Decoupling exposure, risk perception, and preparedness in smallholder systems: Evidence from rural Madagascar

Decoupling exposure, risk perception, and preparedness in smallholder systems: Evidence from rural Madagascar | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Decoupling exposure, risk perception, and preparedness in smallholder systems: Evidence from rural Madagascar Jacob Emanuel Joseph, Clarisse Umutoni, Joséa Raharison, Anthony M Whitbread, and 1 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-9091846/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Household climate risk in Madagascar reflects the interaction of hazard exposure, perceived risk, and structural capacity to prepare for shocks. As climate variability and extremes intensify across tropical smallholder regions, understanding how these dimensions align at the household level becomes increasingly important. This study examines these dynamics in two contrasting regions, cyclone-exposed Sava and drought-prone Atsimo Andrefana. Using data from 245 stratified households, we integrate a composite Exposure Index, perceived likelihood and severity measures, and a multidimensional Shock Readiness Index capturing assets, credit access, and financial buffers. Results show a decoupling between exposure, perceived risk, and readiness. In Sava, households report higher exposure and perceived risk but lower readiness, associated with limited credit access, liquidity constraints, and specialized crop-based livelihoods. In contrast, households in Atsimo Andrefana perceive lower risk yet demonstrate comparatively higher readiness, supported by diversified livelihoods and convertible livestock assets. These patterns indicate a perception–readiness gap in which heightened awareness does not translate into preparedness under binding structural constraints. Early warning access shows limited association with readiness, suggesting that information alone has limited influence where response options are constrained. By integrating perceived risk with structural preparedness, the study clarifies why preparedness does not systematically track exposure in climate-exposed rural systems. In the context of ongoing global climate change, the findings suggest that strengthening structural capacity through improved financial access, livelihood diversification, and institutional support will be critical to narrowing preparedness gaps in smallholder regions beyond Madagascar. Climate risk exposure early warning systems Madagascar resilience Shock Readiness Index (SRI) Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Figure 7 Figure 8 1. Introduction Climate risks are intensifying across low-income rural systems, yet effective climate risk management requires more than identifying hazard exposure alone. Established risk frameworks emphasize that disaster risk emerges from the interaction of hazard, exposure, and vulnerability, including the capacity to anticipate and respond (Cardona et al., 2012 ). At the household level, this interaction depends not only on experienced shocks but also on how risks are perceived and whether households possess the structural resources required to act. Evidence shows that awareness of hazards does not automatically translate into protective action (Sullivan-Wiley & Gianotti, 2017 ). Even where exposure and risk awareness are high, structural and economic constraints can inhibit preparedness (Noll et al., 2021). Such perception–readiness gaps are particularly consequential in low-income agricultural systems, where liquidity, assets, and institutional support condition adaptive capacity (Carter & Barrett, 2006 ). Madagascar provides a pertinent context for examining these dynamics. The northeast is regularly affected by cyclones and flooding, while the semi-arid southwest faces recurrent droughts and prolonged dry spells. These hazards intersect with poverty, weak infrastructure, and limited market access, constraining households’ ability to anticipate, absorb, and recover from shocks (Pret et al., 2025 ). At the same time, regional climate assessments project greater variability, more intense tropical cyclones, and heightened drought risk under continued warming (IPCC, 2022 ). These trends are likely to exacerbate livelihood vulnerabilities and deepen structural inequalities in rural systems. Despite this context, most climate-risk assessments in Madagascar operate at national or regional scales and rely on aggregated indicators (Schatte and Meyer, 2025 ; Van der Lee et al., 2022 ). While useful for planning, these approaches obscure household-level variation in exposure, behavioral responses, and structural preparedness. Few studies analyze exposure, perception, and readiness together, even though these dimensions often diverge in ways that shape resilience outcomes (Brooks et al., 2005 ; Jones and Tanner, 2016 ). Integrating these components within a unified empirical framework enables clearer differentiation of household risk profiles beyond uniform adaptation assumptions. Gender further conditions these dynamics. Women often report higher perceived climate risk, yet limited access to credit, land, and information technologies constrains their ability to translate awareness into adaptive action (Acanga et al., 2025 ). Unequal access to assets and decision-making authority therefore shapes who can convert risk information into preparedness. This study addresses these gaps through a household-level climate-risk assessment in two ecologically distinct regions of Madagascar: cyclone-exposed Sava in the northeast and drought-prone Atsimo Andrefana in the southwest. Using data from 245 stratified households, the analysis quantifies exposure to climatic hazards, perceived likelihood and severity, and a multidimensional Shock Readiness Index capturing assets, credit access, and financial buffers. By integrating behavioral appraisal with structural capacity, the framework links perceived risk to the ability to act under contrasting hazard and livelihood contexts (Constable et al., 2022 ; Ginbo & Hansson, 2023 ). The study addresses three questions: (1) How do household exposure and perceived climate risk differ across hazard environments? (2) To what extent does perceived risk align with structural preparedness capacity? (3) Does access to early warning information strengthen preparedness once structural constraints are considered? By examining how exposure, perception, and structural capacity interact across contrasting contexts, the study provides a diagnostic perspective on household vulnerability under intensifying climate variability. 2. Materials and methods 2.1 Conceptual framework The study adopts the IPCC and Cardona et al. ( 2012 ) risk framework, in which climate risk emerges from interactions among hazard, exposure, and vulnerability, while adaptive capacity shapes response potential. Building on this perspective, household climate risk is examined through three interrelated dimensions: exposure, risk perception, and structural preparedness capacity. Exposure represents households’ cumulative experience with climatic shocks, reflecting the intensity and diversity of hazard events encountered over time. Risk perception captures households’ appraisal of the likelihood and potential severity of climatic hazards. Structural preparedness capacity reflects resource-based conditions that enable or constrain responses to shocks, including access to productive assets, credit, and short-term financial buffers. In this study, structural preparedness capacity is operationalized through the Shock Readiness Index (SRI). Access to early warning systems is treated as a contextual informational factor that may interact with household capacity but remains analytically distinct from the structural components of the SRI. A schematic illustration of the conceptual framework is provided in Supplementary Figure S1 . Composite indices are employed to facilitate systematic comparison across heterogeneous household contexts and are widely used in climate-risk and vulnerability assessments (Birkmann et al., 2013 ; Brooks et al., 2005 ; Islam & Winkel, 2017 ). 2.2 Study design The study was conducted in Sava (humid, cyclone-exposed) and Atsimo Andrefana (semi-arid, drought-prone), two agroecologically contrasting regions of Madagascar. Surveys in Atsimo Andrefana covered two districts (Toliara II and Betioky Atsimo), while data collection in Sava focused on Sambava district. These locations were selected to represent dominant hazard–livelihood configurations, enabling comparison between a cyclone-exposed, crop-oriented system and a drought-prone, diversified agropastoral system. Differences in rainfall regimes, livelihood structures, and hazard profiles provide a basis for examining how exposure, perceived risk, and structural preparedness vary across contrasting contexts. A structured household survey collected information on demographic characteristics, livelihoods, productive assets, credit access, financial buffers, experience of climatic hazards, perceived likelihood and severity, and access to early warning information. Instrument development followed established approaches in climate-risk and vulnerability assessments (Birkmann et al., 2013 ). Sampling used a stratified random design, with strata defined by region, household headship (male vs female), and landholding category (small, medium, large) within each fokontany. Updated village census lists served as the sampling frame. A total of 245 households were retained for analysis (121 in Sava and 124 in Atsimo Andrefana) after removing incomplete questionnaires. The sample included 86 male- and 30 female-headed households in Sava and 94 male- and 35 female-headed households in Atsimo Andrefana. Preliminary power calculations indicated that approximately 120 households per region were sufficient to detect moderate standardized differences (Cohen’s d ≈ 0.5) at α = 0.05. 2.3 Data collection Data in both regions were collected in September 2025 to ensure temporal consistency across study sites. Enumerators received structured training covering questionnaire interpretation, translation into Malagasy, gender-sensitive interviewing, asset scoring procedures, and ethical conduct. Training was followed by a one-day field pilot conducted in one village in each region to refine translations and confirm respondent comprehension. Daily supervision and consistency checks were conducted during fieldwork to maintain data quality and consistent implementation across regions. Respondents were asked about major climatic shocks experienced during a common five-year recall period to ensure cross-regional comparability. Savings were recorded using locally meaningful value ranges (e.g., 3 million) to facilitate recall and reduce reporting error. Missing data were minimal. Item-level missingness was below 5% for all variables. Questionnaires with more than 5% missing responses were excluded, resulting in a final analytic sample of 245 households. Excluded cases were few and did not materially affect regional balance. All respondents provided informed verbal consent in accordance with the ethical procedures of the Food Systems Resilience Program (FSRP 1 ) in Madagascar. 2.4 Analytical framework The analytical framework operationalizes the three dimensions introduced above, namely exposure, risk perception, and structural preparedness capacity, through composite household-level indices. The analysis examines experienced shocks, subjective threat appraisal, and structural conditions shaping response potential as complementary dimensions of household climate-risk dynamics (Cardona et al., 2012 ). All components were normalized using min–max scaling to ensure comparability across variables measured on different scales. Equal weighting was applied across components, a widely used approach in multidimensional climate-risk and vulnerability assessments when dimensions represent conceptually distinct elements and theoretical balance across components is desired (Stevens et al., 2023 ). Because exposure, perception, and preparedness capacity represent analytically distinct and complementary aspects of household climate risk, equal weighting provides a transparent aggregation rule. Sensitivity analyses assessing scaling choices and index robustness are documented in Supplementary S2. 2.4.1 Exposure Index Exposure captures households’ cumulative experience with climatic shocks within the standardized recall period. Three components were used: the total number of events experienced, the maximum reported severity, and the number of distinct hazard types. Severity was recorded on a five-point scale ranging from 1 (very low impact) to 5 (extremely severe impact), as reported by respondents. Reported events refer to major climatic shocks experienced during this period. Composite exposure indices that integrate multiple indicators into a single relative measure are increasingly applied in climate, environmental health, and risk assessment research to support comparisons across heterogeneous populations (Martenies et al., 2023 ; van Asselt & Aung, 2025 ). Self-reported shock histories are widely used in climate-risk diagnostics and are recognized as an appropriate basis for constructing household-level exposure indicators, particularly in rural and data-sparse settings (Nguyen & Nguyen, 2019 ). For household 𝑖, the Exposure Index is: $${E}_{i}=\frac{1}{3}\left(\frac{{\text{E}\text{v}\text{e}\text{n}\text{t}\text{s}}_{\text{i}}}{max\left(Events\right)}+\frac{{Severity}_{i}}{max\left(Severity\right)}+\frac{{Types}_{i}}{max\left(Types\right)}\right)$$ This formulation draws on validated exposure metrics in climate-risk and livelihood-shock literature (Cardona et al., 2012 ; Doan et al., 2023 ). 2.4.2 Risk perception Index Risk perception was assessed through Likert‑scaled ratings of the perceived likelihood and expected severity of major hazards, averaged across all hazards reported by each respondent. Likert responses were mapped to numerical values (1–5). Although items were developed specifically for this survey, they were piloted during field testing to ensure comprehension and cultural appropriateness. This approach captures households’ subjective appraisal of climate hazards, reflecting the cognitive dimension of vulnerability. For \(j\) hazards reported by household \(i\) : $${P}_{i}=\frac{1}{j}\sum_{h=1}^{j}\frac{{Likelihood}_{ih}+{Severity}_{ih}}{2}$$ This formulation builds on established risk perception frameworks (Cardona et al., 2012 ; Van Valkengoed et al., 2023 ). 2.4.3 Shock Readiness Index (SRI) The Shock Readiness Index (SRI) represents households’ structural preparedness capacity, combining productive assets, credit access, and financial buffers. These three dimensions are widely recognized in adaptive‑capacity research as critical enablers of preparedness (Brooks et al., 2005 ; Islam & Winkel, 2017 ). 1. Asset index This consisted of assets condition score (Good = 1, Fair = 0.5, Poor = 0) and proportion of assets not lost in recent shocks and given by \({A}_{i}=\frac{1}{2}({Cond}_{i}+{\text{N}\text{o}\text{L}\text{o}\text{s}\text{s}}_{i}\) ) 2. Credit index It combines usage (Yes/No) and credit limit (Small, Medium, Large → 0.33, 0.66, 1). \({C}_{i}=\frac{1}{2}({Used}_{i}+{\text{L}\text{i}\text{m}\text{i}\text{t}}_{i}\) ) 3. Buffer index The buffer dimension includes access to emergency funds and size of typical savings (Small/Medium/Large). \({B}_{i}=\frac{1}{2}({Access}_{i}+{\text{A}\text{m}\text{o}\text{u}\text{n}\text{t}}_{i}\) ) The \(SRI\) is computed as: \({SRI}_{i}=\frac{1}{3}({A}_{i}+{\text{B}}_{i}+{\text{C}}_{i}\) ) The resulting index ranges from 0 to 1, with higher values indicating greater structural preparedness capacity. Robustness checks for index consistency, scaling sensitivity, and distributional assumptions are documented in Supplementary S2. 2.5 Statistical analysis The analysis proceeded in three stages. First, descriptive statistics (means, medians, and distribution summaries) were used to characterize household profiles and compare patterns across regions and household headship groups. Second, visual exploration used distribution plots (boxplots and raincloud plots), bivariate scatterplots, and quadrant matrices positioning households according to their exposure and readiness scores. Third, inferential tests assessed group differences and associations among the key indices. Welch’s tests were used for regional and gender comparisons due to unequal variances, and effect sizes were quantified using Cohen’s d. Pearson and Spearman correlations examined relationships between exposure, perceived risk, and readiness. Gender-disaggregated analyses followed established guidance for integrating social differentiation into climate-risk assessments (Ravera et al., 2016 ). To examine multivariate associations, ordinary least squares (OLS) regression models were estimated. These models predicted (i) perceived risk from region, household headship, and exposure, and (ii) the Shock Readiness Index (SRI) from region, headship, exposure, perceived risk, and early warning variables. Model estimates, confidence intervals, and diagnostics are reported in Supplementary Tables S4 and S5. Statistical significance was evaluated at α = 0.05, and 95% confidence intervals and effect sizes are reported where relevant. 