Impact of energy poverty on cognitive and mental health among middle-aged and older adults in China

The study investigates whether and how energy poverty (EPOV), conceptualized as inadequate affordability of household energy services, affects the cognitive and mental health (CMH) of middle-aged and older adults in China. Against the backdrop of rapid population aging and known links between poverty and health, the paper addresses a gap in the literature concerning EPOV’s effects on cognitive health (CH) and mental health (MH). China provides a salient context due to high aging rates, widespread CMH concerns (e.g., dementia and depression), and notable levels of EPOV. The research aims to quantify EPOV prevalence using affordability-based measures and estimate its impacts on CH and MH, while exploring physical health (PH) as a potential mediating mechanism and examining heterogeneity by age group.

The paper reviews concepts and measures of energy poverty, emphasizing Reddy’s definition that includes access, affordability, reliability, quality, safety, and environmental friendliness. Given China’s near-universal electricity access since 2013, affordability is deemed more relevant than availability for measuring EPOV. Prior work has commonly used fixed thresholds (e.g., 10% of income), amended thresholds for low-income groups, and relative measures such as double-median expenditure shares, as well as LIHC-type approaches. Empirical literature links EPOV to adverse health outcomes, with limited direct evidence on CMH. Indirect evidence connects EPOV to poorer academic achievement (cognitive proxy) and lower subjective wellbeing (related to MH). The review highlights PH as a plausible mediator: EPOV may worsen PH through low-quality fuels, insufficient energy use, and trade-offs like the “heat or eat” dilemma, while PH is strongly correlated with both MH and CH in older adults.

Data: Three waves (2014, 2016, 2018) of the nationally representative China Family Panel Studies (CFPS). The analytic sample includes individuals aged 45+, excluding households using cheap solid fuels and those with zero income. The final unbalanced panel comprises 19,664 individuals from 11,108 households (44,981 person-wave observations); 9,972 individuals are observed in all three waves. Outcomes: CH proxied by self-reported memory over the past week (1–5 scale). MH measured as the average of six CES-D short-form items (depressed, hard to do things, sleepless, lonely, sad, cannot go on with life) on a 1–4 scale (higher is better). Robustness outcomes include math and verbal test scores (CH), and happiness (1–10) and life satisfaction (1–5) (MH). Key explanatory variables: Six affordability-based EPOV measures at the household level (allocated to individuals): EPOV1 (energy share >10% of income), EPOV2 (as EPOV1 and income in bottom 50%), EPOV3 (energy share > double the sample median), EPOV4 (energy share > double the province-specific median), EPOV5 (income bottom 25% and energy share > sample median), and EPOV6 (equal-weight composite of EPOV1–EPOV5, range 0–1). Control variables: age, gender, education dummies, employment, exercise frequency, self-rated social status, marital status, and urban hukou. Empirical strategy: Two-way fixed effects (FE) regressions for CH and MH with provincial and year dummies and individual FE. To address endogeneity (omitted time-varying factors, reverse causality, measurement error, selection), two-stage least squares (2SLS) and FE-IV models use the provincial average of EPOV6 as an instrument. Robustness checks use alternative EPOV measures and alternative outcome measures (limited to the 2018 wave for certain variables). Mechanism analysis tests PH mediation via self-reported health (SRH, 1–5) using a two-step approach (EPOV → SRH; then SRH added to outcome models). Further analyses assess lagged EPOV effects (t−1, t−2) on CMH and heterogeneity by age group (middle-aged <60 vs. older adults ≥60).

- Prevalence: Depending on measure, 24.3%–27.8% of middle-aged and older adults lived in EPOV. - Baseline associations (EPOV6): OLS shows significant negative effects on CH (−0.213, SE 0.015) and MH (−0.146, SE 0.008). With controls and province/year dummies: CH −0.180 (0.015), MH −0.130 (0.008). FE estimates remain negative: CH −0.061 (0.020), MH −0.042 (0.009). - IV results: Using the provincial mean of EPOV6 as an instrument (first-stage F>10), 2SLS and FE-IV show larger negative effects. CH: 2SLS −1.176 (0.252); FE-IV −0.464 (0.263). MH: 2SLS −0.910 (0.127); FE-IV −0.637 (0.127). These suggest baseline estimates were downward biased. - Robustness to alternative EPOV measures (FE): For CH, effects range from −0.040 to −0.052; for MH, from −0.028 to −0.035 (all statistically significant). Alternative outcomes (OLS, 2018): EPOV6 reduces math score (−0.280, 0.071), verbal score (−0.629, 0.192), happiness (−0.368, 0.054), and life satisfaction (−0.091, 0.022). - Mechanism via physical health: EPOV6 lowers SRH (≈−0.148). SRH positively predicts CH (≈0.084) and MH (≈0.065). Adding SRH attenuates EPOV6 effects on CH (to ≈−0.048) and MH (to ≈−0.032), consistent with PH partially mediating the EPOV–CMH relationship. - Lagged effects: EPOV6 at t−1 negatively affects CH (≈−0.105) and MH (≈−0.057); CH effects at t−2 are not evident. Models including current and lagged EPOV show larger cumulative negative impacts, implying prior studies may underestimate EPOV’s total effect when using contemporaneous exposure only. - Heterogeneity by age: Negative effects for both middle-aged and older adults, with stronger impacts on MH among older adults (MH coefficients: −0.024 vs. −0.060; Wald test χ²=5.95, p=0.01). CH differences by age are not statistically significant (χ²=0.08, p=0.78).

The findings demonstrate that affordability-based energy poverty is associated with worse cognitive and mental health among Chinese adults aged 45 and above, addressing a critical evidence gap on CMH outcomes. Results are robust across multiple EPOV definitions and alternative health measures, and remain after addressing endogeneity with IV approaches. The mediation analysis supports a pathway in which EPOV degrades physical health (lower self-reported health), which in turn reduces cognitive and mental health, aligning with prior literature on PH–CMH linkages. The presence of lagged effects indicates that EPOV’s cumulative burden harms CMH beyond immediate periods. Older adults experience more pronounced MH detriments, highlighting vulnerability in later life. These results suggest that reducing energy poverty could contribute meaningfully to healthy aging by improving both mental and cognitive dimensions, not solely physical outcomes.

This study provides nationally representative panel evidence from China that energy poverty, measured through multiple affordability-based indicators, detrimentally affects cognitive and mental health among middle-aged and older adults. Physical health acts as an important channel, and older adults’ mental health is disproportionately affected. The results imply that policies targeting energy affordability (e.g., subsidies, improved energy efficiency, and income support for low-income households) may yield broader health benefits that support healthy aging. Future work could examine additional mediators (e.g., nutrition, indoor air quality), longer-term cognitive outcomes, and external validity across different institutional contexts and countries, using refined causal designs and richer cognitive batteries.

- Potential endogeneity from time-varying unobserved factors, reverse causality, selection, and measurement error is acknowledged; IV strategies mitigate but may not fully eliminate bias. - CMH measures rely on self-reported memory and CES-D items (subject to reporting bias); alternative cognitive test scores and wellbeing measures are available only for specific waves, limiting panel comparability. - EPOV is measured via affordability proxies; availability and quality dimensions are less emphasized given China’s universal access, which may omit other EPOV facets. - The panel is unbalanced; although FE methods and robustness checks are used, attrition could influence estimates.