Gender inequality in the impact of climate shock on physical health and its spillover effect

Abstract

Background

Physical health disparities between male and female constitute a significant yet underexplored manifestation of gender inequality, a challenge increasingly exacerbated by climate shock.

Methods

This paper explores the gender inequality in the impact of climate shock on physical health and its spillover effect. We develop a theoretical framework of intra-household resource allocation, and employ micro-level panel datasets matched with a climate physical risk dataset to estimate the impacts of climate shock using panel fixed-effect models.

Results

We find that climate shock significantly reduces the activity of daily living and health expenditure, with female disproportionately affected. In male-preferred households, the deterioration of male’s activity of daily living generates negative spillovers that further undermine female’s activity of daily living, whereas in households without a strong male preference, this spillover is attenuated. Channel analysis shows that climate shock impairs the activity of daily living through reduced exercise and social participation. Heterogeneity analysis reveals that the physical health-reducing effects of climate shock are more pronounced among rural, low-educated, unmarried and unretired subgroups. Improving access to local employment opportunities and healthcare resources is effective in enhancing health adaptation to climate shock.

Conclusions

These findings underscore the need for gender-responsive adaptation strategies that prioritize equitable access to healthcare resources to strengthen physical health resilience under escalating climate risks.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12913-026-14791-w.

Introduction

Promoting universal health and functional well-being remains a cornerstone of global development, as anchored in the United Nations’ Sustainable Development Goals 3. However, the escalating climate crisis has emerged as a systemic threat to these objectives, disproportionately eroding the physical health and functional capacity of vulnerable populations. Activities of Daily Living (ADL), a critical barometer of physical autonomy and quality of life, are increasingly compromised by climate-induced physiological stress and restricted mobility [1]. While climate physical risks are exogenous and ubiquitous, their physical health manifestations are far from uniform. Growing evidence suggests that gendered vulnerabilities, rooted in physiological differences, socio-economic disparities, and intra-household power dynamics, dictate the “unequal burden” individuals bear under identical climate exposure [2]. In low- and middle-income countries, where institutional support is often fragmented, these physical health disparities not only diminish individual productivity but also deepen systemic gender inequality, necessitating a profound understanding of how climate shock reshapes the landscape of physical health across genders [3]. Crucially, this paper argues that the household is not merely a passive recipient of climate shocks but an active site where pre-existing gender preferences determine who bears the health cost and who receives financial protection. Therefore, this study aims to uncover the intra-household spillover effect, specifically gendered resource allocation, that transform a universal climate shock into a gendered physical health penalty.

Extensive research has documented the physiological toll of climate shock, traditionally focusing on mortality rates or acute morbidities such as cardiovascular and respiratory diseases. However, a nascent but vital body of literature is shifting attention toward functional health indicators, particularly ADL, which serve as a more comprehensive measure of long-term climate resilience and human capital maintenance [4, 5]. While the direct physical impacts of extreme temperatures and environmental shocks are increasingly recognized, existing studies often treat individuals as isolated units, neglecting the complex intra-household dynamics that mediate these impacts [6]. This black box of the household is where gender inequalities are most acutely manifested. Yet the literature has largely focused on aggregate socio-economic outcomes rather than the granular reallocation of health resources within the domestic sphere [7, 8]. Specifically, the potential for physical health spillovers, where the functional deterioration of one household member, particularly a male, triggers a chain reaction of resource competition or increased caregiving burdens that further undermine the physical health of female members, remains a critical yet underexplored dimension of climate-induced inequality.

Complementing the erosion of functional capacity, the financial strain of climate shock serves as a secondary yet equally potent driver of physical health inequality. Extreme weather events inevitably escalate healthcare costs as households attempt to mitigate physiological damage and functional impairment [9]. However, these financial burdens are rarely distributed equitably within the domestic sphere. In resource-constrained environments, healthcare expenditures often exhibit a pro-male bias, where health-related investments are prioritized for male members as they are frequently perceived as the primary economic pillars of the family [10, 11]. This creates a precarious double burden for women: they are simultaneously more vulnerable to the functional decline triggered by climate shocks and more likely to be marginalized during the subsequent allocation of remedial medical resources [12]. Existing scholarship, while acknowledging the aggregate surge in health costs, has yet to sufficiently scrutinize the disconnect between physical impairment and financial mitigation at the micro-level. Specifically, how gendered preferences transform a universal environmental risk into an asymmetrical health penalty remains a critical blind spot that this study aims to illuminate.

This research seeks to provide a granular analysis of how climate shock influences physical health and health expenditure, with a specific focus on gender inequality and intra-household spillover. Our study is structured four primary objectives. First, we study the gender inequality in the impact of climate shock on physical health and health expenditure. Second, we rigorously examine the intra-household spillover effects, exploring how the functional deterioration of male members, conditioned by gendered allocation preferences, asymmetrically exacerbates the physical health burdens of females. Third, we delineate the channels through which climate shock impair physical health, specifically investigating the roles of curtailed physical exercise and social activity. Finally, we assess the role of improved access to local employment opportunities and healthcare resources in mitigating the gender-differentiated physical health impacts of climate shock, as well as its effectiveness in supporting health adaptation to climate shock.

This paper has three primary contributions. First, we redefine the climate-health nexus by focusing on ADL, a structural measure of human capital, rather than subjective well-being. By utilizing the intra-household bargaining perspective, we move beyond individual-centric analyses to uncover the health-financial disconnect, the phenomenon where climate-induced physical health impairment in females fails to trigger the same compensatory financial protection as observed in males. This provides a more rigorous empirical basis for understanding how intra-household gender preference transform universal climate shock into physical health inequality. Second, this study demonstrates that climate shock doesn’t merely direct affect physical health, but specifically through the restriction of exercise and social activity. Third, we evaluate the framework of health adaptation to climate shock. We provide evidence that the restoration of physical health is more contingent on the accessibility of healthcare resources than on employment. This distinction is vital for developing gender-responsive health adaptation to climate shock that targets the specific physical health vulnerabilities exacerbated by climate shock.

Conceptual framework

To better guide the empirical analysis, we establish a conceptual framework to model gender inequality in the effects of climate shock on individuals’ physical health and health expenditure [13, 14]. Specifically, we explore how gender preferences shape both the direction and magnitude of these effects.

We begin with a simplified model of intra-household resource allocation, where haggle over price to decide the savings and health expenditure of male and female. The model considers a household with two individuals, including a male and a female. Following [15], we assume that households possess a utility function (U^T), which is separable and depends on sub-utility derived from earnings (E) and other outcomes of interest (C) [16–18]. The earnings E is defined over the earnings of two individuals (E_f and E_m), which depend on their physical health (H_f and H_m). We define E_i = θ_iH_i for i = f (female), m (male). The return to health expenditure θ_i is a productivity parameter that translates physical health into earnings, reflecting the standard assumption in health economics that better physical health enhances labor productivity and thus income. θ_i can vary depending on the individual’s gender. Other outcomes of interest C are captured in the composite investment good I within the budget constraint, representing consumption and non-health investments. Accordingly, the households’ sub-utility from earnings can be expressed as:

1

Households decide how to allocate financial resources between individuals’ health expenditure and general consumption through intra-household bargaining. Given a total resource pool of y for family health investment, they aim to maximize sub-utility function U. Therefore, for ease of illustration, we focus on the maximization problem of the sub-utility function U subject to the budget constraint y. Individuals also face a monetary constraint where they divide their monetary (M_i) between health expenditure (p_iH_i) and investment (I_i). Here, H_idenotes the quantity of health investment received by individual i, while p_i represents the unit cost (price) of health investment, so that total health expenditure equals p_iH_i. Investment (I_i) in the model represents a composite good that encompasses all other uses of monetary resources beyond health expenditure, including both consumption of non-health goods and services as well as savings or financial investments, thereby ensuring the model accounts for the full allocation of household income. In this model, we ignore the sequential nature of the decision-making and treat health expenditure allocation as a one-period problem, taking the number of gender composition as given.