3. Results 3.1 Household characteristics and livelihood The two study regions exhibited measurable differences in demographic composition, education, dependency structure, and livelihood activities ( Table S3 ). Statistically significant differences were observed for three of the four indicators. Households in Sava reported a smaller average household size (4.0 persons) than households in Atsimo Andrefana (5.0 persons; t = − 6.24, p < 0.001). Dependency ratios also differed, with lower mean dependency in Sava (0.58) than in Atsimo Andrefana (0.74; t = − 3.88, p < 0.001). Mean years of education were higher in Sava (6.0 years) than in Atsimo Andrefana (4.0 years; t = 5.85, p < 0.001). The proportion of female-headed households did not differ significantly between regions (p = 0.57). Figure 1 shows the distribution of households’ primary livelihood activities by region and household headship. A chi-square test indicated significant variation in the composition of livelihoods across the two settings. In Atsimo Andrefana, households reported a more diverse set of primary activities: 80% identified agropastoral production, and smaller shares reported engagement in animal husbandry (10%), crafts and artisanal work (4%), trade and commerce (3%), services (2%), fishing (< 1%), and construction or manual labor (< 1%). In contrast, households in Sava predominantly identified crop-based agricultural production as their main livelihood: 98% reported agriculture as their primary activity, 1% reported artisanal work, and 1% reported animal husbandry. Livelihood differences by household headship mirrored these regional contrasts. In Atsimo Andrefana, female-headed households reported agropastoral production (80%), livestock husbandry (6%), crafts (6%), and trade and commerce (5%), whereas male-headed households reported a similar distribution (90% in agropastoral production, 7% in animal husbandry, 2% in services, 1% in fishing, and 2% in construction). In Sava, livelihood profiles remained largely uniform: female-headed households reported 97% engagement in agricultural production and 3% in manual labor, while male-headed households reported 98% and 1%, respectively. 3.2 Household capacity and resilience resources 3.2.1 Labor availability per season Table 1 summarizes household labor availability across seasons in the two study regions. Clear seasonal differences were observed. In Sava, labor supply peaked during the rainy season (mean 72, median 90) and remained relatively high during harvest (mean 27, median 27), but was substantially lower in the dry season (mean 7, median 5). By contrast, in Atsimo Andrefana labor availability was highest during the dry season (mean 47, median 15) and remained elevated during the rainy season (mean 42, median 15), while harvest labor was comparatively low (mean 8, median 8) and sowing season averaged 13 labor-days. Table 1 also reports labor availability by household headship. Across most seasons and in both regions, male-headed households reported higher adult-equivalent labor-days than female-headed households, although several exceptions were observed. In Atsimo Andrefana, female-headed households reported higher labor availability during the dry season (53 vs. 43 labor-days), while both groups reported similar availability during harvest (8 labor-days). Male-headed households reported higher labor availability during the rainy and sowing seasons. In Sava, male-headed households generally reported higher labor availability across seasons, except during the rainy season, when female-headed households reported slightly higher values (73 vs. 66 labor-days). Full seasonal and headship-specific values are presented in Table 1 . Table 1 Household labor availability (adult-equivalent labor-days per season) by region and household headship Season Atsimo Andrefana Sava Atsimo Andrefana Sava Male-HH Female-HH Male-HH Female-HH Dry 47 7 43 53 8 2 Harvest 8 27 8 8 27 18 Rainy 42 72 44 35 66 73 Sowing 13 27 15 7 28 26 3.2.2 Asset inventory Figure 2 summarizes household asset conditions and the share of assets usable during shocks across the two study regions. In Atsimo Andrefana, assets were predominantly rated as fair (71%), with 25% classified as poor and only 4% as good (Fig. 2 a). In contrast, households in Sava reported a larger share of assets in fair (88%) or good condition (5%), while only 7% were classified as poor. A chi-square test confirmed that differences in asset condition between the regions were statistically significant (χ² = 35.54, df = 2, p < 0.001). Despite these differences in asset composition, the usability of assets during shocks showed a different pattern (Fig. 2 b). In Sava, 46% of assets were reported as usable during shocks compared with 38% in Atsimo Andrefana. However, the proportion of assets reported as lost or damaged during shocks was broadly similar across the two regions, at approximately 64–65%. 3.2.3 Credit access Credit access profiles differed across the two regions (Fig. 3 ). In Atsimo Andrefana, households relied mainly on credit from family and friends (56%), followed by shopkeeper credit and microfinance institutions. Overall, 77% of households reported borrowing during the previous 12 months. In Sava, family and friends remained the dominant source (59%), but community savings groups and microfinance institutions represented a relatively larger share of borrowing. Recent borrowing was less common than in Atsimo Andrefana, with 67% of households reporting credit use in the previous year. Table S4 in the Supplementary Material summarizes credit sources by region and household headship. Family and friends were the dominant source of borrowing across both regions, accounting for 50–57% of borrowing among male-headed households and reaching 74% among female-headed households in Sava. In Atsimo Andrefana, female-headed households relied slightly less on family-based credit (50%) than male-headed households (57%), whereas the opposite pattern was observed in Sava (74% vs. 51%). Shopkeeper credit was particularly common among male-headed households in Sava (23%), while bank credit contributed a larger share among female-headed households in Atsimo Andrefana (19%). Microfinance institutions and community savings groups accounted for smaller shares of borrowing across household types. 3.2.4 Household buffering capacity Regional differences in household buffering strategies were evident across the two study areas (Fig. 4 ). In Atsimo Andrefana, households relied most heavily on food or commodity stocks (39%), followed by family or community support (26%) and livestock that could be sold quickly (20%). Smaller shares reported buffering through cash at home (13%), mobile money balances (1%), or other sources (1%). In Sava, the most common buffering resource was cash at home (34%), followed by food or commodity stocks (28%), livestock sales (16%), and mobile money balances (12%). Reliance on family or community support (10%) was lower than in Atsimo Andrefana. Across regions, households reported high short-term accessibility to their primary buffers. As shown in Fig. 4 (bottom panel), 93% of households in Atsimo Andrefana and 89% in Sava indicated that they could access their main buffer within 72 hours. The regional difference was not statistically significant (p = 0.16). 3.3 Household hazard exposure, risk perception, and readiness dynamics 3.3.1 Hazard profile Figure 5 shows reported hazard events in Sava and Atsimo Andrefana. Households most frequently reported floods (49%) and cyclones (43%) in Sava, while drought (3%), crop pest or disease outbreaks (2%), and livestock disease (1%) were reported less frequently. In Atsimo Andrefana, hazards were more evenly distributed: cyclones were most frequently reported (60%), followed by drought (13%), floods (11%), and crop pest or disease outbreaks (11%), while livestock disease accounted for 3% of reported events. Overall, flood events were reported more often in Sava, whereas cyclone reporting was higher in Atsimo Andrefana, where hazards were distributed across a broader set of categories. 3.3.2 Exposure and risk perception by region and household type Household exposure to climate-related hazards, measured using a 0–1 index, differed systematically by region and only modestly by household headship (Fig. 6 a). In Atsimo Andrefana, male- and female-headed households showed similar median exposure levels, with no statistically significant difference between groups (Welch t = 0.81, p = 0.425). In Sava, exposure scores were higher overall, and male-headed households reported slightly higher values than female-headed households (t = 2.266, p = 0.026). Cross-regional comparisons showed significantly greater exposure in Sava for both household types (all p < 0.01). Summary statistics and confidence intervals are reported in Supplementary Tables S5A–S5B. Perceived risk scores, measured on a 1–4 scale, showed similarly strong regional contrasts (Fig. 6 b). Within Atsimo Andrefana, male- and female-headed households reported comparable levels of perceived risk (t = − 1.788, p = 0.079). In Sava, perceived risk scores were higher overall, and differences between household types were not statistically significant (t = 0.467, p = 0.642). Cross-regional comparisons indicated significantly higher perceived risk in Sava for both household types (all p < 0.001). Corresponding descriptive statistics and confidence intervals are provided in Supplementary Tables S5C–S5D. 3.3.3 Household shock readiness Shock Readiness Index (SRI) scores showed relatively limited variation across regions and household headship groups (Supplementary Table S5E). Mean SRI values ranged between 0.53 and 0.60 across all groups. In Atsimo Andrefana, mean scores were 0.60 for female-headed households and 0.53 for male-headed households. In Sava, corresponding values were 0.55 for female-headed households and 0.58 for male-headed households. Overall, differences by region and household headship were modest, with group means clustering around 0.55–0.58. Within-region Welch t-tests (Supplementary Table S5F) indicated no statistically significant differences between male- and female-headed households in either region (Atsimo Andrefana: t = − 1.567, p = 0.12; Sava: t = 0.904, p = 0.37). Across regions, only the male-headed household comparison was statistically significant, with male-headed households in Atsimo Andrefana reporting lower SRI scores than those in Sava (t = − 2.369, p = 0.02). The corresponding comparison for female-headed households was not statistically significant (t = 1.382, p = 0.17). To better understand the sources of these modest readiness differences, the asset, credit, and buffer components of the SRI were compared across regions (Fig. 7 ). Asset scores were similar between regions (0.37 in Atsimo Andrefana; 0.36 in Sava), whereas credit and buffer scores were slightly higher in Atsimo Andrefana (credit = 0.69, buffer = 0.70) than in Sava (credit = 0.61, buffer = 0.68). To complement these comparisons, multivariate OLS models were estimated to examine correlates of perceived risk and shock readiness (Supplementary Tables S6A–S6B). In the perceived-risk model, both region (β = 0.854, 95% CI: 0.716–0.992, p < 0.001) and exposure index (β = 1.207, 95% CI: 0.483–1.919, p = 0.001) were significant predictors, whereas household headship was not. The model explained 51% of the variance (R² = 0.51). In the SRI model, perceived risk was the only variable significantly associated with readiness (β = −0.0796, 95% CI: −0.110 to − 0.048, p < 0.001). Region showed a small positive coefficient (β = 0.0457, p = 0.043), while household headship and exposure index were not significant. The model explained 10.7% of the variance (R² = 0.107). Figure 8 summarizes the relationships between exposure, perceived risk, and shock readiness across the two study regions. Panel (a) plots perceived risk against the exposure index. In Atsimo Andrefana, perceived-risk values cluster narrowly between 1.5 and 2.5 across the observed exposure range (0.15–0.60), and the fitted line indicates a weak association. In Sava, the relationship is more pronounced, with households experiencing higher exposure scores (up to approximately 0.8) generally reporting higher perceived risk. Panel (b) shows perceived risk in relation to the SRI. In both regions, the fitted regression lines slope downward, indicating that households reporting higher perceived risk tend to have lower readiness scores. The negative association is stronger in Atsimo Andrefana, where SRI values decline from approximately 0.7–0.8 at lower perceived-risk levels to below 0.5 at higher perceived-risk levels. In Sava, the slope is weaker and SRI values display greater variability across the full perceived-risk range. Panel (c) presents the distribution of households across the four exposure–readiness quadrants. In Atsimo Andrefana, households are relatively evenly distributed: 29% fall in the high-exposure/high-readiness quadrant, 9% in high-exposure/low-readiness, 29% in low-exposure/high-readiness, and 34% in low-exposure/low-readiness. In Sava, the distribution is more skewed, with 48% of households in the high-exposure/low-readiness quadrant. The remaining households are distributed across high-exposure/high-readiness (17%), low-exposure/high-readiness (25%), and low-exposure/low-readiness (9%). 3.3.4 Early warning access and household shock readiness Shock Readiness Index (SRI) scores were compared across households receiving different types of early warnings and across varying numbers of warning channels (Supplementary Figure S4; Supplementary Tables S7A–S7H). TV warnings were reported by only one household and were therefore excluded from the analysis. Across both regions, differences in SRI between households that did and did not receive radio, SMS, or neighbor warnings were small. In Atsimo Andrefana, median SRI values for recipients and non-recipients clustered between 0.50 and 0.60, with overlapping interquartile ranges across household headship groups. In Sava, households reporting radio or SMS warnings showed slightly higher median SRI values (approximately 0.55–0.65) than non-recipients (approximately 0.50–0.55), although distributions overlapped substantially. Grouping households by the number of warning channels accessed showed similarly modest differences. In the pooled sample, mean SRI increased slightly from about 0.54 among households with no channels to approximately 0.57–0.60 among households with one or two channels. A one-way ANOVA restricted to households with zero, one, or two channels detected no statistically significant differences (p ≈ 0.49), and the estimated effect size was small (η² ≈ 0.006). Multivariate linear regression models were estimated to assess whether early warning access predicted readiness after controlling for region, household headship, exposure index, and perceived risk (Supplementary Tables S7A–S7H). In the baseline model including the number of warning channels, perceived risk was the only statistically significant predictor of SRI (β ≈ −0.08, p < 0.001). Coefficients for number of channels, region, and household headship were not statistically significant. Adding warning-type indicators (radio, SMS, neighbor) produced only a small increase in model fit (R² ≈ 0.10 to R² ≈ 0.14), and the warning-type coefficients remained non-significant. A region × channels interaction term was also not statistically significant. 4. Discussion Integrating behavioral appraisal with structural preparedness shows that hazard exposure, perceived risk, and adaptive capacity do not align predictably at the household level. Across the two regions studied, households in Sava experienced higher exposure and stronger perceived risk but lower readiness, whereas households in Atsimo Andrefana reported lower perceived risk alongside slightly higher preparedness. These patterns indicate that preparedness reflects the interaction between behavioral responses and structural enabling conditions rather than exposure alone, reinforcing evidence that vulnerability evolves through context-specific and path-dependent processes in low-income agricultural systems (Cinner et al., 2018 ; Miller et al., 2010 ). Under intensifying climate variability, rising exposure may therefore increase perceived risk without proportionate gains in preparedness (IPCC, 2022 ). 