The influence and mechanism of female-headed households on household debt risk: empirical evidence from China

As society progresses, improving women’s social status has become a global trend, and gender gaps have narrowed in many dimensions. Within households, women’s decision-making power has risen, leading to a growing prevalence of female-headed households (defined here as households in which the woman has the dominant right to make decisions on family affairs). In China, survey evidence shows substantial increases in wives’ participation in major family decisions between 2010 and 2020. Household debt is a key macro-financial concern because excessive household leverage can threaten financial stability. At the same time, women are typically more risk-averse and more cautious in financial decision-making than men. This study asks whether female-headed households reduce household debt risk and through what mechanisms (e.g., risk attitudes and asset allocation). Understanding these relationships can inform household financial management, risk prevention, and policies to promote gender equality. Using CHFS2019 microdata, the paper measures household debt risk by the debt-to-income ratio and, given its zero truncation, employs a Tobit model. It addresses endogeneity using instrumental variables and self-selection using propensity score matching and a treatment effects model. It also explores heterogeneity by city development level and family demographic structure and performs robustness checks (alternative risk indicator and models). The study contributes by (1) aligning the definition of “household head” with the family’s de facto decision-maker, (2) documenting that female-headed households significantly reduce household debt risk, and (3) clarifying mechanisms via risk aversion (partial mediation) and housing property holding (masking effect).

The literature comprises two strands. 1) Gender characteristics and debt: Corporate finance studies find mixed effects of female leadership on leverage—some show female executives reduce debt due to more cautious, less overconfident financing preferences and tighter credit constraints; others find higher short-term debt shares or more debt in certain ownership contexts. Household-level studies show gender influences indebtedness likelihood, debt levels, and channels: males often increase the probability of indebtedness; females are more hesitant to take unnecessary debt, exhibit higher risk aversion and prudence, and may face discrimination or barriers in formal credit markets, sometimes relying more on informal finance. Risk attitudes, subjective debt burden, and financial self-efficacy mediate gendered debt behaviors. 2) Determinants of household debt: Micro factors include income (with mixed effects on leverage and debt distress), demographic structure (dependency ratios), financial literacy (lower literacy linked to excessive/high-cost borrowing), expectations (optimism raising leverage and distress), and decision-maker characteristics (age, health, education, personality). Macro factors include housing markets (house price rises), inflation, credit market development, inequality, and digital finance/payment technologies, which can increase household debt. Gaps: Limited research examines how female-led household financial decision-making directly affects household debt risk and the mechanisms (e.g., risk attitudes and asset allocation) through which female-headed households operate.

Data: CHFS2019, covering 29 provinces, 343 districts/counties, 1360 village/neighborhood committees; 34,643 households and 107,008 individuals. Sample processing: exclude missing key variables; exclude negative income or zero consumption expenditure; winsorize income and liabilities; exclude households with total assets >100 million RMB. Final N=20,919 households. Variables: - Core explanatory variable: Female-headed household (1 if household head is female; CHFS household head is the de facto decision-maker). - Explained variable: Household debt risk measured by debt-to-income ratio (DIR), with total debt including 11 liability categories and total income including 5 income types. - Mediators: (i) Risk aversion (binary=1 if respondent prefers lower risk/return or no risk in investment choice question), (ii) Housing property holding (ratio of housing assets to total assets). - Controls: Individual (age, education years, self-reported health of head); Family (family size, number participating in social security, number in medical insurance, owner-occupied housing indicator, log household consumption expenditure, log household deposit assets, internet use), Regional (rural vs. urban indicator; region fixed effects for East/Central/West/Northeast). Modeling strategy: - Baseline: Left-censored Tobit for DIR: DIR_i = β0 + β1 Fhead_i + β2 Controls_i + Region FE + ε_i. - Mediation: M_i = c0 + c1 Fhead_i + c2 Controls_i + Region FE + ε_i; Tobit(DIR_i) = γ0 + γ1 Fhead_i + γ2 M_i + γ3 Controls_i + Region FE + ε_i. Mediation assessed using Wen and Ye (2014) and distinguishing mediating vs. masking effects per MacKinnon et al. (2000). - Endogeneity checks: (i) IV Tobit using female household head rate in the same community as instrument (relevance strong; exclusion argued by local peer effects). (ii) Propensity score matching (one-to-four, radius, kernel), with balance tests and Rosenbaum bounds (Gamma). (iii) Treatment effects model (two-step): first-stage Probit for treatment; second-stage outcome with LR endogeneity test. - Heterogeneity: Split by city development level (first/new first-tier; second-tier; third-tier and below) and by family composition (0,1,2+ children aged ≤14; 0,1,2+ elderly aged ≥65). - Robustness: Alternative risk indicator (asset-liability ratio via Tobit); alternative estimator (OLS); provincial fixed effects (province dummies). Descriptive statistics: Mean DIR=0.853; asset-liability ratio=0.115; female-headed=14%; risk aversion=0.644; mean housing property holding=0.649; owner-occupied housing≈99.6%; average head: age 56.7, health 3.25/5, education 8.79 years; internet use≈69.7%. Univariate comparisons show female-headed households have lower DIR and asset-liability ratios; higher risk aversion; higher housing property shares.