Based on this set-up, the sub-utility optimization problem of households can be formulated as

2

where for i = f, m. γ_f and γ_m capture the household’s priority for female and male members, respectively. They are a combination of the household’s altruistic parameters (δ_f and δ_m) and returns to physical health (θ_f and θ_m) for the female and male. η_i captures the effective return to health investment for individual_i, reflecting how improvements in physical health translate into utility-relevant outcomes (e.g., productivity or earnings). The altruistic parameters (δ_f and δ_m) capture the household’s intrinsic valuation of the well-being of female and male members, respectively, which is central to our analysis of gender preferences in intra-household resource allocation. Therefore, a stronger preference for males may be influenced by intra-household bargaining being more altruistic towards the male (δ_m > δ_f), or the male having a higher physical health return rate (η_m > η_f). σ∈ (−∞, 1] captures the inequality aversion of households. A smaller σ indicates that households are more inequality averse. Individual’s returns on non-health investment are depended by r_f and r_m. The variables p_i and p_m represent the unit costs of physical health investment for females and males, respectively, and thus govern total health expenditure through p_iH_i. Individuals’ health expenditure may vary by their physical health level. Following exposure to climate shock, individuals whose physical health is less affected by climate shock face lower health expenditure than those whose family members are less affected by climate shock.

The derived demand functions of individuals’ physical health are and , where is the maximum investment that a family can earn when neither individual requires medical treatment. The effective costs are and . The corresponding utility-compensated demand functions that minimize the expenditure with respect to a given level of utility U are denoted by and , where subscript h represents the Hicksian demand.

The model set-up allows us to describe the direct and spillover effects of climate shock, which is equivalent to the own- and cross-price effects of changes in female physical health ( and ). We are also interested in how own- and cross-price effects vary by household’s male preference, captured by γ_m. Using the household optimization problem outlined in Eq. (2), we derive following equations:

3

4

5

where . The derivation of these equations is provided in Appendix a.

Based on Eqs. (3), (4) and (5), we generate predictions regarding the impacts of climate shock:

PROPOSITION 1.

A decline in physical health levels caused by climate shock makes resident spend more money on health expenditures and less on investments.

PROPOSITION 2.

When σ <0, a rise in male health spending has a positive spillover effect on female physical health outcomes. When 0 < σ <1, a rise in male health spending has a negative spillover effect on female physical health outcomes.

PROPOSITION 3.

When the physical health outcomes of the two individuals are substitutes (0 < σ <1) andholds, the negative spillover effect on females increases as household preference for females decreases.

Proofs of these propositions are discussed in Appendix A. It is important to note that the empirical analysis that follows does not attempt to structurally estimate the parameters σ or γ. Instead, the theoretical framework is used to derive qualitative comparative statics, which guide the interpretation of reduced-form empirical patterns.

To better understand the model’s predictions of the direct and spillover effects of climate shock, Fig. 1 illustrates simulated climate shock impacts on health expenditure under different gender preference scenarios. The observed difference in health expenditure between male and female increases with climate shock severity. However, heterogeneity exists in the degree of expenditure change across genders under different gender preference scenarios. In the male-preference scenario (Panel a), marginal health expenditures are higher for male than for female under equivalent climate shock conditions. Conversely, in the female-preference scenario (Panel b), males exhibit lower marginal health expenditures than females at the same level of climate exposure. The magnitude of gender-based expenditure differences, as climate shock intensifies, quantifies how the utility function depends on households’ gender preferences.

Fig. 1.

Illustration of climate-dependent health expenditure functions in different gender preference scenarios

Data and variables

Data source

Data

This study primarily relies on longitudinal microdata from China Health and Retirement Longitudinal Study (CHARLS), a nationally representative survey administered by Peking University. CHARLS is designed to collect information on the demographic characteristics, health status, healthcare utilization, and socioeconomic conditions of respondents in China, as well as their households. The survey adopts a multi-stage, stratified probability sampling strategy, covering 450 village-level units across 150 county-level administrative divisions in mainland China, thereby ensuring broad geographic and socioeconomic representativeness. Data are collected at the individual, household, and community levels, allowing for detailed analysis of intra-household resource allocation. To construct a balanced and consistent panel suitable for longitudinal analysis, we combine five consecutive survey waves (2011, 2013, 2015, 2018, and 2020). All waves are harmonized following the official CHARLS panel guidelines to ensure consistency in variable definitions and measurement over time. This multi-wave structure enables us to exploit within-individual variation and to identify the dynamic effects of climate shock on physical health and health expenditure.

Ethical considerations

The Biomedical Ethics Committee of Peking University have verified that the study complies with research ethics (ethical code: IRB00001052-11015). All procedures involving human participants adhered to ethical standards, including the Declaration of Helsinki. Before holding the interviews, the participants were asked to sign an informed consent form if they were willing to participate in the study. The participants were also informed that they were free to withdraw at any point of the research and the time of the interviews would be set by their agreement. For performing the observations in accordance with the ethical considerations, the participants were observed with prior notice and in an overt manner.

Measures of variables and descriptive statistics

Climate shock

We use a climate physical risk (CPR) index as a proxy for regional climate shock [19]. Four types of extreme climate, including extreme low temperature days (LTD), extreme high temperature days (HTD), extreme rainfall days (ERD), and extreme drought day (EDD), are measured by constructing algorithms in Appendix B. Since the four extreme climate measures differ in nature, direct comparisons are not feasible. To address this, we apply the min-max standardization approach to normalize the data and generate a general index within each category. Taking LTD as an example:

6

where K represents the total number of cities, and L is the number of years in sample set. m represents the mth city, and n represents the nth year. In this way, four sub-indices LTD, HTD, ERD and EDD can be obtained.

Once standardized, four sub-indices are used to construct a general CPR index for each city. Here the simple average is used such as:

7

where, ω_iis set to 0.25. Obviously, weights can be adjusted to serve any individual purpose in a particular study.

Detailed descriptive statistics and classification metrics for the CPR index across the surveyed urban centers are presented in Table C1. The longitudinal progression captured in Column 5 highlights the evolving nature of climate physical risks during the observation window. To further elucidate these trends, Fig. 2 employs kernel density estimation to depict the distributional heterogeneity of CPR, while Figure 3 provides a geospatial mapping of its variation between 2010 and 2020. The results underscore a pronounced geographical imbalance, with the western interior exhibiting significantly higher risk exposure than the eastern coastal regions. This East-West divide has notably widened over the decade, driven by an accelerated intensification of climate stressors in western territories. The visual evidence in Fig. 3 further identifies significant spatial autocorrelation, revealing persistent hotspots where climate risks are increasingly concentrated, thereby setting the stage for our analysis of functional physical health vulnerabilities.