4.1 Livelihood structure and preparedness Livelihood structure and socio-demographic characteristics shaped household risk profiles in distinct ways across the two regions. Households in Sava exhibited characteristics often associated with adaptive capacity, including smaller household sizes, lower dependency ratios, and higher education levels. However, these advantages coexisted with strong livelihood specialization within a vanilla-dominated agropastoral economy. Such concentration can increase sensitivity to climatic and market variability by limiting diversification opportunities (Barrett et al., 2001 ), indicating that human capital advantages do not necessarily translate into higher preparedness. In contrast, households in Atsimo Andrefana combined agriculture with livestock keeping, fishing, artisanal work, small trade, and seasonal labor. This diversified livelihood structure aligns with evidence that multiple income streams can buffer households against climatic and market shocks (Barrett et al., 2001 ) and may support absorptive capacity even where perceived risk is comparatively modest. Buffering strategies further distinguished the regions. Households in Sava relied more heavily on financial liquidity, including cash and mobile money, whereas households in Atsimo Andrefana depended more on livestock assets, food stocks, and reciprocal social networks. These patterns reinforce that preparedness reflects locally configured combinations of assets, markets, and social relations rather than a simple regional ranking of vulnerability (IPCC, 2022 ). 4.2 Structural readiness, credit access, and vulnerability constraints The Shock Readiness Index highlighted structural and financial factors as important correlates of preparedness. Although regional differences were moderate, households in Atsimo Andrefana consistently showed slightly stronger asset protection, broader credit use, and more stable financial buffers. The clearest divergence concerned credit access: despite higher exposure and perceived risk, households in Sava reported lower credit use and smaller borrowing capacity. While the data did not allow separation of credit supply and demand constraints, this pattern is consistent with interpretations from vulnerability-trap literature, in which households recognize risk but remain unable to act because of structural barriers (Carter & Barrett, 2006 ; Beaman et al., 2014 ; Kaila et al., 2020 ). Likewise, more stable savings patterns observed in Atsimo Andrefana may help reduce welfare losses and limit reliance on erosive coping strategies such as asset liquidation (Janzen & Carter, 2019). These results suggest that financial enabling conditions shape preparedness outcomes, helping explain why higher perceived risk in Sava did not translate into higher readiness. 4.3 Hazard exposure, risk perception, and the perception–readiness gap The relationship between exposure, perception, and readiness differed systematically between the regions. In Sava, frequent cyclones, rainfall shocks, and crop losses were associated with higher perceived risk, consistent with behavioral evidence showing that recent hazard experience elevates risk appraisal (Ginbo & Hansson, 2023 ). However, increased awareness did not correspond to higher readiness, indicating a perception–readiness gap that is plausibly explained by structural constraints rather than differences in risk awareness. If hazard frequency or intensity increases (IPCC, 2022 ), rising exposure may elevate perceived risk without proportionate gains in readiness where structural barriers persist (Wachinger et al., 2013). In Atsimo Andrefana, lower perceived risk coexisted with slightly higher readiness. This pattern can be interpreted as a form of behavioral adjustment under recurrent stress; whereby repeated exposure normalizes hazards while reinforcing routine preparedness practices embedded within diversified livelihoods. Because behavioral adaptation was not directly measured, this interpretation should be viewed as a plausible explanation rather than a demonstrated mechanism. Gender differences provided additional insight. Female-headed households reported higher perceived risk yet showed similar readiness levels compared with male-headed households. Although decision-making authority and resource control were not explicitly measured, this pattern suggests that greater risk awareness does not necessarily translate into greater preparedness capacity. This interpretation is consistent with broader evidence showing that gendered vulnerability is often shaped less by differences in risk perception than by structural constraints in asset ownership, financial access, and livelihood opportunities that limit the ability to act on perceived risk (Erman et al., 2021 ). The findings therefore point toward a potential gendered pathway of vulnerability that warrants further investigation. 4.4 Early warning access and the limits of information without capacity Early warning access showed weak and uneven associations with household readiness across both regions. In Atsimo Andrefana, households receiving radio, SMS, or neighbor alerts did not demonstrate substantially higher preparedness than those without warnings. In Sava, warning access was associated with slightly higher readiness, particularly among female-headed households, but these differences remained modest and inconsistent. Similarly, access to multiple warning channels showed limited association with preparedness. Although multi-channel systems are often linked with improved comprehension and trust (Pescaroli et al., 2025 ; Rokhideh et al., 2025 ), regression results indicated that warning receipt did not meaningfully predict readiness once exposure and perceived risk were considered. Importantly, the analysis captured warning access rather than warning quality, timing, or specific behavioral responses, which constrains conclusions about the effectiveness of early warning systems. Overall, the findings indicate that information alone may be insufficient to generate meaningful preparedness gains where households face binding resource constraints. Erman et al. ( 2021 ) similarly highlight that adaptive capacity depends on enabling conditions such as asset ownership, financial access, and institutional support. These results suggest that the effectiveness of climate information services may depend on the broader systems within which they are embedded. 4.5 Limitations Several limitations should be acknowledged. Exposure and perception indicators relied on self-reported shock histories and were therefore susceptible to recall bias, particularly for less salient events. The cross-sectional design prevented causal inference between exposure, perception, and readiness, and potential endogeneity between perception and preparedness cannot be excluded. Composite indices required assumptions regarding scaling and weighting. Although sensitivity checks improved confidence in the results, aggregation may have masked variation among underlying readiness components. Finally, the study focused on two regional contexts rather than national representation; however, the selected regions capture contrasting hazard and livelihood archetypes commonly observed in climate-exposed rural systems. 4.6 Synthesis and broader implications The results demonstrate that household climate risk cannot be inferred from hazard exposure or perceived threat alone. Preparedness emerged from the interaction between hazard experience and structural enabling conditions, consistent with established risk frameworks (Cardona et al., 2012 ). Across the two regions, differences in livelihood structure, asset portfolios, and financial access helped explain why exposure and readiness did not align. Under projected increases in hydroclimatic variability across many tropical smallholder regions, exposure is likely to strengthen in ways that do not automatically expand household preparedness (IPCC, 2022 ). The present findings suggest that without parallel strengthening of structural capacity, intensifying hazard signals may widen perception–readiness gaps rather than close them. The exposure–readiness quadrant provides a practical diagnostic framework for differentiating household risk profiles. Within this framework, high-exposure/low-readiness households represent contexts where financial and institutional support may be prioritized, whereas high-exposure/high-readiness households may benefit more from anticipatory or information-based interventions. Low-exposure/high-readiness contexts reflect resilience-maintenance situations, while low-exposure/low-readiness households may require longer-term structural investment. Rather than prescribing fixed interventions, this typology offers a decision-support lens for tailoring climate risk management strategies to local household conditions. The limited association between early warning access and readiness further reinforces the importance of structural capacity. Climate information appears to function primarily as a reinforcing input rather than an independent driver of preparedness when households lack feasible protective options. Mwangi et al. ( 2021 ) similarly show that the influence of climate information services on decision-making depends on institutional context and user characteristics. Effective climate risk management in low-income rural systems therefore requires integrating climate information services with financial, livelihood, and institutional mechanisms that enable households to translate awareness into action. 5. Conclusion This study demonstrates that, in a context of intensifying climate variability, household preparedness to climate risk in rural Madagascar is associated more strongly with structural and livelihood conditions than with exposure or perceived risk alone. Across contrasting hazard environments, households in Atsimo Andrefana exhibited slightly higher readiness, likely supported by diversified livelihoods and convertible livestock assets, whereas households in Sava combined higher exposure and stronger risk perception with lower preparedness and more limited financial flexibility. These contrasting profiles indicate that preparedness reflects the interaction between hazard context, livelihood structure, and enabling capacity rather than exposure or perception alone. The findings further suggest a decoupling between perceived risk and preparedness when financial and material constraints limit households’ ability to act. Early warning access showed limited association with readiness once exposure and perceived risk were considered, highlighting that information alone may not translate into preparedness where structural constraints persist. By integrating exposure indicators, risk-perception measures, and a multidimensional readiness index, the study contributes a household-level diagnostic framework that clarifies how behavioral and structural dimensions of climate risk interact in low-income, climate-exposed settings. These results imply that expanding climate information services alone is unlikely to strengthen preparedness where households face binding structural constraints. Improving access to credit, protecting productive assets, supporting livelihood diversification, and strengthening institutional support appear central to enabling households to translate risk awareness into effective action. Without reinforcing these structural conditions, increasing hazard exposure under continued warming may heighten awareness without narrowing preparedness gaps. Declarations Author Contribution J.E.J. conceptualized the study, collected the data, conducted the investigation and analysis, developed the software, and wrote the main manuscript text. C.U. contributed to conceptualization, supervision, investigation, and manuscript review. J.R. contributed to data collection and investigation. A.M.W. contributed to conceptualization, supervision, funding acquisition, and manuscript review. A.W. contributed to conceptualization, supervision, and manuscript review. All authors reviewed and approved the final manuscript. Acknowledgement This work was funded by the World Bank Group through the Food Systems Resilience Program (FSRP) – Madagascar. The authors gratefully acknowledge this support, which enabled the completion of the present study. Data Availability The datasets generated and analyzed during this study are not publicly available due to confidentiality considerations associated with household survey data but are available from the corresponding author on reasonable request. References Acanga A, Matovu B, Murale V, Arlikatti S (2025) Gender perspectives in disaster response: An evidence-based review. Progress Disaster Sci 26:100416. ttps://doi.org/10.1016/j.pdisas.2025.100416 Adger WN, Brown K, Nelson DR, Berkes F, Eakin H, Folke C, Galvin K, Gunderson L, Goulden M, O’Brien K, Ruitenbeek J, Tompkins EL (2011) Resilience implications of policy responses to climate change. Wiley Interdisciplinary Reviews Clim Change 2(5):757–766. ttps://doi.org/10.1002/wcc.133 Barrett CB, Reardon T, Webb P (2001) Nonfarm income diversification and household livelihood strategies in rural Africa: concepts, dynamics, and policy implications. Food Policy 26(4):315–331. ttps://doi.org/10.1016/s0306-9192(01)00014-8 Beaman L, Karlan D, Thuysbaert B, Udry C (2014) Selection into Credit Markets: Evidence from Agriculture in Mali. ttps://doi.org/10.3386/w20387 Birkmann J, Cardona OD, Carreño ML, Barbat AH, Pelling M, Schneiderbauer S, Kienberger S, Keiler M, Alexander D, Zeil P, Welle T (2013) Framing vulnerability, risk and societal responses: the MOVE framework. Nat Hazards 67(2):193–211. ttps://doi.org/10.1007/s11069-013-0558-5 Brooks N, Adger WN, Kelly PM (2005) The determinants of vulnerability and adaptive capacity at the national level and the implications for adaptation. Glob Environ Change 15(2):151–163. ttps://doi.org/10.1016/j.gloenvcha.2004.12.006 Cardona O, Van Aalst MK, Birkmann J, Fordham M, McGregor G, Perez R, Pulwarty RS, Schipper ELF, Sinh BT, Décamps H, Keim M, Davis I, Ebi KL, Lavell A, Mechler R, Murray V, Pelling M, Pohl J, Smith A, Thomalla F (2012) Determinants of risk: exposure and vulnerability. In Cambridge University Press eBooks (pp. 65–108). ttps://doi.org/10.1017/cbo9781139177245.005 Carter MR, Barrett CB (2006) The economics of poverty traps and persistent poverty: An asset-based approach. J Dev Stud 42(2):178–199. ttps://doi.org/10.1080/00220380500405261 Cinner JE, Adger WN, Allison EH, Barnes ML, Brown K, Cohen PJ, Gelcich S, Hicks CC, Hughes TP, Lau J, Marshall NA, Morrison TH (2018) Building adaptive capacity to climate change in tropical coastal communities. Nat Clim Change 8(2):117–123. ttps://doi.org/10.1038/s41558-017-0065-x Constable AJ, French S, Karoblyte V, Viner D (2022) Decision-Making for managing Climate-Related risks: Unpacking the decision process to avoid Trial-and-Error responses. Front Clim. 4ttps://doi.org/10.3389/fclim.2022.754264 Doan MK, Hill R, Hallegatte S, Corral P, Brunckhorst B, Nguyen M, Freije-Rodriguez S, Naikal E (2023) Counting people exposed to, vulnerable to, or at high risk from climate shocks: a methodology. In World Bank. ttps://doi.org/10.1596/1813-9450-10619. Washington, DC eBooks Erman A, De Vries Robbé SA, Thies SF, Kabir K, Maruo M, International Bank for Reconstruction and Development / The World Bank (2021) Gender dimensions of disaster risk and resilience. International Bank for Reconstruction and Development / The World Bank Ginbo T, Hansson H (2023) Intra-household risk perceptions and climate change adaptation in sub-Saharan Africa. Eur Rev Agric Econ 50(3):1039–1063. ttps://doi.org/10.1093/erae/jbad011 IPCC (2022) Climate Change 2022: Impacts, adaptation and Vulnerability: Contribution of Working Group II to the Sixth Assessment Report of the Intergovernmental Panel on Climate Change (H.