, and approved the entirety of the text, verified each reference, and assume full responsibility for the content of the article.

Sample composition

Descriptive statistics

Table 3 and Figure 1 present the descriptive statistics. In Concern about flooding, the highest mean corresponded to R1, concern that the home will flood during the season (M = 3.17, SD = 1.33), and the lowest to R2, difficulty sleeping when it rains heavily (M = 2.17, SD = 0.98). In Housing resistance to rain, R6, “My home withstands heavy rains,” obtained the same response in all six cases (M = 4.00, SD = 0.00), whereas R9, the safe place inside the home, and R10, the protection of belongings, reached means of 3.83 (SD = 0.98 and 0.75), and R8, location on high ground, had the lowest in the dimension (M = 2.83, SD = 0.75). In Support among neighbors in the event of flooding, the highest mean was that of R12, mutual warning before storms (M = 3.50, SD = 0.55), and the lowest was that of R15, response from authorities (M = 2.83, SD = 1.17); R14, knowing where to go in case of evacuation, showed the greatest dispersion in the set (SD = 1.47).

In descriptive terms, worry scores were around or below the midpoint of the scale, those for housing resistance were above it, and those for neighborhood support were around it. The absence of variation in R6 prevents calculating its skewness, its kurtosis, and its correlation with the rest of its dimension, and is itself a finding: in this sample, all participants reported agreeing that their housing withstands heavy rains, which coexists with intermediate means for water ingress and terrain location. With six cases, the skewness and kurtosis indices, which range from -1.44 (R9) to 0.89 (R13) and from -3.33 (R12) to 3.60 (R9), lack stability and should not be interpreted as a description of the population.

Internal consistency

Table 4 brings together the item analysis by dimension calculated by Factoriza, and Table 5 presents the alpha and omega coefficients from the complementary program. In Concern about flooding, alpha coincided in both calculations (.843) and omega was .870. In Support among neighbors in the face of flooding, alpha also coincided (.739) and omega was .757. In Housing resistance to rain, Factoriza reported an alpha of .682 with the 5 items, including R6 without variation, whereas the program, which excludes items without variation, estimated an alpha of .727 and an omega of .772 with 4 items. Both values are consistent with each other, because .727 is exactly the alpha if R6, as reported by Factoriza, is eliminated.

The item-by-item analysis identifies, in each dimension, an item that is poorly coherent with the rest. In Concern about flooding, R4, calmness when a storm is announced in its recoded version, presented a corrected item-total correlation of .214, and its removal would raise the dimension's alpha to .901; in contrast, R2 reached the highest correlation (.957). In Support among neighbors in the event of flooding, R12, neighbors warning each other about storms, showed a corrected correlation of .049, and its removal would raise alpha to .797, while R15, attention from the authorities, presented the highest (.768). In Housing resistance to rain, R8, location on high ground, had the highest correlation (.856), and its removal would reduce alpha to .437, which indicates that the consistency of this dimension rests largely on a single item.

Internal structure

The planned exploratory factor analysis could not be estimated interpretably with this cut, and for that reason its results are not presented in tables. After excluding R6 due to lack of variation, the program worked with 14 items and 6 complete cases. With fewer cases than items, the correlation matrix is not invertible, so the overall Kaiser-Meyer-Olkin measure of sampling adequacy, the item-level adequacy, and Bartlett's test of sphericity could not be computed. Although parallel analysis suggested 3 factors and the program completed the extraction, the solution presented communalities greater than 1 for R2 (1.016) and R7 (1.025), that is, Heywood cases, which indicate an improper solution. These results, which are recorded in the reproducibility appendix, do not provide evidence on the dimensionality of the instrument and should not be cited as such.

Relationship between dimensions

DISCUSSION

The purpose of this study was to describe the initial behavior of a scale of habitability and psychosocial vulnerability to floods in the preliminary phase of its pilot survey. The first result, and the most important for the authors, is methodological: with six cases it is not possible to gather evidence of internal structure, and the consistency coefficients obtained should be understood as a first reading of the behavior of the items and not as estimates of the instrument's reliability. The simulation literature on alpha and omega coefficients has been developed with samples much larger than that of this preliminary phase (Malkewitz et al., 2023), and the literature guiding item factor analysis places sample adequacy as a prerequisite for any interpretation (Ferrando et al., 2022; Goretzko et al., 2021).

With that caveat, the cut offers useful signals for reviewing the instrument. The lack of variation in R6 suggests that the general statement that housing withstands heavy rains may generate socially desirable or poorly discriminating responses, especially when it coexists with more specific items, such as water entry or terrain location, which did show dispersion. From the perspective of vulnerability, in which housing precariousness is the most frequently reiterated factor (Hussainzad y Gou, 2024), items that describe observable conditions are more informative than global assessments. The strong dependence of this dimension's alpha on R8 reinforces the idea that the dimension behaves more as a set of housing conditions than as a homogeneous scale.

In the worry dimension, R4's behavior merits attention. Feeling calm at the announcement of a storm, worded in reverse, was the item least consistent with its dimension. A possible explanation is that the reported calm reflects confidence in preparedness or in warnings, and not only the absence of worry, consistent with the distinction between risk perception and appraisal of coping capacity (Bubeck et al., 2012). Another is that reverse wording generated comprehension difficulties. The items referring to insomnia and the feeling of danger to the family, by contrast, showed high consistency, which aligns with the description of anticipatory anxiety about storms in exposed populations (Fothergill et al., 2021; Twiddy et al., 2022).

In the neighborhood support dimension, mutual warning in the event of storms (R12) was the least consistent item, whereas attention from the authorities (R15) was the most consistent with the total. This pattern is compatible with the distinction between bonding capital and linking capital (Azad and Pritchard, 2023): warning among neighbors could be a widespread and weakly differentiating practice, whereas the perception of institutional attention and of concrete support in the event of flooding would vary more across households. The positive and high, although imprecise, relationship between perceived housing resistance and neighborhood support is compatible with evidence linking social capital to adaptive capacity (Pazhuhan et al., 2023; Zhao et al., 2025), but with six cases it cannot be distinguished from a sampling coincidence, and the cross-sectional design would in any case prevent establishing its direction.

Finally, the review of the instrument should consider that some items presuppose experiences that not all people have, such as remembering past floods (R5) or the need to evacuate (R14). The literature indicates that previous flood experience is related to current concern (Freimund et al., 2022; Twiddy et al., 2022), so recording whether the household has been flooded and when would make it possible to separate experience from concern and better describe the psychosocial vulnerability of households in Escuinapa and El Rosario.