Fig. 2.

Distribution of climate shock in China

Fig. 3.

The CPR by city in 2010 and 2020. Blanks indicate the presence of missing values

Dependent variable is the physical health of respondents, which is assessed through the activities of daily living (ADL) [20, 21]. ADL consists of six key activities: can or can’t dress (Dress), bath (Bath), eat (Dine), getting in and out of bed (Get-up), go to the toilet (WC), and pee and poo (Relieve), independently. To assess ADL, each activity is assigned a value of 1 if the individual is able to perform it independently without assistance. Scores for all six activities are then added together. The total score ranges from a maximum of 6, indicating complete independence in all activities, to a minimum of 0, indicating that assistance is required for each activity.

We operationalize health expenditure using respondents’ total medical costs over the past month, serving as a proxy indicator for respondents’ resource investments in physical health improvement [22]. This measure inherently assumes that healthcare utilization reflects intentional efforts to address physical health well-being. We interpret higher health expenditure as a household’s revealed preference to invest in an individual’s physical health recovery. However, we acknowledge it also correlates with illness severity, making the contrast between rising expenditure for male and stagnant expenditure for female particularly telling of allocation preferences. We analyze physical health and health expenditure within households by classifying members into six demographic groups. Table 1 documents an observable pattern that female demonstrate higher physical health levels but lower health expenditure compared to male.

Table 1.

Physical health and health expenditure by age and sex

	Boys	Girls	Adult men	Adult women	Elderly men	Elderly women
ADL	5.78	5.00	5.58	5.71	5.37	5.57
	(0.67)	(1.73)	(1.03)	(0.89)	(1.24)	(1.10)
Health expense	0.00	0.00	230.92	189.00	269.98	264.23
	(0.00)	(0.00)	(2526.97)	(2861.81)	(1961.51)	(1941.48)
Observations	10	6	24320	20095	19563	18110

Notes: Boy (0–18 years), girl (0–18 years), adult men (18–60 years), adult women (18–60 years), older men (60+ years), and older women (60+ years). Numbers in parentheses are standard deviations. The samples for boys and girls are smaller. The samples for boys and girls are notably smaller, as these observations are derived from the household rosters of the primary adult respondents in the CHARLS dataset. While these subgroups provide valuable insights into intra-household structure, their limited size necessitates cautious interpretation of group-specific estimates. Nevertheless, their inclusion allows for a more comprehensive analysis of climate shock impacts across all age segments within families

Control variables

Estimation models also control for age (Age), gender (Gender), educational level (Education), marital status (Marriage), retirement status (Retire) and whether or not alcohol (Drink) is consumed at the individual level. In addition, the model controls for other household-level characteristics: place of residence (Rural), household size (Size), annual household consumption per capita (Expense) and whether or not broadband is installed (Broadband) [23]. Related statistics are provided in Appendix Table C2.

Empirical strategy

To rigorously identify the causal nexus between climate shock and physical health, we employ a high-dimensional fixed-effects panel framework [14]. The CPR metric is constructed to capture the annual frequency of extreme weather days, defined as periods where localized meteorological parameters breach region-specific historical thresholds. A fundamental premise of our identification strategy lies in the decoupling of long-run climatic shifts from short-term idiosyncratic shocks. Unlike secular climate trends which may exhibit predictable spatial patterns, these threshold-exceeding events are characterized by idiosyncratic temporal volatility and are serially uncorrelated. This inherent stochasticity ensures that such weather-driven shocks are orthogonal to unobserved individual or regional characteristics, satisfying the requirements for treating them as exogenous variations in our baseline specification.

8

In our primary specification, Y_i,h,c,t denotes the outcomes of interest—physical functional capacity (measured by ADL scores) and healthcare expenditure—for individual i in household h within city c at time t. The model incorporates a comprehensive vector of Control_i,h,c,t to account for time-varying individual and household-level characteristics. To purge the estimates of latent confounding factors, we integrate a multi-level fixed-effect structure: δ_t captures macro-level temporal shocks common to all units, while λ_i and η_c absorb time-invariant unobservables at the individual and city levels, respectively. Standard errors are conservatively clustered at the individual level to provide robust inference against potential heteroskedasticity and serial correlation. While the coefficient β yields the average treatment effect, it may obscure the asymmetric vulnerabilities existing within the domestic sphere. To unveil these hidden disparities, we extend Eq. (8) by introducing an interaction framework between CPR and a series of categorical variables representing distinct age-gender cohorts, thereby facilitating a granular analysis of how climate shocks disproportionately reshape the physical health and resource allocation of specific household members.

9

where Gender_i,h,c,t is a categorical variable, the interaction term CPR_c,t×Gender_i,h,c,t×Age_i,h,c,t captures the direct influence of climate shock on specific groups. Equation (10) provides the marginal impacts of climate shock on physical health and health expenditures for different age-specific gender differences.

10

where, ρ ∈ {1, … , 6} denotes age-gender groups. Household gender preferences intensify the impact of external climate shock on health expenditure inequality. To explore the gender preference inequality in health consumption smoothing, we introduce a categorical variable into Eq. (11) for household gender composition exclude oneself, interacting it with climate shock and individual’s age group.

11

where Inter_Gender_i,h,c,t serves as binary indicators taking the value 1 if there is a member of the opposite sex within household, excluding individual i, and 0 otherwise. The partial derivatives of Eq. (12) with respect to climate shock give the marginal impact of climate shock on physical health and healthcare expenditure for different household gender preferences. The four-way interaction term CPRc,t×Genderi,ℎ,c,t×Agei,ℎ,c,t×Inter_Genderi,ℎ,c,tcaptures how the effect of climate shock on individual outcomes varies jointly by the individual’s gender, age, and the household’s gender composition. Specifically, the coefficient β₁₁measures the differential impact of CPR on outcome Y for individuals of a given gender-age group who live in households with members of the opposite sex, relative to otherwise comparable individuals in same-gender households. Therefore, β₁₁reflects whether and how intra-household gender structure moderates the gender- and age-specific effects of climate shock.

12

In scenarios where gender preferences are pronounced, if the marginal effects of climate shock are more substantial, it suggests that these gender preferences do not aid in lessening the uneven gender-specific impacts of climate shock on physical health and healthcare expenditure. Our reduced-form approach does not aim to recover structural preference parameters, but rather to assess whether observed behavioral responses to climate shocks align with the theoretical implications of gendered intra-household allocation.

We estimate the baseline model using a panel fixed-effect model, which allows us to control for time-invariant unobserved heterogeneity across individuals and common shocks over time. This approach is well suited to our setting, as it mitigates bias arising from omitted factors that are constant within individuals but correlated with both climate exposure and physical health outcomes. However, the two-way fixed effect estimates may be sensitive to heterogeneous treatment effects across groups and time. To address these limitations, we complement the baseline specification with a series of robustness checks.