-O. Pörtner, D. C. Roberts, M. M. B. Tignor, E. Poloczanska, & K. Mintenbeck, Eds.). Cambridge University Press. ttps://doi.org/10.1017/9781009325844 Islam N, Winkel J (2017) Climate change and social inequality (DESA Working Paper No. 152). UNITED NATIONS Department of Economic and Social Affairs. Retrieved November 27, 2025, from https://www.un.org/esa/desa/papers/2017/wp152_2017.pdf Janzen SA, Carter MR (2018) After the Drought: The impact of microinsurance on consumption smoothing and asset protection. Am J Agric Econ 101(3):651–671. ttps://doi.org/10.1093/ajae/aay061 Jones L, Tanner T (2016) Subjective resilience’: using perceptions to quantify household resilience to climate extremes and disasters. Reg Envriron Chang 17(1):229–243. ttps://doi.org/10.1007/s10113-016-0995-2 Kaila H, Singhal S, Tuteja D (2020) Development programs, security, and violence reduction: Evidence from an insurgency in India. World Dev 130:104911. ttps://doi.org/10.1016/j.worlddev.2020.104911 Martenies SE, Zhang M, Corrigan AE, Kvit A, Shields T, Wheaton W, Him DA, Aschner J, Talavera-Barber MM, Barrett ES, Bastain TM, Bendixsen C, Breton CV, Bush NR, Cacho F, Camargo CA, Carroll KN, Carter BS, Cassidy-Bushrow AE, Kress AM (2023) Developing a National-Scale Exposure Index for Combined Environmental Hazards and Social Stressors and Applications to the Environmental Influences on Child Health Outcomes (ECHO) cohort. Int J Environ Res Public Health 20(14):6339. ttps://doi.org/10.3390/ijerph20146339 Miller F, Osbahr H, Boyd E, Thomalla F, Bharwani S, Ziervogel G, Walker B, Van Der Birkmann J, Rockström J, Hinkel J, Downing T, Folke C, Nelson D (2010) Resilience and vulnerability: complementary or conflicting concepts? Ecol Soc 15(3). ttps://doi.org/10.5751/es-03378-150311 Mwangi M, Kituyi E, Ouma G (2021) A systematic review of the literature on the contribution of past climate information services pilot projects in climate risk management. Sci Afr 14:e01005. ttps://doi.org/10.1016/j.sciaf.2021.e01005 Nguyen G, Nguyen TT (2019) Exposure to weather shocks: A comparison between self-reported record and extreme weather data. Econ Anal Policy 65:117–138. ttps://doi.org/10.1016/j.eap.2019.11.009 Pescaroli G, Dryhurst S, Karagiannis GM (2025) Bridging gaps in research and practice for early warning systems: new datasets for public response. Front Communication 10. ttps://doi.org/10.3389/fcomm.2025.1451800 Pret V, Falconnier GN, Affholder F, Corbeels M, Chikowo R, Descheemaeker K (2025) Farm resilience to climatic risk. A review. Agron Sustain Dev 45(1):10. ttps://doi.org/10.1007/s13593-024-00998-w Ravera F, Martín-López B, Pascual U, Drucker A (2016) The diversity of gendered adaptation strategies to climate change of Indian farmers: A feminist intersectional approach. Ambio 45(S3):335–351. ttps://doi.org/10.1007/s13280-016-0833-2 Rokhideh M, Fearnley C, Budimir M (2025) Multi-Hazard Early Warning Systems in the SenDai Framework for Disaster Risk Reduction: Achievements, gaps, and future directions. Int J Disaster Risk Sci 16(1):103–116. ttps://doi.org/10.1007/s13753-025-00622-9 Schatte P, Meyer M (2025) Assessing holistic agroecological resilience of agroecosystems from a landscape perspective: a systematic review. Ecol Soc 30(2). ttps://doi.org/10.5751/es-16137-300224 Sullivan-Wiley KA, Gianotti AGS (2017) Risk perception in a Multi-Hazard environment. World Dev 97:138–152. ttps://doi.org/10.1016/j.worlddev.2017.04.002 Stevens SM, Joy MK, Abrahamse W, Milfont TL, Petherick LM (2023) Composite environmental indices—a case of rickety rankings. PeerJ 11:e16325. ttps://doi.org/10.7717/peerj.16325 van Asselt J, Aung ZW (2025) Community and household shocks and coping strategies: Findings from the ninth round of the Myanmar Household Welfare Survey (April - October 2025). Myanmar SSP Working Paper 74. Washington, DC: International Food Policy Research Institute. https://hdl.handle.net/10568/179640 Van Der Lee J, Kangogo D, Gülzari ŞÖ, Dentoni D, Oosting S, Bijman J, Klerkx L (2022) Theoretical positions and approaches to resilience assessment in farming systems. A review. Agron Sustain Dev 42(2). ttps://doi.org/10.1007/s13593-022-00755-x Van Valkengoed AM, Perlaviciute G, Steg L (2023) From believing in climate change to adapting to climate change: The role of risk perception and efficacy beliefs. Risk Anal 44(3):553–565. ttps://doi.org/10.1111/risa.14193 Wachinger G, Renn O, Begg C, Kuhlicke C (2012) The Risk Perception Paradox—Implications for governance and communication of Natural Hazards. Risk Anal 33(6):1049–1065. ttps://doi.org/10.1111/j.1539-6924.2012.01942.x Footnotes More information about the project: https://ewsdata.rightsindevelopment.org/files/documents/66/WB-P178566.pdf Additional Declarations No competing interests reported. Supplementary Files SupplementarymaterialMadagascar.docx Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-9091846","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":620981568,"identity":"27aeb514-5afa-4c84-a9da-752ff5efa132","order_by":0,"name":"Jacob Emanuel Joseph","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA4UlEQVRIiWNgGAWjYBAC9gb+B2AGH3sDA0MCMVoYG3ggDDaeAyRrkSBKPUQL64afOw7LsUk+fvziQc09Bn7p4xcIaOE/drP3zGFjNuk0M4uEY8UMkn05BYRsYbvB23Y4sU06h80ggS2BweAMD34ngrTc/Nt2uL5N8gxQyz8itAgCtdwG2pLAJsHD/CCxDaSF/QBeLdLMQC2ybemGbTxpZgyJfQk8kj08eHUAY7D/2c23bdby/OyHH3/88S1Bjp+H/QF+PcwIJpsEkABawWOAXwuy7g8QmpAto2AUjIJRMNIAAI5lQUVuzUWgAAAAAElFTkSuQmCC","orcid":"","institution":"International Livestock Research Institute","correspondingAuthor":true,"prefix":"","firstName":"Jacob","middleName":"Emanuel","lastName":"Joseph","suffix":""},{"id":620981569,"identity":"2d8247e1-76b7-4d02-8b9d-975210d0ef48","order_by":1,"name":"Clarisse Umutoni","email":"","orcid":"","institution":"International Livestock Research Institute","correspondingAuthor":false,"prefix":"","firstName":"Clarisse","middleName":"","lastName":"Umutoni","suffix":""},{"id":620981570,"identity":"1bcca71a-9e0d-49d7-82ce-1047704ca27f","order_by":2,"name":"Joséa Raharison","email":"","orcid":"","institution":"International Livestock Research Institute","correspondingAuthor":false,"prefix":"","firstName":"Joséa","middleName":"","lastName":"Raharison","suffix":""},{"id":620981571,"identity":"18cee3bb-1d91-4615-9f45-f580ec08ae00","order_by":3,"name":"Anthony M Whitbread","email":"","orcid":"","institution":"International Livestock Research Institute","correspondingAuthor":false,"prefix":"","firstName":"Anthony","middleName":"M","lastName":"Whitbread","suffix":""},{"id":620981572,"identity":"c17a78ca-db94-444f-ae7b-406659ff6d99","order_by":4,"name":"Abdrahmane Wane","email":"","orcid":"","institution":"International Livestock Research Institute","correspondingAuthor":false,"prefix":"","firstName":"Abdrahmane","middleName":"","lastName":"Wane","suffix":""}],"badges":[],"createdAt":"2026-03-11 08:23:17","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-9091846/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-9091846/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":106968457,"identity":"5a1716e9-fb86-494b-a840-dcca5f4efa10","added_by":"auto","created_at":"2026-04-15 10:08:35","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":66587,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003ePrimary livelihood activities reported by households in Atsimo Andrefana and Sava.\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003ePanel (a) shows overall regional differences; panel (b) shows household headship differences\u003c/strong\u003e.\u003c/p\u003e","description":"","filename":"1.png","url":"https://assets-eu.researchsquare.com/files/rs-9091846/v1/c3d530eebea597fb7d0eb4b9.png"},{"id":106968460,"identity":"41e2cfab-e5fc-4984-8552-b90f50553b27","added_by":"auto","created_at":"2026-04-15 10:08:35","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":88004,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eHousehold asset condition and resilience by region. (a) Share of assets in poor, fair, and good condition (%). (b) Share of assets usable during shocks versus lost/damaged (%).\u003c/strong\u003e\u003c/p\u003e","description":"","filename":"2.png","url":"https://assets-eu.researchsquare.com/files/rs-9091846/v1/2e9958fcdd3dafe4f42bf4d3.png"},{"id":106968272,"identity":"0317fb49-8aff-4186-a7e8-c47f2f54542a","added_by":"auto","created_at":"2026-04-15 10:07:24","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":103762,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eHousehold credit sources and borrowing by region. Donut charts show source distribution; bars show the share accessing credit in the past 12 months\u003c/strong\u003e.\u003c/p\u003e","description":"","filename":"3.png","url":"https://assets-eu.researchsquare.com/files/rs-9091846/v1/a174e2cd7d25c520c2bde7d1.png"},{"id":106968286,"identity":"4ac62aef-fad0-4146-9bee-0ce50e9f5cd8","added_by":"auto","created_at":"2026-04-15 10:07:37","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":99218,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eHousehold buffering capacity by region. Top panel: Share of households relying on different buffer sources (%). Bottom panel: Share of households able to access buffers within 72 hours (%).\u003c/strong\u003e\u003c/p\u003e","description":"","filename":"4.png","url":"https://assets-eu.researchsquare.com/files/rs-9091846/v1/63c7e241ee46afce267393e3.png"},{"id":106968327,"identity":"ddba3ce2-7dec-43ad-9ceb-086fb83cf783","added_by":"auto","created_at":"2026-04-15 10:07:56","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":103192,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eRegional distribution of reported hazard events across Atsimo Andrefana and Sava.\u003c/strong\u003e\u003c/p\u003e","description":"","filename":"5.png","url":"https://assets-eu.researchsquare.com/files/rs-9091846/v1/3f59c64cfb86812df3d93b63.png"},{"id":106968421,"identity":"2530ec5b-0cca-49d2-8618-fd82c3543dab","added_by":"auto","created_at":"2026-04-15 10:08:23","extension":"png","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":41783,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eExposure and perceived risk by region and household headship. (a) Household exposure index (0–1), where higher values indicate greater exposure based on the number of hazard events, maximum severity, and diversity of hazard types experienced. (b) Perceived risk score (1–4), where 1 = low, 2 = moderate, 3 = high/likely, and 4 = extremely high/very likely. Boxplots show distributions for male- and female-headed households in Atsimo Andrefana and Sava (median, interquartile range, and outliers).\u003c/strong\u003e\u003c/p\u003e","description":"","filename":"6.png","url":"https://assets-eu.researchsquare.com/files/rs-9091846/v1/eafa9765673527f474aa8d9f.png"},{"id":106968274,"identity":"0d50bedc-9dec-41ab-9387-b4635f0acf70","added_by":"auto","created_at":"2026-04-15 10:07:24","extension":"png","order_by":7,"title":"Figure 7","display":"","copyAsset":false,"role":"figure","size":55621,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eMean readiness component scores (0–1) for assets, credit, and financial buffers in Atsimo Andrefana and Sava. Higher values indicate greater coping capacity.\u003c/strong\u003e\u003c/p\u003e","description":"","filename":"7.png","url":"https://assets-eu.researchsquare.com/files/rs-9091846/v1/e3b0bcb2ecfda6fd2e80a35e.png"},{"id":106968328,"identity":"1117703f-ee34-4e8e-93fa-54994c766bfc","added_by":"auto","created_at":"2026-04-15 10:07:56","extension":"png","order_by":8,"title":"Figure 8","display":"","copyAsset":false,"role":"figure","size":386157,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eExposure, perceived risk, and shock readiness relationships in Atsimo Andrefana and Sava. (a) Exposure vs. perceived risk with fitted trends. (b) Perceived risk vs. SRI. (c) Exposure–readiness quadrants shown as percentage heatmaps by region.\u003c/strong\u003e\u003c/p\u003e","description":"","filename":"8.png","url":"https://assets-eu.researchsquare.com/files/rs-9091846/v1/7adf2b6e3176e256abfdef8f.png"},{"id":106971077,"identity":"2dad63d7-3c01-440d-98e6-03db6b08da9b","added_by":"auto","created_at":"2026-04-15 10:17:23","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":2640638,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-9091846/v1/4d68f4d3-494d-4d56-9d02-8da33a2ee9f9.pdf"},{"id":106968287,"identity":"93e45fe6-3357-4e37-a9c9-70059f24899a","added_by":"auto","created_at":"2026-04-15 10:07:37","extension":"docx","order_by":0,"title":"","display":"","copyAsset":false,"role":"supplement","size":843111,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementarymaterialMadagascar.docx","url":"https://assets-eu.researchsquare.com/files/rs-9091846/v1/1b0e3aa520e21312d8f6d5c7.docx"}],"financialInterests":"No competing interests reported.","formattedTitle":"Decoupling exposure, risk perception, and preparedness in smallholder systems: Evidence from rural Madagascar","fulltext":[{"header":"1. Introduction","content":"\u003cp\u003eClimate risks are intensifying across low-income rural systems, yet effective climate risk management requires more than identifying hazard exposure alone. Established risk frameworks emphasize that disaster risk emerges from the interaction of hazard, exposure, and vulnerability, including the capacity to anticipate and respond (Cardona et al., \u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e2012\u003c/span\u003e). At the household level, this interaction depends not only on experienced shocks but also on how risks are perceived and whether households possess the structural resources required to act. Evidence shows that awareness of hazards does not automatically translate into protective action (Sullivan-Wiley \u0026amp; Gianotti, \u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e2017\u003c/span\u003e). Even where exposure and risk awareness are high, structural and economic constraints can inhibit preparedness (Noll et al., 2021). Such perception\u0026ndash;readiness gaps are particularly consequential in low-income agricultural systems, where liquidity, assets, and institutional support condition adaptive capacity (Carter \u0026amp; Barrett, \u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e2006\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eMadagascar provides a pertinent context for examining these dynamics. The northeast is regularly affected by cyclones and flooding, while the semi-arid southwest faces recurrent droughts and prolonged dry spells. These hazards intersect with poverty, weak infrastructure, and limited market access, constraining households\u0026rsquo; ability to anticipate, absorb, and recover from shocks (Pret et al., \u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e2025\u003c/span\u003e). At the same time, regional climate assessments project greater variability, more intense tropical cyclones, and heightened drought risk under continued warming (IPCC, \u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). These trends are likely to exacerbate livelihood vulnerabilities and deepen structural inequalities in rural systems.\u003c/p\u003e \u003cp\u003eDespite this context, most climate-risk assessments in Madagascar operate at national or regional scales and rely on aggregated indicators (Schatte and Meyer, \u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e2025\u003c/span\u003e; Van der Lee et al., \u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). While useful for planning, these approaches obscure household-level variation in exposure, behavioral responses, and structural preparedness. Few studies analyze exposure, perception, and readiness together, even though these dimensions often diverge in ways that shape resilience outcomes (Brooks et al., \u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e2005\u003c/span\u003e; Jones and Tanner, \u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e2016\u003c/span\u003e). Integrating these components within a unified empirical framework enables clearer differentiation of household risk profiles beyond uniform adaptation assumptions.\u003c/p\u003e \u003cp\u003eGender further conditions these dynamics. Women often report higher perceived climate risk, yet limited access to credit, land, and information technologies constrains their ability to translate awareness into adaptive action (Acanga et al., \u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e2025\u003c/span\u003e). Unequal access to assets and decision-making authority therefore shapes who can convert risk information into preparedness.\u003c/p\u003e \u003cp\u003eThis study addresses these gaps through a household-level climate-risk assessment in two ecologically distinct regions of Madagascar: cyclone-exposed Sava in the northeast and drought-prone Atsimo Andrefana in the southwest. Using data from 245 stratified households, the analysis quantifies exposure to climatic hazards, perceived likelihood and severity, and a multidimensional Shock Readiness Index capturing assets, credit access, and financial buffers. By integrating behavioral appraisal with structural capacity, the framework links perceived risk to the ability to act under contrasting hazard and livelihood contexts (Constable et al., \u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e2022\u003c/span\u003e; Ginbo \u0026amp; Hansson, \u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e2023\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eThe study addresses three questions: (1) How do household exposure and perceived climate risk differ across hazard environments? (2) To what extent does perceived risk align with structural preparedness capacity? (3) Does access to early warning information strengthen preparedness once structural constraints are considered? By examining how exposure, perception, and structural capacity interact across contrasting contexts, the study provides a diagnostic perspective on household vulnerability under intensifying climate variability.