- Baseline Tobit: Female-headed households significantly reduce household debt risk. Marginal effect of female-headed household on DIR = -0.143 (t/z≈-3.20), i.e., a 14.29 percentage-point reduction, confirming H1. - Key controls: Older and healthier heads, more members in social security, and higher deposit assets reduce DIR; larger family size, internet use, higher consumption expenditure, owner-occupied housing, and rural households associate with higher DIR; education becomes insignificant after fuller controls. - IV Tobit: Instrument (community female head rate) is strong (first-stage F=143.02; t=41.15). Second-stage coefficient of female-headed household on DIR = -1.766 (t≈-3.20), Wald test p=0.022, reinforcing a causal negative effect. - PSM (ATT on DIR): One-to-four matching: -0.245 (SE=0.071, t=-3.46, p=0.003); Radius: -0.157 (SE=0.057, t=-2.73, p=0.005); Kernel: -0.155 (SE=0.057, t=-2.70, p=0.005). Rosenbaum bounds show no sensitivity for Gamma∈[1,2], indicating robustness to hidden bias. - Treatment effects model: Second-stage effect = -0.746 (z=-3.48), LR endogeneity test χ²=4.59 (p=0.032), consistent with a significant inhibitory effect. - Mediation: • Risk aversion channel: Female-headed → higher risk aversion (c1=0.023, p<0.05); risk aversion → lower DIR (γ2=-0.180, p<0.001); direct effect of female-headed remains negative (γ1=-0.138, p<0.001). Since indirect and direct effects have same sign and |β_total|=0.143 > |γ1|=0.138, risk aversion partially mediates (H2 supported). • Housing property holding channel: Female-headed → higher housing asset share (c1=0.049, p<0.001); housing property holding → higher DIR (γ2=0.554, p<0.001); direct effect of female-headed remains negative (γ1=-0.169, p<0.001). Indirect and direct effects have opposite signs and |β_total|=0.143 < |γ1|=0.169, indicating a masking effect (H3 supported). - Heterogeneity: • By city level: Effect significant only in third-tier and below (coef=-0.645, t=-3.23); insignificant in first/new first-tier and second-tier cities. • By family structure: Significant reductions for households with 0 children (coef=-0.654, t=-3.43), 0 elderly (coef=-0.482, t=-2.58), and 2+ elderly (coef=-0.596, t=-1.90). Not significant for 1 child, 2+ children, or 1 elderly. - Robustness: • Alternative risk metric (asset-liability ratio): female-headed marginal effect = -0.025 (p<0.001). • OLS on DIR: coefficient = -0.1656 (p<0.01). • Provincial fixed effects: coefficient = -0.373 (p<0.05).

The findings confirm that when women are the de facto financial decision-makers, household debt risk is lower. This aligns with evidence that females are more risk-averse and cautious in borrowing and debt management, thereby mitigating leverage-related vulnerabilities. The control variable patterns are economically consistent: greater age and health reduce debt reliance; larger family size and higher consumption needs raise borrowing; financial buffers (deposits) and social security participation lower risk; internet use is associated with greater access to borrowing and consumption, increasing leverage; owner-occupied housing raises risk due to mortgage debt; rural status correlates with higher leverage due to lower income and limited formal credit access. Mechanisms: Risk aversion partially mediates the negative effect—female heads more often prefer safer choices, restraining borrowing and reducing debt burdens. Conversely, female heads also tilt portfolios toward housing assets, and because housing acquisition and equity-based borrowing increase liabilities, this channel raises debt risk and masks part of the beneficial direct effect. Heterogeneity: The stronger effect in less-developed cities likely reflects thinner financial infrastructure and higher reliance on household-level prudence where female-headed caution has more impact. Family composition shapes needs and borrowing purposes: without children or elderly, households face fewer expenditure pressures, enabling savings and lower leverage; with multiple elderly members, improved elderly-care systems and heightened caution against medical shocks may drive more prudent borrowing. Where dependents increase expenditure needs, prudence of female heads is counterbalanced by debt demand, making effects insignificant. Overall, empowering women in household decision-making supports prudent debt management and financial stability, but overconcentration in low-risk housing assets can inadvertently elevate leverage; balanced portfolio diversification within acceptable risk is advisable.

Using nationally representative CHFS2019 data, the study shows that female-headed households significantly reduce household debt risk. Causal and robustness checks (IV Tobit, PSM with Rosenbaum bounds, treatment effects, alternate metrics/models, and provincial FE) corroborate the result. Mechanistically, risk aversion partially mediates the reduction in risk, while greater housing property holding by female-headed households introduces a masking effect by increasing debt risk. Effects are strongest in third-tier cities and below, and among households without children, without elderly, and with two or more elderly members. Policy implications: - Encourage women’s participation in household economic decision-making to enhance financial prudence and stability. - Strengthen financial education for women to calibrate risk tolerance and avoid excessive concentration in low-risk, high-leverage assets (e.g., housing). - Improve financial infrastructure and credit access in less-developed regions to reduce reliance on costly/informal credit. - Continue enhancing pension and caregiving policy frameworks to alleviate elderly-related financial pressures. Future research should refine household debt risk metrics by incorporating liability structures, interest rates, income stability, and credit histories, and use panel data to study dynamics over time.

- Measurement: Debt-to-income and asset-liability ratios capture leverage and repayment capacity but not the composition and riskiness of assets/liabilities, interest rates, income stability, employment type, or credit histories. Future work should construct more comprehensive risk indicators incorporating these dimensions. - Data design: Cross-sectional CHFS2019 limits dynamic inference. Panel data would allow analysis of temporal changes, causal dynamics, and persistence in household debt risk.