The impact of climate shock on physical health

Impact of climate shock on physical health and health expenditure

Table 2 reports the estimated effects of climate shock on ADL and health expenditure based on Eq. (8). Columns (1)-(3) show that a one-unit increase in the CPR index is associated with a statistically significant decrease in ADL. The estimated coefficient remains stable across specifications, ranging from −0.003 to −0.003, and is robust to the inclusion of individual, household, and city fixed effects. Columns (4)-(6) present the corresponding results for health expenditure. The estimates indicate that higher CPR is associated with a significant increase in healthcare expenditure, with coefficients between 7.492 and 7.941, suggesting that climate-related risks raise households’ health-related financial burden. These results indicate that climate shock exposure is associated with deteriorating physical health and increased health expenditure, providing empirical support for PROPOSITION 1.

Table 2.

Estimated effects of climate shock on ADL and health expenditure

	ADL			Health expenditure
	(1)	(2)	(3)	(4)	(5)	(6)
CPR	−0.003***	−0.003***	−0.003***	7.492**	7.940*	7.941*
	(0.001)	(0.001)	(0.001)	(3.764)	(4.045)	(4.050)
Survey FE	Yes	Yes	Yes	Yes	Yes	Yes
Individual FE	Yes	Yes	Yes	Yes	Yes	Yes
City FE			Yes			Yes
Individual controls	Yes	Yes	Yes	Yes	Yes	Yes
Household controls		Yes	Yes		Yes	Yes
Observations	41327	38333	38333	41097	38180	38180
R2	0.588	0.587	0.587	0.392	0.437	0.437

Notes: Robust standard errors clustered at individual level are reported in parentheses. * p < 0.1, ** p < 0.05, *** p < 0.01

Although the marginal effect of CPR on ADL appears numerically small, it is economically meaningful when placed in context. The standard deviation of ADL in the sample is 1.065 implying that a one-unit increase in CPR corresponds to approximately 0.28% of a standard deviation decline in functional health. Moreover, given that ADL is a bounded index capturing discrete losses in basic daily functioning, even modest average shifts can translate into a non-negligible increase in the probability of functional impairment, particularly among older individuals whose baseline ADL is already close to the disability threshold. When exposure to climate shocks persists over multiple periods, these effects may accumulate, resulting in substantial deterioration in functional independence.

Gendered impacts of climate shock on physical health and health expenditure

Examining the gendered impacts of climate shock on intra-household physical health and health expenditure is critical, particularly given persistent gender and age disparities in vulnerability and resource allocation. Significant differences in physical health responses across household members necessitate a disaggregated analysis to better understand distributional consequences and inform targeted policy interventions. To quantify heterogeneous effects across age-gender cohorts, we estimate Eq. (9), with Table 3 reporting interaction-adjusted marginal effects of CPR.

Table 3.

Marginal effects of climate shock on ADL and health expenditure by age and gender cohorts

	Boys	Girls	Adult men	Adult women	Elderly men	Elderly women
	(1)	(2)	(3)	(4)	(5)	(6)
Panel A Dependent variable: ADL
CPR	−0.009***	0.025	−0.004***	−0.002	−0.004**	−0.003**
	(0.001)	(0.038)	(0.001)	(0.002)	(0.002)	(0.002)
Observations	38353
R2	0.587
Panel B Dependent variable: health expenditure
CPR	16.713***	16.600**	13.122**	4.316	10.718*	1.839
	(6.113)	(7.968)	(5.356)	(7.968)	(5.736)	(5.853)
Observations	40741
R2	0.392

Notes: Individual fixed effect, survey fixed effect, city fixed effect, and household and individual control variables are controlled. Robust standard errors clustered at individual level are reported in parentheses. * p < 0.1, ** p < 0.05, *** p < 0.01

Panel A reports the effects of CPR on ADL. The results reveal pronounced heterogeneity across demographic groups. CPR is associated with statistically significant deteriorations in ADL among adult men, elderly men, and elderly women. By contrast, the estimated effect for boys is negative but relatively small in magnitude, indicating that although boys’ physical health status worsens under climate shock, the extent of deterioration is substantially more limited than that observed among older cohorts. For girls, the estimated coefficient is positive and statistically insignificant, suggesting no robust adverse effect of CPR on ADL. This pattern can be rationalized by age- and gender-specific vulnerability profiles. Children, especially boys, typically exhibit stronger baseline physical resilience and faster physiological recovery, which may buffer the functional health consequences of climate shock. In addition, boys often face fewer caregiving responsibilities and lower cumulative exposure to chronic health risks within households, limiting the extent to which climate stressors translate into observable declines in daily functioning. In contrast, adults and the elderly are more likely to suffer from pre-existing physical health conditions and cumulative exposure to environmental stress, making their ADL more sensitive to climate risks.

Panel B presents the corresponding effects on health expenditure. CPR induces significant increases in healthcare expenditure for boys and girls of similar magnitude, indicating that households respond to climate shock affecting children through increased medical utilization regardless of gender. While households appear willing to increase healthcare spending for both boys and girls in response to climate risks, the returns to such spending are heterogeneous. For boys, increased expenditure does not fully compensate for heightened vulnerability to climate-induced functional health shocks, whereas for girls, similar expenditure levels may be sufficient to prevent measurable deterioration in ADL. However, for adults, particularly women, the rise in health expenditure is smaller or statistically insignificant, despite larger deteriorations in ADL. This mismatch suggests that increased physical health needs under climate stress do not always translate into proportionate healthcare investments, especially for female and older household members. These findings indicate that climate shock affects physical health status and healthcare utilization through distinct age- and gender-specific channels. The relatively muted deterioration in boys’ ADL reflects lower biological vulnerability and stronger resilience, whereas the more pronounced declines observed among adults and the elderly highlight accumulated physical health risks and unequal intra-household allocation of health resources. These results underscore the importance of age- and gender-responsive adaptation policies and provide empirical support for PROPOSITION 2, where 0<σ < 1 implies that a rise in male health spending has a negative spillover effect on female physical health outcomes.

Intra-household spillover effects of climate shock impacts on physical health and health expenditure

To explore intra-household spillover effects of climate shock on physical health and health expenditure, we classify individuals according to whether they co-reside with members of the opposite sex. Specifically, male individuals are divided into those living with at least one female household member and those living without females, while female individuals are divided into those living with at least one male household member and those living without males. This classification is motivated by the theoretical framework of intra-household resource allocation under gender preference. In households where both genders are present, climate-induced physical health shocks may trigger reallocations of limited health resources across members, with prioritization shaped by gender-based preferences. In contrast, in same-gender households, such gendered allocation constraints are absent, allowing us to isolate the role of gender composition in mediating intra-household spillovers. While spillovers may also exist among individuals of the same gender, our focus on intra-household spillovers related to gender preferences, which emphasizes that unequal treatment arises primarily along gender preference rather than within-gender age hierarchies. Examining opposite-gender co-residence therefore provides a more direct test of whether gender preference amplifies the unequal physical health consequences of climate shock.

The results for male individuals using Eq. (11) are presented in Table 4. For males, CPR leads to a significant decrease in ADL scores across most cohorts (Panel A), particularly for boys in households with females (Panel A, column 4), suggesting a substantial decline in their physical health. However, their health expenditure tends to decrease in households where no females are present (Panel B), potentially reflecting a lack of caregiving support or budget constraints in single-male households. This suggests that when male members’ physical health is compromised by climate shocks, households, driven by male-centered preferences, allocate significant financial resources to mitigate these impacts. In contrast, Table 5 reveals a protection gap for females. While girls and women also suffer physical health declines under climate shock, their health expenditure does not show a commensurate increase in households where males are present. Specifically, for females in households with no males, health expenditure increases significantly to buffer the shock (Table 5, Column 2). However, this protective investment diminishes or even turns negative when they reside with males. This confirms that in male-preferred households, resources are prioritized for male physical health recovery, leaving females to bear a disproportionate physical health burden without adequate financial support for medical care.