\u003c/p\u003e"},{"header":"2. Materials and methods","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003e2.1 Conceptual framework\u003c/h2\u003e \u003cp\u003eThe study adopts the IPCC and Cardona et al. (\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e2012\u003c/span\u003e) risk framework, in which climate risk emerges from interactions among hazard, exposure, and vulnerability, while adaptive capacity shapes response potential. Building on this perspective, household climate risk is examined through three interrelated dimensions: exposure, risk perception, and structural preparedness capacity. Exposure represents households\u0026rsquo; cumulative experience with climatic shocks, reflecting the intensity and diversity of hazard events encountered over time. Risk perception captures households\u0026rsquo; appraisal of the likelihood and potential severity of climatic hazards. Structural preparedness capacity reflects resource-based conditions that enable or constrain responses to shocks, including access to productive assets, credit, and short-term financial buffers. In this study, structural preparedness capacity is operationalized through the Shock Readiness Index (SRI).\u003c/p\u003e \u003cp\u003eAccess to early warning systems is treated as a contextual informational factor that may interact with household capacity but remains analytically distinct from the structural components of the SRI. A schematic illustration of the conceptual framework is provided in Supplementary Figure \u003cspan refid=\"MOESM1\" class=\"InternalRef\"\u003eS1\u003c/span\u003e. Composite indices are employed to facilitate systematic comparison across heterogeneous household contexts and are widely used in climate-risk and vulnerability assessments (Birkmann et al., \u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e2013\u003c/span\u003e; Brooks et al., \u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e2005\u003c/span\u003e; Islam \u0026amp; Winkel, \u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e2017\u003c/span\u003e).\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec4\" class=\"Section2\"\u003e \u003ch2\u003e2.2 Study design\u003c/h2\u003e \u003cp\u003eThe study was conducted in Sava (humid, cyclone-exposed) and Atsimo Andrefana (semi-arid, drought-prone), two agroecologically contrasting regions of Madagascar. Surveys in Atsimo Andrefana covered two districts (Toliara II and Betioky Atsimo), while data collection in Sava focused on Sambava district. These locations were selected to represent dominant hazard\u0026ndash;livelihood configurations, enabling comparison between a cyclone-exposed, crop-oriented system and a drought-prone, diversified agropastoral system. Differences in rainfall regimes, livelihood structures, and hazard profiles provide a basis for examining how exposure, perceived risk, and structural preparedness vary across contrasting contexts.\u003c/p\u003e \u003cp\u003eA structured household survey collected information on demographic characteristics, livelihoods, productive assets, credit access, financial buffers, experience of climatic hazards, perceived likelihood and severity, and access to early warning information. Instrument development followed established approaches in climate-risk and vulnerability assessments (Birkmann et al., \u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e2013\u003c/span\u003e). Sampling used a stratified random design, with strata defined by region, household headship (male vs female), and landholding category (small, medium, large) within each fokontany. Updated village census lists served as the sampling frame.\u003c/p\u003e \u003cp\u003eA total of 245 households were retained for analysis (121 in Sava and 124 in Atsimo Andrefana) after removing incomplete questionnaires. The sample included 86 male- and 30 female-headed households in Sava and 94 male- and 35 female-headed households in Atsimo Andrefana. Preliminary power calculations indicated that approximately 120 households per region were sufficient to detect moderate standardized differences (Cohen\u0026rsquo;s d\u0026thinsp;\u0026asymp;\u0026thinsp;0.5) at α\u0026thinsp;=\u0026thinsp;0.05.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec5\" class=\"Section2\"\u003e \u003ch2\u003e2.3 Data collection\u003c/h2\u003e \u003cp\u003eData in both regions were collected in September 2025 to ensure temporal consistency across study sites. Enumerators received structured training covering questionnaire interpretation, translation into Malagasy, gender-sensitive interviewing, asset scoring procedures, and ethical conduct. Training was followed by a one-day field pilot conducted in one village in each region to refine translations and confirm respondent comprehension. Daily supervision and consistency checks were conducted during fieldwork to maintain data quality and consistent implementation across regions.\u003c/p\u003e \u003cp\u003eRespondents were asked about major climatic shocks experienced during a common five-year recall period to ensure cross-regional comparability. Savings were recorded using locally meaningful value ranges (e.g., \u0026lt;\u0026thinsp;1\u0026nbsp;million Ariary; 1\u0026ndash;3\u0026nbsp;million; \u0026gt;3\u0026nbsp;million) to facilitate recall and reduce reporting error. Missing data were minimal. Item-level missingness was below 5% for all variables. Questionnaires with more than 5% missing responses were excluded, resulting in a final analytic sample of 245 households. Excluded cases were few and did not materially affect regional balance. All respondents provided informed verbal consent in accordance with the ethical procedures of the Food Systems Resilience Program (FSRP\u003csup\u003e1\u003c/sup\u003e) in Madagascar.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec6\" class=\"Section2\"\u003e \u003ch2\u003e2.4 Analytical framework\u003c/h2\u003e \u003cp\u003eThe analytical framework operationalizes the three dimensions introduced above, namely exposure, risk perception, and structural preparedness capacity, through composite household-level indices. The analysis examines experienced shocks, subjective threat appraisal, and structural conditions shaping response potential as complementary dimensions of household climate-risk dynamics (Cardona et al., \u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e2012\u003c/span\u003e). All components were normalized using min\u0026ndash;max scaling to ensure comparability across variables measured on different scales. Equal weighting was applied across components, a widely used approach in multidimensional climate-risk and vulnerability assessments when dimensions represent conceptually distinct elements and theoretical balance across components is desired (Stevens et al., \u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e2023\u003c/span\u003e). Because exposure, perception, and preparedness capacity represent analytically distinct and complementary aspects of household climate risk, equal weighting provides a transparent aggregation rule. Sensitivity analyses assessing scaling choices and index robustness are documented in Supplementary S2.\u003c/p\u003e \u003cdiv id=\"Sec7\" class=\"Section3\"\u003e \u003ch2\u003e2.4.1 Exposure Index\u003c/h2\u003e \u003cp\u003eExposure captures households\u0026rsquo; cumulative experience with climatic shocks within the standardized recall period. Three components were used: the total number of events experienced, the maximum reported severity, and the number of distinct hazard types. Severity was recorded on a five-point scale ranging from 1 (very low impact) to 5 (extremely severe impact), as reported by respondents. Reported events refer to major climatic shocks experienced during this period. Composite exposure indices that integrate multiple indicators into a single relative measure are increasingly applied in climate, environmental health, and risk assessment research to support comparisons across heterogeneous populations (Martenies et al., \u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e2023\u003c/span\u003e; van Asselt \u0026amp; Aung, \u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e2025\u003c/span\u003e). Self-reported shock histories are widely used in climate-risk diagnostics and are recognized as an appropriate basis for constructing household-level exposure indicators, particularly in rural and data-sparse settings (Nguyen \u0026amp; Nguyen, \u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e2019\u003c/span\u003e). For household \u0026#119894;, the Exposure Index is:\u003cdiv id=\"Equa\" class=\"Equation\"\u003e\u003cdiv format=\"TEX\" class=\"mathdisplay\" id=\"FileID_Equa\" name=\"EquationSource\"\u003e\n$${E}_{i}=\\frac{1}{3}\\left(\\frac{{\\text{E}\\text{v}\\text{e}\\text{n}\\text{t}\\text{s}}_{\\text{i}}}{max\\left(Events\\right)}+\\frac{{Severity}_{i}}{max\\left(Severity\\right)}+\\frac{{Types}_{i}}{max\\left(Types\\right)}\\right)$$\u003c/div\u003e\u003c/div\u003e\u003c/p\u003e \u003cp\u003eThis formulation draws on validated exposure metrics in climate-risk and livelihood-shock literature (Cardona et al., \u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e2012\u003c/span\u003e; Doan et al., \u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e2023\u003c/span\u003e).\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec8\" class=\"Section3\"\u003e \u003ch2\u003e2.4.2 Risk perception Index\u003c/h2\u003e \u003cp\u003eRisk perception was assessed through Likert‑scaled ratings of the perceived likelihood and expected severity of major hazards, averaged across all hazards reported by each respondent. Likert responses were mapped to numerical values (1\u0026ndash;5). Although items were developed specifically for this survey, they were piloted during field testing to ensure comprehension and cultural appropriateness. This approach captures households\u0026rsquo; subjective appraisal of climate hazards, reflecting the cognitive dimension of vulnerability. For \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(j\\)\u003c/span\u003e\u003c/span\u003e hazards reported by household \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(i\\)\u003c/span\u003e\u003c/span\u003e:\u003cdiv id=\"Equb\" class=\"Equation\"\u003e\u003cdiv format=\"TEX\" class=\"mathdisplay\" id=\"FileID_Equb\" name=\"EquationSource\"\u003e\n$${P}_{i}=\\frac{1}{j}\\sum_{h=1}^{j}\\frac{{Likelihood}_{ih}+{Severity}_{ih}}{2}$$\u003c/div\u003e\u003c/div\u003e\u003c/p\u003e \u003cp\u003eThis formulation builds on established risk perception frameworks (Cardona et al., \u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e2012\u003c/span\u003e; Van Valkengoed et al., \u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e2023\u003c/span\u003e).\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec9\" class=\"Section3\"\u003e \u003ch2\u003e2.4.3 Shock Readiness Index (SRI)\u003c/h2\u003e \u003cp\u003eThe Shock Readiness Index (SRI) represents households\u0026rsquo; structural preparedness capacity, combining productive assets, credit access, and financial buffers. These three dimensions are widely recognized in adaptive‑capacity research as critical enablers of preparedness (Brooks et al., \u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e2005\u003c/span\u003e; Islam \u0026amp; Winkel, \u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e2017\u003c/span\u003e).\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e\n\u003ch3\u003e1. Asset index\u003c/h3\u003e\n\u003cp\u003e \u003cdiv class=\"BlockQuote\"\u003e \u003cp\u003eThis consisted of assets condition score (Good\u0026thinsp;=\u0026thinsp;1, Fair\u0026thinsp;=\u0026thinsp;0.5, Poor\u0026thinsp;=\u0026thinsp;0) and proportion of assets not lost in recent shocks and given by\u003c/p\u003e \u003cp\u003e \u003cspan class=\"InlineEquation\"\u003e \u003cspan class=\"mathinline\"\u003e\\({A}_{i}=\\frac{1}{2}({Cond}_{i}+{\\text{N}\\text{o}\\text{L}\\text{o}\\text{s}\\text{s}}_{i}\\)\u003c/span\u003e \u003c/span\u003e)\u003c/p\u003e \u003c/div\u003e \u003c/p\u003e\n\u003ch3\u003e2. Credit index\u003c/h3\u003e\n\u003cp\u003e \u003cdiv class=\"BlockQuote\"\u003e \u003cp\u003eIt combines usage (Yes/No) and credit limit (Small, Medium, Large \u0026rarr; 0.33, 0.66, 1).\u003c/p\u003e \u003cp\u003e \u003cspan class=\"InlineEquation\"\u003e \u003cspan class=\"mathinline\"\u003e\\({C}_{i}=\\frac{1}{2}({Used}_{i}+{\\text{L}\\text{i}\\text{m}\\text{i}\\text{t}}_{i}\\)\u003c/span\u003e \u003c/span\u003e)\u003c/p\u003e \u003c/div\u003e \u003c/p\u003e\n\u003ch3\u003e3. Buffer index\u003c/h3\u003e\n\u003cp\u003e \u003cdiv class=\"BlockQuote\"\u003e \u003cp\u003eThe buffer dimension includes access to emergency funds and size of typical savings (Small/Medium/Large).\u003c/p\u003e \u003cp\u003e \u003cspan class=\"InlineEquation\"\u003e \u003cspan class=\"mathinline\"\u003e\\({B}_{i}=\\frac{1}{2}({Access}_{i}+{\\text{A}\\text{m}\\text{o}\\text{u}\\text{n}\\text{t}}_{i}\\)\u003c/span\u003e \u003c/span\u003e)\u003c/p\u003e \u003c/div\u003e \u003c/p\u003e \u003cp\u003eThe \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(SRI\\)\u003c/span\u003e\u003c/span\u003e is computed as:\u003cdiv class=\"BlockQuote\"\u003e\u003cp\u003e \u003cspan class=\"InlineEquation\"\u003e \u003cspan class=\"mathinline\"\u003e\\({SRI}_{i}=\\frac{1}{3}({A}_{i}+{\\text{B}}_{i}+{\\text{C}}_{i}\\)\u003c/span\u003e \u003c/span\u003e)\u003c/p\u003e\u003c/div\u003e\u003c/p\u003e \u003cp\u003eThe resulting index ranges from 0 to 1, with higher values indicating greater structural preparedness capacity. Robustness checks for index consistency, scaling sensitivity, and distributional assumptions are documented in Supplementary S2.\u003c/p\u003e \u003cdiv id=\"Sec13\" class=\"Section2\"\u003e \u003ch2\u003e2.5 Statistical analysis\u003c/h2\u003e \u003cp\u003eThe analysis proceeded in three stages. First, descriptive statistics (means, medians, and distribution summaries) were used to characterize household profiles and compare patterns across regions and household headship groups. Second, visual exploration used distribution plots (boxplots and raincloud plots), bivariate scatterplots, and quadrant matrices positioning households according to their exposure and readiness scores. Third, inferential tests assessed group differences and associations among the key indices. Welch\u0026rsquo;s tests were used for regional and gender comparisons due to unequal variances, and effect sizes were quantified using Cohen\u0026rsquo;s d. Pearson and Spearman correlations examined relationships between exposure, perceived risk, and readiness. Gender-disaggregated analyses followed established guidance for integrating social differentiation into climate-risk assessments (Ravera et al., \u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e2016\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eTo examine multivariate associations, ordinary least squares (OLS) regression models were estimated. These models predicted (i) perceived risk from region, household headship, and exposure, and (ii) the Shock Readiness Index (SRI) from region, headship, exposure, perceived risk, and early warning variables. Model estimates, confidence intervals, and diagnostics are reported in Supplementary Tables S4 and S5. Statistical significance was evaluated at α\u0026thinsp;=\u0026thinsp;0.05, and 95% confidence intervals and effect sizes are reported where relevant.\u003c/p\u003e \u003c/div\u003e"},{"header":"3. Results","content":"\u003cdiv id=\"Sec15\" class=\"Section2\"\u003e \u003ch2\u003e3.1 Household characteristics and livelihood\u003c/h2\u003e \u003cp\u003eThe two study regions exhibited measurable differences in demographic composition, education, dependency structure, and livelihood activities (\u003cb\u003eTable S3\u003c/b\u003e). Statistically significant differences were observed for three of the four indicators. Households in Sava reported a smaller average household size (4.0 persons) than households in Atsimo Andrefana (5.0 persons; t = \u0026minus;\u0026thinsp;6.24, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001). Dependency ratios also differed, with lower mean dependency in Sava (0.58) than in Atsimo Andrefana (0.74; t = \u0026minus;\u0026thinsp;3.88, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001). Mean years of education were higher in Sava (6.0 years) than in Atsimo Andrefana (4.0 years; t\u0026thinsp;=\u0026thinsp;5.85, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001). The proportion of female-headed households did not differ significantly between regions (p\u0026thinsp;=\u0026thinsp;0.57).\u003c/p\u003e \u003cp\u003eFigure \u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e shows the distribution of households\u0026rsquo; primary livelihood activities by region and household headship. A chi-square test indicated significant variation in the composition of livelihoods across the two settings. In Atsimo Andrefana, households reported a more diverse set of primary activities: 80% identified agropastoral production, and smaller shares reported engagement in animal husbandry (10%), crafts and artisanal work (4%), trade and commerce (3%), services (2%), fishing (\u0026lt;\u0026thinsp;1%), and construction or manual labor (\u0026lt;\u0026thinsp;1%). In contrast, households in Sava predominantly identified crop-based agricultural production as their main livelihood: 98% reported agriculture as their primary activity, 1% reported artisanal work, and 1% reported animal husbandry.\u003c/p\u003e \u003cp\u003eLivelihood differences by household headship mirrored these regional contrasts. In Atsimo Andrefana, female-headed households reported agropastoral production (80%), livestock husbandry (6%), crafts (6%), and trade and commerce (5%), whereas male-headed households reported a similar distribution (90% in agropastoral production, 7% in animal husbandry, 2% in services, 1% in fishing, and 2% in construction). In Sava, livelihood profiles remained largely uniform: female-headed households reported 97% engagement in agricultural production and 3% in manual labor, while male-headed households reported 98% and 1%, respectively.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec16\" class=\"Section2\"\u003e \u003ch2\u003e3.2 Household capacity and resilience resources\u003c/h2\u003e \u003cdiv id=\"Sec17\" class=\"Section3\"\u003e \u003ch2\u003e3.2.1 Labor availability per season\u003c/h2\u003e \u003cp\u003eTable\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e summarizes household labor availability across seasons in the two study regions. Clear seasonal differences were observed. In Sava, labor supply peaked during the rainy season (mean 72, median 90) and remained relatively high during harvest (mean 27, median 27), but was substantially lower in the dry season (mean 7, median 5). By contrast, in Atsimo Andrefana labor availability was highest during the dry season (mean 47, median 15) and remained elevated during the rainy season (mean 42, median 15), while harvest labor was comparatively low (mean 8, median 8) and sowing season averaged 13 labor-days.\u003c/p\u003e \u003cp\u003eTable\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e also reports labor availability by household headship. Across most seasons and in both regions, male-headed households reported higher adult-equivalent labor-days than female-headed households, although several exceptions were observed. In Atsimo Andrefana, female-headed households reported higher labor availability during the dry season (53 vs. 43 labor-days), while both groups reported similar availability during harvest (8 labor-days). Male-headed households reported higher labor availability during the rainy and sowing seasons. In Sava, male-headed households generally reported higher labor availability across seasons, except during the rainy season, when female-headed households reported slightly higher values (73 vs. 66 labor-days). Full seasonal and headship-specific values are presented in Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab1\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eHousehold labor availability (adult-equivalent labor-days per season) by region and household headship\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"7\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c7\" colnum=\"7\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003eSeason\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003eAtsimo Andrefana\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003eSava\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colspan=\"2\" nameend=\"c5\" namest=\"c4\"\u003e \u003cp\u003eAtsimo Andrefana\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colspan=\"2\" nameend=\"c7\" namest=\"c6\"\u003e \u003cp\u003eSava\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003e\u003cem\u003eMale-HH\u003c/em\u003e\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003e\u003cem\u003eFemale-HH\u003c/em\u003e\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c6\"\u003e \u003cp\u003e\u003cem\u003eMale-HH\u003c/em\u003e\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c7\"\u003e \u003cp\u003e\u003cem\u003eFemale-HH\u003c/em\u003e\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eDry\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e47\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e7\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e43\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e53\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e8\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e2\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eHarvest\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e8\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e27\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e8\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e8\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e27\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e18\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eRainy\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e42\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e72\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e44\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e35\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e66\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e73\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eSowing\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e13\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e27\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e15\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e7\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e28\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e26\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec18\" class=\"Section3\"\u003e \u003ch2\u003e3.2.2 Asset inventory\u003c/h2\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eFigure \u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e summarizes household asset conditions and the share of assets usable during shocks across the two study regions. In Atsimo Andrefana, assets were predominantly rated as fair (71%), with 25% classified as poor and only 4% as good (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003ea). In contrast, households in Sava reported a larger share of assets in fair (88%) or good condition (5%), while only 7% were classified as poor. A chi-square test confirmed that differences in asset condition between the regions were statistically significant (χ\u0026sup2; = 35.54, df\u0026thinsp;=\u0026thinsp;2, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001).\u003c/p\u003e \u003cp\u003eDespite these differences in asset composition, the usability of assets during shocks showed a different pattern (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eb). In Sava, 46% of assets were reported as usable during shocks compared with 38% in Atsimo Andrefana. However, the proportion of assets reported as lost or damaged during shocks was broadly similar across the two regions, at approximately 64\u0026ndash;65%.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec19\" class=\"Section3\"\u003e \u003ch2\u003e3.2.3 Credit access\u003c/h2\u003e \u003cp\u003eCredit access profiles differed across the two regions (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e). In Atsimo Andrefana, households relied mainly on credit from family and friends (56%), followed by shopkeeper credit and microfinance institutions. Overall, 77% of households reported borrowing during the previous 12 months. In Sava, family and friends remained the dominant source (59%), but community savings groups and microfinance institutions represented a relatively larger share of borrowing. Recent borrowing was less common than in Atsimo Andrefana, with 67% of households reporting credit use in the previous year.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eTable S4 in the Supplementary Material summarizes credit sources by region and household headship. Family and friends were the dominant source of borrowing across both regions, accounting for 50\u0026ndash;57% of borrowing among male-headed households and reaching 74% among female-headed households in Sava. In Atsimo Andrefana, female-headed households relied slightly less on family-based credit (50%) than male-headed households (57%), whereas the opposite pattern was observed in Sava (74% vs. 51%). Shopkeeper credit was particularly common among male-headed households in Sava (23%), while bank credit contributed a larger share among female-headed households in Atsimo Andrefana (19%). Microfinance institutions and community savings groups accounted for smaller shares of borrowing across household types.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec20\" class=\"Section3\"\u003e \u003ch2\u003e3.2.4 Household buffering capacity\u003c/h2\u003e \u003cp\u003eRegional differences in household buffering strategies were evident across the two study areas (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003e). In Atsimo Andrefana, households relied most heavily on food or commodity stocks (39%), followed by family or community support (26%) and livestock that could be sold quickly (20%). Smaller shares reported buffering through cash at home (13%), mobile money balances (1%), or other sources (1%). In Sava, the most common buffering resource was cash at home (34%), followed by food or commodity stocks (28%), livestock sales (16%), and mobile money balances (12%). Reliance on family or community support (10%) was lower than in Atsimo Andrefana. Across regions, households reported high short-term accessibility to their primary buffers. As shown in Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003e (bottom panel), 93% of households in Atsimo Andrefana and 89% in Sava indicated that they could access their main buffer within 72 hours. The regional difference was not statistically significant (p\u0026thinsp;=\u0026thinsp;0.16).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv id=\"Sec21\" class=\"Section2\"\u003e \u003ch2\u003e3.3 Household hazard exposure, risk perception, and readiness dynamics\u003c/h2\u003e \u003cdiv id=\"Sec22\" class=\"Section3\"\u003e \u003ch2\u003e3.3.1 Hazard profile\u003c/h2\u003e \u003cp\u003eFigure \u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003e shows reported hazard events in Sava and Atsimo Andrefana. Households most frequently reported floods (49%) and cyclones (43%) in Sava, while drought (3%), crop pest or disease outbreaks (2%), and livestock disease (1%) were reported less frequently. In Atsimo Andrefana, hazards were more evenly distributed: cyclones were most frequently reported (60%), followed by drought (13%), floods (11%), and crop pest or disease outbreaks (11%), while livestock disease accounted for 3% of reported events. Overall, flood events were reported more often in Sava, whereas cyclone reporting was higher in Atsimo Andrefana, where hazards were distributed across a broader set of categories.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec23\" class=\"Section3\"\u003e \u003ch2\u003e3.3.2 Exposure and risk perception by region and household type\u003c/h2\u003e \u003cp\u003eHousehold exposure to climate-related hazards, measured using a 0\u0026ndash;1 index, differed systematically by region and only modestly by household headship (Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003ea). In Atsimo Andrefana, male- and female-headed households showed similar median exposure levels, with no statistically significant difference between groups (Welch t\u0026thinsp;=\u0026thinsp;0.81, p\u0026thinsp;=\u0026thinsp;0.425). In Sava, exposure scores were higher overall, and male-headed households reported slightly higher values than female-headed households (t\u0026thinsp;=\u0026thinsp;2.266, p\u0026thinsp;=\u0026thinsp;0.026). Cross-regional comparisons showed significantly greater exposure in Sava for both household types (all p\u0026thinsp;\u0026lt;\u0026thinsp;0.01). Summary statistics and confidence intervals are reported in Supplementary Tables S5A\u0026ndash;S5B.\u003c/p\u003e \u003cp\u003ePerceived risk scores, measured on a 1\u0026ndash;4 scale, showed similarly strong regional contrasts (Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003eb). Within Atsimo Andrefana, male- and female-headed households reported comparable levels of perceived risk (t = \u0026minus;\u0026thinsp;1.788, p\u0026thinsp;=\u0026thinsp;0.079). In Sava, perceived risk scores were higher overall, and differences between household types were not statistically significant (t\u0026thinsp;=\u0026thinsp;0.467, p\u0026thinsp;=\u0026thinsp;0.642). Cross-regional comparisons indicated significantly higher perceived risk in Sava for both household types (all p\u0026thinsp;\u0026lt;\u0026thinsp;0.001). Corresponding descriptive statistics and confidence intervals are provided in Supplementary Tables S5C\u0026ndash;S5D.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec24\" class=\"Section3\"\u003e \u003ch2\u003e3.3.3 Household shock readiness\u003c/h2\u003e \u003cp\u003eShock Readiness Index (SRI) scores showed relatively limited variation across regions and household headship groups (Supplementary Table S5E). Mean SRI values ranged between 0.53 and 0.60 across all groups. In Atsimo Andrefana, mean scores were 0.60 for female-headed households and 0.53 for male-headed households. In Sava, corresponding values were 0.55 for female-headed households and 0.58 for male-headed households. Overall, differences by region and household headship were modest, with group means clustering around 0.55\u0026ndash;0.58.\u003c/p\u003e \u003cp\u003eWithin-region Welch t-tests (Supplementary Table S5F) indicated no statistically significant differences between male- and female-headed households in either region (Atsimo Andrefana: t\u0026thinsp;=\u0026thinsp;\u0026minus;\u0026thinsp;1.567, p\u0026thinsp;=\u0026thinsp;0.12; Sava: t\u0026thinsp;=\u0026thinsp;0.904, p\u0026thinsp;=\u0026thinsp;0.37). Across regions, only the male-headed household comparison was statistically significant, with male-headed households in Atsimo Andrefana reporting lower SRI scores than those in Sava (t\u0026thinsp;=\u0026thinsp;\u0026minus;\u0026thinsp;2.369, p\u0026thinsp;=\u0026thinsp;0.02). The corresponding comparison for female-headed households was not statistically significant (t\u0026thinsp;=\u0026thinsp;1.382, p\u0026thinsp;=\u0026thinsp;0.17).\u003c/p\u003e \u003cp\u003eTo better understand the sources of these modest readiness differences, the asset, credit, and buffer components of the SRI were compared across regions (Fig.\u0026nbsp;\u003cspan refid=\"Fig7\" class=\"InternalRef\"\u003e7\u003c/span\u003e). Asset scores were similar between regions (0.37 in Atsimo Andrefana; 0.36 in Sava), whereas credit and buffer scores were slightly higher in Atsimo Andrefana (credit\u0026thinsp;=\u0026thinsp;0.69, buffer\u0026thinsp;=\u0026thinsp;0.70) than in Sava (credit\u0026thinsp;=\u0026thinsp;0.61, buffer\u0026thinsp;=\u0026thinsp;0.68).\u003c/p\u003e \u003cp\u003eTo complement these comparisons, multivariate OLS models were estimated to examine correlates of perceived risk and shock readiness (Supplementary Tables S6A\u0026ndash;S6B). In the perceived-risk model, both region (β\u0026thinsp;=\u0026thinsp;0.854, 95% CI: 0.716\u0026ndash;0.992, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001) and exposure index (β\u0026thinsp;=\u0026thinsp;1.207, 95% CI: 0.483\u0026ndash;1.919, p\u0026thinsp;=\u0026thinsp;0.001) were significant predictors, whereas household headship was not. The model explained 51% of the variance (R\u0026sup2; = 0.51). In the SRI model, perceived risk was the only variable significantly associated with readiness (β = \u0026minus;0.0796, 95% CI: \u0026minus;0.110 to \u0026minus;\u0026thinsp;0.048, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001). Region showed a small positive coefficient (β\u0026thinsp;=\u0026thinsp;0.0457, p\u0026thinsp;=\u0026thinsp;0.043), while household headship and exposure index were not significant. The model explained 10.7% of the variance (R\u0026sup2; = 0.107).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eFigure \u003cspan refid=\"Fig8\" class=\"InternalRef\"\u003e8\u003c/span\u003e summarizes the relationships between exposure, perceived risk, and shock readiness across the two study regions. Panel (a) plots perceived risk against the exposure index. In Atsimo Andrefana, perceived-risk values cluster narrowly between 1.5 and 2.5 across the observed exposure range (0.15\u0026ndash;0.60), and the fitted line indicates a weak association. In Sava, the relationship is more pronounced, with households experiencing higher exposure scores (up to approximately 0.8) generally reporting higher perceived risk.