Table 4.

Marginal effects of climate shock on ADL and health expenditure heterogeneity by male with or without female

	Without female			With female
	Boys	Adult men	Elderly men	Boys	Adult men	Elderly men
	(1)	(2)	(3)	(4)	(5)	(6)
Panel A Dependent variable: ADL
CPR	0.038	−0.004	−0.003	−0.010***	−0.002	−0.004**
	(0.053)	(0.004)	(0.004)	(0.002)	(0.002)	(0.002)
Observations	20054
R2	0.581
Panel B Dependent variable: health expenditure
CPR	−25.519*	−23.444**	−20.142*	37.063***	11.461*	10.276
	(13.935)	(10.829)	(10.561)	(7.728)	(6.732)	(6.819)
Observations	20043
R2	0.376

Notes: Individual fixed effect, survey fixed effect, city fixed effect, and household and individual control variables are controlled. Robust standard errors clustered at individual level are reported in parentheses. * p < 0.1, ** p < 0.05, *** p < 0.01

Table 5.

Marginal effects of climate shock on ADL and health expenditure heterogeneity by female with or without male

	Without male			With male
	Girls	Adult women	Elderly women	Girls	Adult women	Elderly women
	(1)	(2)	(3)	(4)	(5)	(6)
Panel A Dependent variable: ADL
CPR		−0.003**	−0.003*	−0.010**	−0.005	−0.006
		(0.001)	(0.002)	(0.004)	(0.005)	(0.005)
Observations	18299
R2	0.591
Panel B Dependent variable: health expenditure
CPR		14.270**	9.409	9.030	0.835	−6.778
		(6.421)	(6.828)	(8.302)	(6.556)	(5.928)
Observations	18155
R2	0.475

Notes: Individual fixed effect, survey fixed effect, city fixed effect, and household and individual control variables are controlled. Robust standard errors clustered at individual level are reported in parentheses. Blank spaces indicate missing study samples. * p < 0.1, ** p < 0.05, *** p < 0.01

The contrast between Table 4 and Table 5 reveals that under climate shock, male-preferred households prioritize the physical health needs or functional recovery of male members. In such households, females experience significant health declines but don’t receive a corresponding increase in health expenditure. This confirms that females in male-oriented domestic structures bear a disproportionate physical health burden, as household resources are skewed toward mitigating the impacts on male members, leaving females with inadequate financial protection for their physical health recovery. This suggests that gender differences in intra-household allocation of health resources under climate shock shocks are influenced by inequitable gender preferences, thereby providing empirical support for PROPOSITION 3, specifically that when the physical health outcomes of the two individuals are substitutes (0 < σ < 1) and holds, the negative spillover effect on females increases as household preference for females decreases. These spillover patterns should be understood as reduced-form outcomes reflecting the interaction of gender, life-cycle stage, caregiving responsibilities, and household risk-sharing arrangements, rather than as evidence of a single dominant mechanism.

Robustness checks

Changing clustering levels. To account for potential intra-group correlation and unobserved spatial dependencies, we re-estimate the baseline model by shifting the clustering of standard errors to the household and community strata. Given that individuals residing within the same domestic unit or geographic proximity are often exposed to identical environmental stressors and socio-economic shocks, ignoring such correlations could lead to over-precisely estimated t-statistics. As reported in Table 6 and Appendix Tables C3–C4, the standard errors remain stable and all key coefficients maintain their statistical significance at the conventional levels. This persistence confirms that our primary findings are not biased by localized error structures.

Table 6.

Changing the standard error clustering level

	ADL			Health expenditure
	(1)	(2)	(3)		(4)
CPR	−0.003**	−0.003**	7.941*		7.858*
	(0.001)	(0.001)	(4.043)		(4.545)
Survey FE	Yes	Yes	Yes		Yes
Individual FE	Yes	Yes	Yes		Yes
City FE	Yes	Yes	Yes		Yes
Individual controls	Yes	Yes	Yes		Yes
Household controls	Yes	Yes	Yes		Yes
Cluster level	Household	Community	Household		Community
Observations	38333	38333	38180		38180
R2	0.587	0.587	0.437		0.437

Notes: Robust standard errors are reported in parentheses. * p < 0.1, ** p < 0.05, *** p < 0.01

Alternative measures. We examine the sensitivity of our results to alternative variable definitions to ensure the findings are not artifactual to specific measurement choices. While ADL scores serve as our primary physical health indicator, we substitute them with hospitalization frequency, which is a measure of acute healthcare utilization (Table 7, Column 1 and Appendix Table C5, Panel A). Similarly, for health-related financial burdens, we employ expenditure on health insurance to capture precautionary investments and risk-mitigation strategies (Table 7, Column 2 and Appendix Table C5, Panel B). The high degree of consistency between these results and the baseline estimations reinforces the robustness of the identified CPR-health nexus.

Table 7.

Alternative measures and Additional control variables

	Alternative measures		Additional control variables
	Hospital visit	Medical insurance expenditure	ADL	Health expenditure
	(1)	(2)	(3)	(4)
CPR	0.002	2.708***	−0.003***	7.941*
	(0.001)	(0.979)	(0.001)	(4.050)
Survey FE	Yes	Yes	Yes	Yes
Individual FE	Yes	Yes	Yes	Yes
City FE	Yes	Yes	Yes	Yes
Individual controls	Yes	Yes	Yes	Yes
Household controls	Yes	Yes	Yes	Yes
City controls			Yes	Yes
Observations	35392	43534	38333	38180
R2	0.539	0.333	0.587	0.437

Notes: Robust standard errors clustered at individual level are reported in parentheses. * p < 0.1, ** p < 0.05, *** p < 0.01

Additional control variables. To strengthen causal inference by addressing potential confounders, we control for three city-level time-varying variables. Gross Domestic Product per capita accounts for the overall economic capacity and development level, which is correlated with both climate exposure and health infrastructure. The number of hospital beds measures the availability of healthcare resources, which could directly affect health expenditures and physical health outcomes and may be correlated with regional development priorities. City-level carbon emissions serve as a proxy for industrial activity, energy use intensity, and associated environmental pressures, which are often linked to both the magnitude of climate risks and the baseline health environment. The robustness of our core findings after including these controls suggests that the estimated impact of CPR on ADL and health expenditure is not driven by these concurrent socioeconomic and environmental factors (Table 7, Column 3–4).

Alternative model specifications. Considering the non-negative and discrete nature of the ADL index, the linear fixed-effects model might only offer a first-order approximation of the relationship. To address potential non-linearities and the count data characteristics of functional health scores, we re-run the analysis using a panel Poisson fixed-effects specification. The results, detailed in Table 8 (Column 1) and Appendix Table C6, show that CPR continues to exert a significant positive pressure on ADL scores (indicating functional decline). This confirms that our baseline conclusions are invariant to the choice of functional form.

Table 8.