\u003c/p\u003e \u003cp\u003ePanel (b) shows perceived risk in relation to the SRI. In both regions, the fitted regression lines slope downward, indicating that households reporting higher perceived risk tend to have lower readiness scores. The negative association is stronger in Atsimo Andrefana, where SRI values decline from approximately 0.7\u0026ndash;0.8 at lower perceived-risk levels to below 0.5 at higher perceived-risk levels. In Sava, the slope is weaker and SRI values display greater variability across the full perceived-risk range.\u003c/p\u003e \u003cp\u003ePanel (c) presents the distribution of households across the four exposure\u0026ndash;readiness quadrants. In Atsimo Andrefana, households are relatively evenly distributed: 29% fall in the high-exposure/high-readiness quadrant, 9% in high-exposure/low-readiness, 29% in low-exposure/high-readiness, and 34% in low-exposure/low-readiness. In Sava, the distribution is more skewed, with 48% of households in the high-exposure/low-readiness quadrant. The remaining households are distributed across high-exposure/high-readiness (17%), low-exposure/high-readiness (25%), and low-exposure/low-readiness (9%).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec25\" class=\"Section3\"\u003e \u003ch2\u003e3.3.4 Early warning access and household shock readiness\u003c/h2\u003e \u003cp\u003eShock Readiness Index (SRI) scores were compared across households receiving different types of early warnings and across varying numbers of warning channels (Supplementary Figure S4; Supplementary Tables S7A\u0026ndash;S7H). TV warnings were reported by only one household and were therefore excluded from the analysis.\u003c/p\u003e \u003cp\u003eAcross both regions, differences in SRI between households that did and did not receive radio, SMS, or neighbor warnings were small. In Atsimo Andrefana, median SRI values for recipients and non-recipients clustered between 0.50 and 0.60, with overlapping interquartile ranges across household headship groups. In Sava, households reporting radio or SMS warnings showed slightly higher median SRI values (approximately 0.55\u0026ndash;0.65) than non-recipients (approximately 0.50\u0026ndash;0.55), although distributions overlapped substantially.\u003c/p\u003e \u003cp\u003eGrouping households by the number of warning channels accessed showed similarly modest differences. In the pooled sample, mean SRI increased slightly from about 0.54 among households with no channels to approximately 0.57\u0026ndash;0.60 among households with one or two channels. A one-way ANOVA restricted to households with zero, one, or two channels detected no statistically significant differences (p\u0026thinsp;\u0026asymp;\u0026thinsp;0.49), and the estimated effect size was small (η\u0026sup2; \u0026asymp; 0.006).\u003c/p\u003e \u003cp\u003eMultivariate linear regression models were estimated to assess whether early warning access predicted readiness after controlling for region, household headship, exposure index, and perceived risk (Supplementary Tables S7A\u0026ndash;S7H). In the baseline model including the number of warning channels, perceived risk was the only statistically significant predictor of SRI (β \u0026asymp; \u0026minus;0.08, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001). Coefficients for number of channels, region, and household headship were not statistically significant. Adding warning-type indicators (radio, SMS, neighbor) produced only a small increase in model fit (R\u0026sup2; \u0026asymp; 0.10 to R\u0026sup2; \u0026asymp; 0.14), and the warning-type coefficients remained non-significant. A region \u0026times; channels interaction term was also not statistically significant.\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e"},{"header":"4. Discussion","content":"\u003cp\u003eIntegrating behavioral appraisal with structural preparedness shows that hazard exposure, perceived risk, and adaptive capacity do not align predictably at the household level. Across the two regions studied, households in Sava experienced higher exposure and stronger perceived risk but lower readiness, whereas households in Atsimo Andrefana reported lower perceived risk alongside slightly higher preparedness. These patterns indicate that preparedness reflects the interaction between behavioral responses and structural enabling conditions rather than exposure alone, reinforcing evidence that vulnerability evolves through context-specific and path-dependent processes in low-income agricultural systems (Cinner et al., \u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e2018\u003c/span\u003e; Miller et al., \u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e2010\u003c/span\u003e). Under intensifying climate variability, rising exposure may therefore increase perceived risk without proportionate gains in preparedness (IPCC, \u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e2022\u003c/span\u003e).\u003c/p\u003e \u003cdiv id=\"Sec27\" class=\"Section2\"\u003e \u003ch2\u003e4.1 Livelihood structure and preparedness\u003c/h2\u003e \u003cp\u003eLivelihood structure and socio-demographic characteristics shaped household risk profiles in distinct ways across the two regions. Households in Sava exhibited characteristics often associated with adaptive capacity, including smaller household sizes, lower dependency ratios, and higher education levels. However, these advantages coexisted with strong livelihood specialization within a vanilla-dominated agropastoral economy. Such concentration can increase sensitivity to climatic and market variability by limiting diversification opportunities (Barrett et al., \u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e2001\u003c/span\u003e), indicating that human capital advantages do not necessarily translate into higher preparedness.\u003c/p\u003e \u003cp\u003eIn contrast, households in Atsimo Andrefana combined agriculture with livestock keeping, fishing, artisanal work, small trade, and seasonal labor. This diversified livelihood structure aligns with evidence that multiple income streams can buffer households against climatic and market shocks (Barrett et al., \u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e2001\u003c/span\u003e) and may support absorptive capacity even where perceived risk is comparatively modest.\u003c/p\u003e \u003cp\u003eBuffering strategies further distinguished the regions. Households in Sava relied more heavily on financial liquidity, including cash and mobile money, whereas households in Atsimo Andrefana depended more on livestock assets, food stocks, and reciprocal social networks. These patterns reinforce that preparedness reflects locally configured combinations of assets, markets, and social relations rather than a simple regional ranking of vulnerability (IPCC, \u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e2022\u003c/span\u003e).\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec28\" class=\"Section2\"\u003e \u003ch2\u003e4.2 Structural readiness, credit access, and vulnerability constraints\u003c/h2\u003e \u003cp\u003eThe Shock Readiness Index highlighted structural and financial factors as important correlates of preparedness. Although regional differences were moderate, households in Atsimo Andrefana consistently showed slightly stronger asset protection, broader credit use, and more stable financial buffers. The clearest divergence concerned credit access: despite higher exposure and perceived risk, households in Sava reported lower credit use and smaller borrowing capacity.\u003c/p\u003e \u003cp\u003eWhile the data did not allow separation of credit supply and demand constraints, this pattern is consistent with interpretations from vulnerability-trap literature, in which households recognize risk but remain unable to act because of structural barriers (Carter \u0026amp; Barrett, \u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e2006\u003c/span\u003e; Beaman et al., \u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e2014\u003c/span\u003e; Kaila et al., \u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e2020\u003c/span\u003e). Likewise, more stable savings patterns observed in Atsimo Andrefana may help reduce welfare losses and limit reliance on erosive coping strategies such as asset liquidation (Janzen \u0026amp; Carter, 2019). These results suggest that financial enabling conditions shape preparedness outcomes, helping explain why higher perceived risk in Sava did not translate into higher readiness.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec29\" class=\"Section2\"\u003e \u003ch2\u003e4.3 Hazard exposure, risk perception, and the perception\u0026ndash;readiness gap\u003c/h2\u003e \u003cp\u003eThe relationship between exposure, perception, and readiness differed systematically between the regions. In Sava, frequent cyclones, rainfall shocks, and crop losses were associated with higher perceived risk, consistent with behavioral evidence showing that recent hazard experience elevates risk appraisal (Ginbo \u0026amp; Hansson, \u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e2023\u003c/span\u003e). However, increased awareness did not correspond to higher readiness, indicating a perception\u0026ndash;readiness gap that is plausibly explained by structural constraints rather than differences in risk awareness. If hazard frequency or intensity increases (IPCC, \u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e2022\u003c/span\u003e), rising exposure may elevate perceived risk without proportionate gains in readiness where structural barriers persist (Wachinger et al., 2013). In Atsimo Andrefana, lower perceived risk coexisted with slightly higher readiness. This pattern can be interpreted as a form of behavioral adjustment under recurrent stress; whereby repeated exposure normalizes hazards while reinforcing routine preparedness practices embedded within diversified livelihoods. Because behavioral adaptation was not directly measured, this interpretation should be viewed as a plausible explanation rather than a demonstrated mechanism.\u003c/p\u003e \u003cp\u003eGender differences provided additional insight. Female-headed households reported higher perceived risk yet showed similar readiness levels compared with male-headed households. Although decision-making authority and resource control were not explicitly measured, this pattern suggests that greater risk awareness does not necessarily translate into greater preparedness capacity. This interpretation is consistent with broader evidence showing that gendered vulnerability is often shaped less by differences in risk perception than by structural constraints in asset ownership, financial access, and livelihood opportunities that limit the ability to act on perceived risk (Erman et al., \u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e2021\u003c/span\u003e). The findings therefore point toward a potential gendered pathway of vulnerability that warrants further investigation.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec30\" class=\"Section2\"\u003e \u003ch2\u003e4.4 Early warning access and the limits of information without capacity\u003c/h2\u003e \u003cp\u003eEarly warning access showed weak and uneven associations with household readiness across both regions. In Atsimo Andrefana, households receiving radio, SMS, or neighbor alerts did not demonstrate substantially higher preparedness than those without warnings. In Sava, warning access was associated with slightly higher readiness, particularly among female-headed households, but these differences remained modest and inconsistent.\u003c/p\u003e \u003cp\u003eSimilarly, access to multiple warning channels showed limited association with preparedness. Although multi-channel systems are often linked with improved comprehension and trust (Pescaroli et al., \u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e2025\u003c/span\u003e; Rokhideh et al., \u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e2025\u003c/span\u003e), regression results indicated that warning receipt did not meaningfully predict readiness once exposure and perceived risk were considered. Importantly, the analysis captured warning access rather than warning quality, timing, or specific behavioral responses, which constrains conclusions about the effectiveness of early warning systems.\u003c/p\u003e \u003cp\u003eOverall, the findings indicate that information alone may be insufficient to generate meaningful preparedness gains where households face binding resource constraints. Erman et al. (\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e2021\u003c/span\u003e) similarly highlight that adaptive capacity depends on enabling conditions such as asset ownership, financial access, and institutional support. These results suggest that the effectiveness of climate information services may depend on the broader systems within which they are embedded.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec31\" class=\"Section2\"\u003e \u003ch2\u003e4.5 Limitations\u003c/h2\u003e \u003cp\u003eSeveral limitations should be acknowledged. Exposure and perception indicators relied on self-reported shock histories and were therefore susceptible to recall bias, particularly for less salient events. The cross-sectional design prevented causal inference between exposure, perception, and readiness, and potential endogeneity between perception and preparedness cannot be excluded. Composite indices required assumptions regarding scaling and weighting. Although sensitivity checks improved confidence in the results, aggregation may have masked variation among underlying readiness components. Finally, the study focused on two regional contexts rather than national representation; however, the selected regions capture contrasting hazard and livelihood archetypes commonly observed in climate-exposed rural systems.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec32\" class=\"Section2\"\u003e \u003ch2\u003e4.6 Synthesis and broader implications\u003c/h2\u003e \u003cp\u003eThe results demonstrate that household climate risk cannot be inferred from hazard exposure or perceived threat alone. Preparedness emerged from the interaction between hazard experience and structural enabling conditions, consistent with established risk frameworks (Cardona et al., \u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e2012\u003c/span\u003e). Across the two regions, differences in livelihood structure, asset portfolios, and financial access helped explain why exposure and readiness did not align. Under projected increases in hydroclimatic variability across many tropical smallholder regions, exposure is likely to strengthen in ways that do not automatically expand household preparedness (IPCC, \u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). The present findings suggest that without parallel strengthening of structural capacity, intensifying hazard signals may widen perception\u0026ndash;readiness gaps rather than close them.\u003c/p\u003e \u003cp\u003eThe exposure\u0026ndash;readiness quadrant provides a practical diagnostic framework for differentiating household risk profiles. Within this framework, high-exposure/low-readiness households represent contexts where financial and institutional support may be prioritized, whereas high-exposure/high-readiness households may benefit more from anticipatory or information-based interventions. Low-exposure/high-readiness contexts reflect resilience-maintenance situations, while low-exposure/low-readiness households may require longer-term structural investment. Rather than prescribing fixed interventions, this typology offers a decision-support lens for tailoring climate risk management strategies to local household conditions.\u003c/p\u003e \u003cp\u003eThe limited association between early warning access and readiness further reinforces the importance of structural capacity. Climate information appears to function primarily as a reinforcing input rather than an independent driver of preparedness when households lack feasible protective options. Mwangi et al. (\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e2021\u003c/span\u003e) similarly show that the influence of climate information services on decision-making depends on institutional context and user characteristics. Effective climate risk management in low-income rural systems therefore requires integrating climate information services with financial, livelihood, and institutional mechanisms that enable households to translate awareness into action.