Alternative model specifications and drop outliers

	Panel poisson model	Drop outliers
	ADL	ADL	Health expenditure
	(1)	(2)	(3)
CPR	−0.007***	−0.003***	5.187**
	(0.002)	(0.001)	(2.592)
Survey FE	Yes	Yes	Yes
Individual FE	Yes	Yes	Yes
City FE	—	Yes	Yes
Individual controls	Yes	Yes	Yes
Household controls	Yes	Yes	Yes
Observations	14511	38333	38180
R2	—	0.587	0.425

Notes: Robust standard errors clustered at individual level are reported in parentheses. * p < 0.1, ** p < 0.05, *** p < 0.01

Outliers. To safeguard our estimates against the undue influence of extreme observations or data anomalies, we perform a sensitivity analysis by trimming the upper and lower 5% tails of the ADL and health expenditure distributions. The regression results for this restricted sample, presented in Table 8 (Columns 2–3) and Appendix Table C7, remain quantitatively and qualitatively consistent with the full-sample baseline. This suggests that the observed physical health-deteriorating effects of climate shocks are a general trend rather than being driven by a small fraction of outlier cases.

Exposure window. Recognizing that the physical health of climate risk may exhibit heterogeneous temporal footprints, we explore the intertemporal dynamics of CPR by introducing lagged exposure windows. Specifically, we replace contemporaneous risk with one-year and two-year lagged CPR metrics to capture delayed manifestations of physical health impairment and financial strain. Table 9 reveals a significant one-year lag in the response of healthcare expenditure, whereas the impact on ADL functional capacity appears more immediate. This temporal asymmetry suggests that while functional decline occurs shortly after climate shocks, the mobilization of financial resources for medical remediation involves a certain gestation period.

Table 9.

Exposure window

	ADL		Health expenditure
	(1)	(2)	(3)	(4)
CPR (Lag one year)	−0.001		8.465***
	(0.001)		(2.992)
CPR (Lag two year)		−0.000		5.474
		(0.001)		(4.154)
Survey FE	Yes	Yes	Yes	Yes
Individual FE	Yes	Yes	Yes	Yes
City FE	Yes	Yes	Yes	Yes
Individual controls	Yes	Yes	Yes	Yes
Household controls	Yes	Yes	Yes	Yes
Observations	38333	38333	38180	38180
R2	0.587	0.587	0.437	0.437

Notes: Robust standard errors clustered at the individual level are reported in parentheses. * p < 0.1, ** p < 0.05, *** p < 0.01

Counterfactual analysis. Finally, we execute a placebo test by constructing a lead-exposure window to rule out the possibility of spurious correlations or pre-existing trends. Theoretically, future climate physical risks should have no causal influence on contemporary functional health or medical spending. By regressing current outcomes on CPR values from the subsequent year, we find no statistically significant coefficients (Table 10). This falsification test confirms that the baseline results are indeed driven by actual climate exposure rather than by random chance or systemic endogeneity, thereby bolstering the causal interpretation of our study.

Table 10.

Counterfactual analysis

	ADL	Health expenditure
	(1)	(3)
CPR (One year ahead)	−0.001	2.078
	(0.001)	(5.483)
Survey FE	Yes	Yes
Individual FE	Yes	Yes
City FE	Yes	Yes
Individual controls	Yes	Yes
Household controls	Yes	Yes
Observations	38333	38180
R2	0.650	0.301

Notes: Robust standard errors clustered at individual level are reported in parentheses. * p < 0.1, ** p < 0.05, *** p < 0.01

Endogeneity. Endogeneity issues, stemming principally from omitted variable bias and simultaneous causality, pose a potential threat to the identification of the causal link between climate shock, physical health, and financial burdens. First, unobserved individual heterogeneity, such as genetic resilience, long-standing lifestyle habits, or inherent environmental adaptability, might correlate with both an individual’s functional health status (ADL) and their exposure to climatic risks. Failure to account for these latent factors could result in biased estimates. Second, a reverse causality channel may exist. Individuals with deteriorating ADL or higher health spending may alter their behavioral patterns (e.g., reducing outdoor mobility or labor supply) in ways that modify their actual exposure to extreme climate events.

To address these identification challenges and isolate the exogenous variation in climate shock, we implement a Two-Stage Least Squares (2SLS) strategy, utilizing city-level altitude as an instrumental variable (IV) for the CPR. Altitude is a standard geographic instrument in environmental economics literature [24, 25]. It satisfies the relevance condition because altitude is a fundamental determinant of local meteorological patterns, strongly predicting temperature variability and precipitation levels. Furthermore, as a time-invariant geological feature determined by natural terrain formation, altitude satisfies the exclusion restriction. It plausibly influences individual physical health outcomes and health expenditures solely through its impact on long-term climatic conditions rather than through direct biological pathways.

Table 11 presents the 2SLS estimation results. Column (1) reports the first-stage regression, revealing that altitude is a robust positive predictor of CPR, statistically significant at the 1% level. The validity of the instrument is supported by rigorous diagnostic tests. The Kleibergen-Paap rk LM statistic (1615.132) strongly rejects the null hypothesis of under-identification (p < 0.01), and the Cragg-Donald Wald F statistic (1930.797) far exceeds the Stock-Yogo critical value of 16.38 (at the 10% level), alleviating concerns regarding weak instrumentation. To verify the exogeneity of the instrumental variable, the Hansen J test for over-identification is conducted. The p-value of the Hansen J statistic is 0.000, indicating that the instrumental variables satisfy the exogeneity assumption.

Table 11.

The treatment of endogeneity

	Treatment of endogeneity
	The first stage	The second stage
	CPR	ADL	Health expenditure
	(1)	(2)	(3)
CPR		−0.011***	33.677**
		(0.003)	(17.051)
Altitude	0.862***
	(0.019)
Survey FE	Yes	Yes	Yes
Individual FE	Yes	——	——
City FE	Yes	——	——
Individual controls	Yes	Yes	Yes
Household controls	Yes	Yes	Yes
Kleibergen-Paap rk LM statistic		1615.132 [0.000]	1610.547 [0.000]
Cragg-Donald Wald F statistic		1930.797 < 16.38>	1909.207 < 16.38>
p value of the Hansen J statistic		0.000	0.000
Observations	46319	45871	45546
R2	0.116	0.092	0.010

Notes: Robust standard errors clustered at individual level are reported in parentheses. * p < 0.1, ** p < 0.05, *** p < 0.01. Values in [] are the p values of the corresponding test statistics. Values in<>are the critical values at the 10% level of the Stock-Yogo weak ID test. Since elevation is a cross-sectional data, the models in treatment of endogeneity only controlled for the time-fixed effects

Columns (2) and (3) display the second-stage results. Even after correcting for potential endogeneity, the impact of climate shock remains robust. The coefficient for ADL is significantly negative, indicating that increased climate shock erodes physical health. Conversely, the coefficient for health expenditure is significantly positive, confirming that climate shocks escalate medical financial burdens. These findings align with the baseline results in both direction and statistical significance, reinforcing the causal interpretation that climate shock deteriorates physical health and drives up healthcare expenditure.

Channels and unequal selection effects

Channels

Climate shock may influence physical health and health expenditure through following channels. First, climate shock may discourage outdoor activities, reduce exercise time and thereby deteriorate cardiovascular physical health and immunity. This, in turn, may heighten physical health risks, decrease expenditures on preventive healthcare, and raise costs associated with treatment [26]. Second, climate shock may undermine social integration by restricting outdoor mobility, community engagement, and interpersonal interactions, thereby intensifying social isolation and functional health deterioration. This decline in social participation may further exacerbate physical functional limitations, as the lack of social stimulation and peer support is known to accelerate functional decay and aging [27]. Against this backdrop, this paper examines the mechanisms through which climate shock affects ADL and health expenditure along three key channels—exercise and social participation—and empirically tests these pathways.

The analysis in this section is designed to explore plausible mechanisms linking climate risks to physical health outcomes rather than to estimate formal causal mediation effects. Specifically, the proposed channels are treated as intermediate outcomes that are influenced by climate risks and are closely related to physical health, allowing us to assess whether climate exposure operates through these dimensions. While formal mediation approaches, such as structural equation modeling with Sobel or Monte Carlo tests, are well suited for identifying causal transmission channels, their implementation requires strong assumptions regarding functional form, error independence, and the absence of unobserved confounding between mediators and physical health outcomes. Given data limitations and identification concerns, we adopt a reduced-form framework and interpret the results as suggestive evidence on potential pathways rather than definitive mediation effects.

Exercise

Maintaining consistent physical activity is a fundamental prerequisite for sustaining functional independence and preventing physiological decay. However, the escalating volatility of climate physical risks tends to obstruct established fitness routines by degrading the quality of the outdoor environment, thereby forcing individuals to recalibrate their health-seeking behaviors [28]. To examine this behavioral channel, we evaluate how climate shocks reshape the intensity and duration of exercise. The empirical evidence presented in Table 122 (Columns 1–3) reveals a significant structural shift: while climate physical risk (CPR) correlates with a marginal increase in low-intensity exercise hours, it exerts a substantial negative pressure on moderate-to-vigorous physical activity (MVPA) and aggregate exercise time. The coefficients for these core variables are both economically and statistically significant. This phenomenon suggests a substitution effect driven by environmental constraints; individuals may pivot toward milder, indoor-compatible movements to circumvent the physiological hazards of extreme outdoor temperatures [29, 30]. Nevertheless, the observed decline in strenuous physical engagement is critical, as it directly undermines musculoskeletal strength and aerobic capacity, providing a plausible pathway for the observed increase in ADL scores and the subsequent deterioration of functional health. Critically, this reduction in moderate-to-high intensity exercise may disproportionately affect male, who traditionally rely more on vigorous physical upkeep to maintain labor productivity. This specific form of physiological capital depreciation likely triggers the intra-household bargaining dynamics observed earlier, compelling households to allocate preferential medical expenditure to males to restore their physical capacity.

Table 12.

Estimated effects of climate shock on exercise and social participation

	Low-intensity exercise	Moderate-intensity exercise	High-intensity exercise	Total exercise	Social participation
	(1)	(2)	(3)	(4)	(5)
CPR	0.002***	−0.001**	−0.002***	−0.002**	−0.001**
	(0.000)	(0.000)	(0.000)	(0.001)	(0.000)
Survey FE	Yes	Yes	Yes	Yes	Yes
Individual FE	Yes	Yes	Yes	Yes	Yes
City FE	Yes	Yes	Yes	Yes	Yes
Individual controls	Yes	Yes	Yes	Yes	Yes
Household controls	Yes	Yes	Yes	Yes	Yes
Observations	43534	43534	43534	43534	42910
R2	0.4394	0.373	0.462	0.468	0.461

Notes: Robust standard errors clustered at individual level are reported in parentheses. * p < 0.1, ** p < 0.05, *** p < 0.01

Social participation

Social participation is critical to maintaining physical functional capacity and informed physical health-related decisions. Climate shock can have far-reaching effects on individuals’ social networks through extreme weather events that act as physical barriers to communal gatherings and informal support systems. We use whether visiting neighbors or interacting with friends as proxy variables for an individual’s level of social participation. At the micro level, climate shock is likely to reduce the frequency of social participation, leading to social withdrawal and diminished functional resilience [31]. The estimation results of Columns (5) in Table 12 show that climate shock significantly reduces social participation. This implies that climate shock weakens the social participation of individuals, further influence individual physical health. This finding offers a vital explanation for the gendered physical health penalty. Given that female’s functional health maintenance often relies more critically on informal social networks and communal support, the climate-induced contraction of social participation disproportionately undermines their ADL status. Unlike the financial compensation observed for males, this loss of social capital is less visible and harder to substitute with monetary expenditure, thereby exacerbating the physical health vulnerability of females.

Unequal selection effects

In this section, we scrutinize the heterogeneous sensitivities of functional health to climate stressors, focusing on how socio-institutional barriers and domestic configurations modulate the baseline impacts. We hypothesize that structural disparities, specifically those related to rural residency and limited educational attainment, exacerbate the functional health penalty for females by restricting their access to adaptive resources. Furthermore, we explore how intra-household dynamics, such as marital status and life-cycle transitions (retirement), influence bargaining power and social support networks, potentially intensifying the physical health-deteriorating effects of CPR. For the purpose of this analysis, ‘low-educated’ is defined as individuals who have not completed middle school. To empirically identify these differential vulnerabilities, we integrate interaction terms between CPR and indicators of structural and domestic characteristics into our primary estimation framework. These associations should be interpreted as reduced-form correlations that are consistent with behavioral adjustment under climate stress, rather than as evidence of a causal or mediating mechanism.

Urban-rural divides and educational attainment

The empirical results regarding structural gradients are presented in Table 13. Individuals residing in rural areas or possessing lower educational levels exhibit a more pronounced surge in ADL, indicating greater functional fragility, and a relative stagnation in health expenditure following climate shocks. These vulnerabilities are consistent across various age-gender cohorts (Appendix Tables C8–C9). This intergroup divergence can be attributed to the protection gap caused by limited physical health literacy and the scarcity of high-quality medical infrastructure in marginalized regions. Our findings align with the theory of asymmetric environmental risk absorption, where lower socioeconomic status diminishes an individual’s capacity to buffer the physiological and behavioral disruptions caused by escalating climate risks [32].

Table 13.

Climate shock affects the structural disparities in ADL and health expenditure

	ADL		Health expenditure
	(1)	(2)	(3)	(4)
CPR × Urban	−0.002		17.283**
	(0.001)		(8.452)
CPR × Rural	−0.003***		3.360
	(0.001)		(3.971)
CPR × Low education		−0.003***		5.836
		(0.001)		(4.184)
CPR × High education		−0.002*		11.004**
		(0.002)		(5.406)
Survey FE	Yes	Yes	Yes	Yes
Individual FE	Yes	Yes	Yes	Yes
City FE	Yes	Yes	Yes	Yes
Individual controls	Yes	Yes	Yes	Yes
Household controls	Yes	Yes	Yes	Yes
Observations	38333	35459	38180	38180
R2	0.587	0.650	0.437	0.437

Notes: Robust standard errors clustered at individual level are reported in parentheses. * p < 0.1, ** p < 0.05, *** p < 0.01

Marriage and lifecycle transitions

Table 14 delineates how domestic configurations and life-cycle stages shape the physical health impacts of climate shock. The results indicate that unmarried and retired individuals are particularly susceptible to climate-induced functional decline, often accompanied by inadequate increases in health spending. This vulnerability is further detailed across specific demographic groups in Appendix Tables C10–C11. For those undergoing lifecycle transitions, such as retirement, the erosion of functional health may be compounded by the loss of work-related social capital and the subsequent contraction of social networks, which limits their behavioral adaptation to climate stressors [33]. Collectively, these findings underscore that climate shock does not merely act on individuals in isolation but interacts with domestic power structures and life stages to reinforce existing physical health inequities.