\u003c/p\u003e \u003c/div\u003e"},{"header":"5. Conclusion","content":"\u003cp\u003eThis study demonstrates that, in a context of intensifying climate variability, household preparedness to climate risk in rural Madagascar is associated more strongly with structural and livelihood conditions than with exposure or perceived risk alone. Across contrasting hazard environments, households in Atsimo Andrefana exhibited slightly higher readiness, likely supported by diversified livelihoods and convertible livestock assets, whereas households in Sava combined higher exposure and stronger risk perception with lower preparedness and more limited financial flexibility. These contrasting profiles indicate that preparedness reflects the interaction between hazard context, livelihood structure, and enabling capacity rather than exposure or perception alone.\u003c/p\u003e \u003cp\u003eThe findings further suggest a decoupling between perceived risk and preparedness when financial and material constraints limit households\u0026rsquo; ability to act. Early warning access showed limited association with readiness once exposure and perceived risk were considered, highlighting that information alone may not translate into preparedness where structural constraints persist. By integrating exposure indicators, risk-perception measures, and a multidimensional readiness index, the study contributes a household-level diagnostic framework that clarifies how behavioral and structural dimensions of climate risk interact in low-income, climate-exposed settings.\u003c/p\u003e \u003cp\u003eThese results imply that expanding climate information services alone is unlikely to strengthen preparedness where households face binding structural constraints. Improving access to credit, protecting productive assets, supporting livelihood diversification, and strengthening institutional support appear central to enabling households to translate risk awareness into effective action. Without reinforcing these structural conditions, increasing hazard exposure under continued warming may heighten awareness without narrowing preparedness gaps.\u003c/p\u003e"},{"header":"Declarations","content":"\u003ch2\u003eAuthor Contribution\u003c/h2\u003e\u003cp\u003eJ.E.J. conceptualized the study, collected the data, conducted the investigation and analysis, developed the software, and wrote the main manuscript text. C.U. contributed to conceptualization, supervision, investigation, and manuscript review. J.R. contributed to data collection and investigation. A.M.W. contributed to conceptualization, supervision, funding acquisition, and manuscript review. A.W. contributed to conceptualization, supervision, and manuscript review. All authors reviewed and approved the final manuscript.\u003c/p\u003e\u003ch2\u003eAcknowledgement\u003c/h2\u003e\u003cp\u003eThis work was funded by the World Bank Group through the Food Systems Resilience Program (FSRP) \u0026ndash; Madagascar. The authors gratefully acknowledge this support, which enabled the completion of the present study.\u003c/p\u003e\u003ch2\u003eData Availability\u003c/h2\u003e\u003cp\u003eThe datasets generated and analyzed during this study are not publicly available due to confidentiality considerations associated with household survey data but are available from the corresponding author on reasonable request.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eAcanga A, Matovu B, Murale V, Arlikatti S (2025) Gender perspectives in disaster response: An evidence-based review. Progress Disaster Sci 26:100416. ttps://doi.org/10.1016/j.pdisas.2025.100416\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAdger WN, Brown K, Nelson DR, Berkes F, Eakin H, Folke C, Galvin K, Gunderson L, Goulden M, O\u0026rsquo;Brien K, Ruitenbeek J, Tompkins EL (2011) Resilience implications of policy responses to climate change. Wiley Interdisciplinary Reviews Clim Change 2(5):757\u0026ndash;766. ttps://doi.org/10.1002/wcc.133\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBarrett CB, Reardon T, Webb P (2001) Nonfarm income diversification and household livelihood strategies in rural Africa: concepts, dynamics, and policy implications. Food Policy 26(4):315\u0026ndash;331. ttps://doi.org/10.1016/s0306-9192(01)00014-8\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBeaman L, Karlan D, Thuysbaert B, Udry C (2014) Selection into Credit Markets: Evidence from Agriculture in Mali. ttps://doi.org/10.3386/w20387\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBirkmann J, Cardona OD, Carre\u0026ntilde;o ML, Barbat AH, Pelling M, Schneiderbauer S, Kienberger S, Keiler M, Alexander D, Zeil P, Welle T (2013) Framing vulnerability, risk and societal responses: the MOVE framework. Nat Hazards 67(2):193\u0026ndash;211. ttps://doi.org/10.1007/s11069-013-0558-5\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBrooks N, Adger WN, Kelly PM (2005) The determinants of vulnerability and adaptive capacity at the national level and the implications for adaptation. Glob Environ Change 15(2):151\u0026ndash;163. ttps://doi.org/10.1016/j.gloenvcha.2004.12.006\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCardona O, Van Aalst MK, Birkmann J, Fordham M, McGregor G, Perez R, Pulwarty RS, Schipper ELF, Sinh BT, D\u0026eacute;camps H, Keim M, Davis I, Ebi KL, Lavell A, Mechler R, Murray V, Pelling M, Pohl J, Smith A, Thomalla F (2012) Determinants of risk: exposure and vulnerability. In Cambridge University Press eBooks (pp. 65\u0026ndash;108). ttps://doi.org/10.1017/cbo9781139177245.005\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCarter MR, Barrett CB (2006) The economics of poverty traps and persistent poverty: An asset-based approach. J Dev Stud 42(2):178\u0026ndash;199. ttps://doi.org/10.1080/00220380500405261\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCinner JE, Adger WN, Allison EH, Barnes ML, Brown K, Cohen PJ, Gelcich S, Hicks CC, Hughes TP, Lau J, Marshall NA, Morrison TH (2018) Building adaptive capacity to climate change in tropical coastal communities. Nat Clim Change 8(2):117\u0026ndash;123. ttps://doi.org/10.1038/s41558-017-0065-x\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eConstable AJ, French S, Karoblyte V, Viner D (2022) Decision-Making for managing Climate-Related risks: Unpacking the decision process to avoid Trial-and-Error responses. Front Clim. 4ttps://doi.org/10.3389/fclim.2022.754264\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDoan MK, Hill R, Hallegatte S, Corral P, Brunckhorst B, Nguyen M, Freije-Rodriguez S, Naikal E (2023) Counting people exposed to, vulnerable to, or at high risk from climate shocks: a methodology. In World Bank. ttps://doi.org/10.1596/1813-9450-10619. Washington, DC eBooks\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eErman A, De Vries Robb\u0026eacute; SA, Thies SF, Kabir K, Maruo M, International Bank for Reconstruction and Development / The World Bank (2021) Gender dimensions of disaster risk and resilience. International Bank for Reconstruction and Development / The World Bank\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGinbo T, Hansson H (2023) Intra-household risk perceptions and climate change adaptation in sub-Saharan Africa. Eur Rev Agric Econ 50(3):1039\u0026ndash;1063. ttps://doi.org/10.1093/erae/jbad011\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eIPCC (2022) Climate Change 2022: Impacts, adaptation and Vulnerability: Contribution of Working Group II to the Sixth Assessment Report of the Intergovernmental Panel on Climate Change (H.-O. P\u0026ouml;rtner, D. C. Roberts, M. M. B. Tignor, E. Poloczanska, \u0026amp; K. Mintenbeck, Eds.). Cambridge University Press. ttps://doi.org/10.1017/9781009325844\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eIslam N, Winkel J (2017) Climate change and social inequality (DESA Working Paper No. 152). UNITED NATIONS Department of Economic and Social Affairs. Retrieved November 27, 2025, from \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.un.org/esa/desa/papers/2017/wp152_2017.pdf\u003c/span\u003e\u003cspan address=\"https://www.un.org/esa/desa/papers/2017/wp152_2017.pdf\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJanzen SA, Carter MR (2018) After the Drought: The impact of microinsurance on consumption smoothing and asset protection. Am J Agric Econ 101(3):651\u0026ndash;671. ttps://doi.org/10.1093/ajae/aay061\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJones L, Tanner T (2016) Subjective resilience\u0026rsquo;: using perceptions to quantify household resilience to climate extremes and disasters. Reg Envriron Chang 17(1):229\u0026ndash;243. ttps://doi.org/10.1007/s10113-016-0995-2\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKaila H, Singhal S, Tuteja D (2020) Development programs, security, and violence reduction: Evidence from an insurgency in India. World Dev 130:104911. ttps://doi.org/10.1016/j.worlddev.2020.104911\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMartenies SE, Zhang M, Corrigan AE, Kvit A, Shields T, Wheaton W, Him DA, Aschner J, Talavera-Barber MM, Barrett ES, Bastain TM, Bendixsen C, Breton CV, Bush NR, Cacho F, Camargo CA, Carroll KN, Carter BS, Cassidy-Bushrow AE, Kress AM (2023) Developing a National-Scale Exposure Index for Combined Environmental Hazards and Social Stressors and Applications to the Environmental Influences on Child Health Outcomes (ECHO) cohort. Int J Environ Res Public Health 20(14):6339. ttps://doi.org/10.3390/ijerph20146339\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMiller F, Osbahr H, Boyd E, Thomalla F, Bharwani S, Ziervogel G, Walker B, Van Der Birkmann J, Rockstr\u0026ouml;m J, Hinkel J, Downing T, Folke C, Nelson D (2010) Resilience and vulnerability: complementary or conflicting concepts? Ecol Soc 15(3). ttps://doi.org/10.5751/es-03378-150311\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMwangi M, Kituyi E, Ouma G (2021) A systematic review of the literature on the contribution of past climate information services pilot projects in climate risk management. Sci Afr 14:e01005. ttps://doi.org/10.1016/j.sciaf.2021.e01005\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eNguyen G, Nguyen TT (2019) Exposure to weather shocks: A comparison between self-reported record and extreme weather data. Econ Anal Policy 65:117\u0026ndash;138. ttps://doi.org/10.1016/j.eap.2019.11.009\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePescaroli G, Dryhurst S, Karagiannis GM (2025) Bridging gaps in research and practice for early warning systems: new datasets for public response. Front Communication 10. ttps://doi.org/10.3389/fcomm.2025.1451800\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePret V, Falconnier GN, Affholder F, Corbeels M, Chikowo R, Descheemaeker K (2025) Farm resilience to climatic risk. A review. Agron Sustain Dev 45(1):10. ttps://doi.org/10.1007/s13593-024-00998-w\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRavera F, Mart\u0026iacute;n-L\u0026oacute;pez B, Pascual U, Drucker A (2016) The diversity of gendered adaptation strategies to climate change of Indian farmers: A feminist intersectional approach. Ambio 45(S3):335\u0026ndash;351. ttps://doi.org/10.1007/s13280-016-0833-2\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRokhideh M, Fearnley C, Budimir M (2025) Multi-Hazard Early Warning Systems in the SenDai Framework for Disaster Risk Reduction: Achievements, gaps, and future directions. Int J Disaster Risk Sci 16(1):103\u0026ndash;116. ttps://doi.org/10.1007/s13753-025-00622-9\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSchatte P, Meyer M (2025) Assessing holistic agroecological resilience of agroecosystems from a landscape perspective: a systematic review. Ecol Soc 30(2). ttps://doi.org/10.5751/es-16137-300224\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSullivan-Wiley KA, Gianotti AGS (2017) Risk perception in a Multi-Hazard environment. World Dev 97:138\u0026ndash;152. ttps://doi.org/10.1016/j.worlddev.2017.04.002\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eStevens SM, Joy MK, Abrahamse W, Milfont TL, Petherick LM (2023) Composite environmental indices\u0026mdash;a case of rickety rankings. PeerJ 11:e16325. ttps://doi.org/10.7717/peerj.16325\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003evan Asselt J, Aung ZW (2025) Community and household shocks and coping strategies: Findings from the ninth round of the Myanmar Household Welfare Survey (April - October 2025). Myanmar SSP Working Paper 74. Washington, DC: International Food Policy Research Institute. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://hdl.handle.net/10568/179640\u003c/span\u003e\u003cspan address=\"https://hdl.handle.net/10568/179640\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eVan Der Lee J, Kangogo D, G\u0026uuml;lzari Ş\u0026Ouml;, Dentoni D, Oosting S, Bijman J, Klerkx L (2022) Theoretical positions and approaches to resilience assessment in farming systems. A review. Agron Sustain Dev 42(2). ttps://doi.org/10.1007/s13593-022-00755-x\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eVan Valkengoed AM, Perlaviciute G, Steg L (2023) From believing in climate change to adapting to climate change: The role of risk perception and efficacy beliefs. Risk Anal 44(3):553\u0026ndash;565. ttps://doi.org/10.1111/risa.14193\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWachinger G, Renn O, Begg C, Kuhlicke C (2012) The Risk Perception Paradox\u0026mdash;Implications for governance and communication of Natural Hazards. Risk Anal 33(6):1049\u0026ndash;1065. ttps://doi.org/10.1111/j.1539-6924.2012.01942.x\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"},{"header":"Footnotes","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003e More information about the project: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://ewsdata.rightsindevelopment.org/files/documents/66/WB-P178566.pdf\u003c/span\u003e\u003cspan address=\"https://ewsdata.rightsindevelopment.org/files/documents/66/WB-P178566.pdf\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"Climate risk, exposure, early warning systems, Madagascar, resilience, Shock Readiness Index (SRI)","lastPublishedDoi":"10.21203/rs.3.rs-9091846/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-9091846/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eHousehold climate risk in Madagascar reflects the interaction of hazard exposure, perceived risk, and structural capacity to prepare for shocks. As climate variability and extremes intensify across tropical smallholder regions, understanding how these dimensions align at the household level becomes increasingly important. This study examines these dynamics in two contrasting regions, cyclone-exposed Sava and drought-prone Atsimo Andrefana. Using data from 245 stratified households, we integrate a composite Exposure Index, perceived likelihood and severity measures, and a multidimensional Shock Readiness Index capturing assets, credit access, and financial buffers.\u003c/p\u003e \u003cp\u003eResults show a decoupling between exposure, perceived risk, and readiness. In Sava, households report higher exposure and perceived risk but lower readiness, associated with limited credit access, liquidity constraints, and specialized crop-based livelihoods. In contrast, households in Atsimo Andrefana perceive lower risk yet demonstrate comparatively higher readiness, supported by diversified livelihoods and convertible livestock assets. These patterns indicate a perception\u0026ndash;readiness gap in which heightened awareness does not translate into preparedness under binding structural constraints. Early warning access shows limited association with readiness, suggesting that information alone has limited influence where response options are constrained.\u003c/p\u003e \u003cp\u003eBy integrating perceived risk with structural preparedness, the study clarifies why preparedness does not systematically track exposure in climate-exposed rural systems. In the context of ongoing global climate change, the findings suggest that strengthening structural capacity through improved financial access, livelihood diversification, and institutional support will be critical to narrowing preparedness gaps in smallholder regions beyond Madagascar.\u003c/p\u003e","manuscriptTitle":"Decoupling exposure, risk perception, and preparedness in smallholder systems: Evidence from rural Madagascar","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-04-15 09:48:09","doi":"10.21203/rs.3.rs-9091846/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"86c9e8f9-ab01-4d64-8691-8e2f49a71eff","owner":[],"postedDate":"April 15th, 2026","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[],"tags":[],"updatedAt":"2026-04-15T09:48:09+00:00","versionOfRecord":[],"versionCreatedAt":"2026-04-15 09:48:09","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-9091846","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-9091846","identity":"rs-9091846","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}