Table 14.

Climate shock affects the intra-household dynamics of ADL and health expenditure

	ADL		Health expenditure
	(1)	(3)	(4)	(6)
CPR × Married	−0.003***		11.283**
	(0.003)		(4.420)
CPR × Other	−0.008**		−16.762**
	(0.003)		(8.107)
CPR × Retired		−0.004		13.320**
		(0.001)		(5.250)
CPR × Unretired		−0.004***		6.432
		(0.001)		(4.057)
Survey FE	Yes	Yes	Yes	Yes
Individual FE	Yes	Yes	Yes	Yes
City FE	Yes	Yes	Yes	Yes
Individual controls	Yes	Yes	Yes	Yes
Household controls	Yes	Yes	Yes	Yes
Observations	43023	38333	38180	38180
R2	0.610	0.587	0.437	0.437

Notes: Robust standard errors clustered at individual level are reported in parentheses. * p < 0.1, ** p < 0.05, *** p < 0.01

Physical health adaptation to climate shock

Institutional support, particularly through the provision of localized medical infrastructure and stable labor markets, serves as a critical buffer against the functional erosion triggered by climate stressors. While enhanced healthcare accessibility provides a direct remedial pathway for mitigating physical impairment, employment stability bolsters functional resilience by securing economic resources and maintaining protective social networks [34, 35]. To quantify the efficacy of these mitigation strategies, we estimate the marginal effects of healthcare and employment availability on the ADL trajectories across different demographic cohorts. Throughout this section, “access to healthcare resources” is interpreted as effective access, the realized ability to obtain timely medical care, rather than as a purely supply-side measure of healthcare availability. As depicted in Fig. 4, both robust healthcare resources and employment stability significantly attenuate the climate-induced increase in ADL scores (i.e., reduce functional health deterioration), indicating a physical health-protective effect, for adult men and women.

Fig. 4.

Marginal effects of climate adaptation for physical health. Employment opportunity represents that the respondent has access to employment opportunities in local city, which is measured by whether the respondent works in local city. Healthcare resource indicates that the respondent has access to healthcare resources in local city, which is measured by whether the respondent cures diseases in local city. Other city represents that the respondent doesn’t have access to employment opportunities and healthcare resources in local city

However, a distinct divergence emerges for elderly women. Their functional resilience appears almost exclusively contingent upon the availability of localized healthcare services, whereas employment opportunities yield negligible mitigating effects on ADL. These findings underscore that in the face of escalating climate volatility, physical health-centered interventions are more potent than purely economic ones in shielding the most vulnerable members from functional decay, especially within gendered household structures that prioritize resource allocation. Consequently, prioritizing the spatial density of primary care and medical facilities is a prerequisite for fostering equitable climate adaptation and narrowing the intra-household physical health gap.

Conclusion

This paper estimates the impact of climate shock on physical health. We construct a theoretical model of intra-household resource allocation and derive gender differences in climate shock impacts on physical health and health expenditure, and intra-household spillovers across gender preference contexts. Based on CHARLS dataset, this paper provides novel empirical evidence on the direct impact of climate shock on physical health and health expenditure, gender inequality and its intra-household spillovers, analyzed using reduced-form model. We also evaluate the effectiveness of access to local employment opportunities and healthcare resources in supporting residents’ physical health adaptation to climate shock.

The estimation results suggest that climate shock exacerbates ADL and increase the burden of health expenditure. In particular, males are in a better position than females to pay for health expenditure, and shocks to males’ physical health caused by climate shock have negative spillover effects on females’ physical health and on health expenditure in male-headed households. The channels of impact are reduced exercise and social participation. Moreover, when structural disparities and intra-household dynamics are taken into account, such impacts show unequal selection effects, which are mainly concentrated in rural, low-educated, unmarried and unretired groups. Access to local employment opportunities and healthcare resources plays a key role in mitigating the adverse effects of climate shock on physical health.

This paper highlights the importance of considering intra-household spillovers for a comprehensive understanding of climate shock impacts. In particular, exposure to the same climate physical risks may affect male and female asymmetrically due to household budget constraints and patriarchy. Such unintended negative spillovers exacerbate gender inequalities in physical health and health expenditure.

Over the past decades, China has implemented a series of physical health adaptation policies aimed at mitigating the adverse effects of climate shock on public health, particularly among vulnerable populations. Notably, the healthy cities pilot policy has facilitated the development of a comprehensive health protection system, and has expanded access to affordable primary healthcare through community health centers and primary care networks, ensuring timely medical support for the dispossessed. While these measures vary in coverage and implementation efficacy, empirical evidence underscores their significant role in improving population health and reducing physical health disparities. As climate shock intensifies physical health risks, policymakers must adopt targeted adaptation strategies that effectively address the physical health needs of diverse demographic groups. Empirical research in this paper, particularly on the negative intra-household spillover effects of climate shock on females’ physical health and health expenditure, offers valuable insights for policy formulation in other developing regions.

Our findings offer several implications for improving health adaptation policies in China. First, the pronounced gender- and age-specific health responses to climate shock suggest that uniform adaptation strategies may be insufficient. Health adaptation policies should incorporate targeted interventions that prioritize vulnerable groups, particularly females and older individuals, whose health appears more adversely affected by climate exposure. Second, the results indicate that increased access to local employment opportunities and healthcare resources is beneficial to physical health and health equity, highlighting the need for adaptation policies to integrate employment services and preventive health care into climate adaptation frameworks. Specifically, this underscores the need for adaptation policies that not only build hospitals but also ensure financial access to primary care for women in rural areas, potentially through targeted subsidies integrated with climate resilience programs. Third, the differential effectiveness of health investments across demographic groups underscores the importance of aligning adaptation policies with intra-household resource allocation dynamics, for example by strengthening public provision of child and adolescent health services to mitigate potential gender biases in household decision-making. Climate adaptation policies in China should move beyond expanding healthcare access and instead emphasize targeted, health-oriented, and equity-sensitive interventions.

Electronic supplementary material

Below is the link to the electronic supplementary material.

Abbreviations

ADL

Activities of daily living

CHARLS

China Health and Retirement Longitudinal Study

CPR

Climate physical risk

LTD

Extreme low temperature days

HTD

Extreme high temperature days

ERD

Extreme rainfall days

EDD

Extreme drought day

MVPA

Moderate-to-vigorous physical activity

Author contributions

Hongkun Zhao: Methodology, Writing – original draft, Writing – review & editing. Zhuo Chen: Formal analysis, Writing – review & editing.

Funding

This research did not receive any specific grant from funding agencies in the public, commercial, or not for profit sectors.

Data availability

Data linking: https://charls.charlsdata.com/pages/data/111/zh-cn.html.

Declarations

Ethics approval and consent to participate

This article was supported by the Biomedical Ethics Committee of Peking University (ethical code: IRB00001052-11015) and written informed consent was obtained from the participants. All procedures involving human participants adhered to ethical standards, including the Declaration of Helsinki.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.