Dynamics of multidimensional poverty and its determinants among the middle-aged and older adults in China

The study addresses the challenge of relative and multidimensional poverty among China’s mid-aged and older adults following the elimination of absolute poverty by 2020. Building on the shift from income-based to multidimensional poverty analysis and from static to dynamic perspectives, the paper identifies a gap: limited research on multidimensional poverty dynamics, especially among older adults, and insufficient integration of poverty durations with transition dynamics. The research question is to understand how multidimensional poverty evolves over time in this population and which factors determine exits from and returns to poverty. The study aims to quantify trends, persistence, and transitions in multidimensional poverty from 2011–2018 and to identify determinants across individual, household, social capital, and spatial domains, hypothesizing that durations reduce transition probabilities and that demographic, family, social capital, and regional characteristics shape exit and reentry risks.

Prior work expanded poverty assessment from income to multidimensional frameworks (e.g., Alkire-Foster method), typically using cross-sectional data to measure incidence and correlates. Dynamic poverty research has largely focused on income poverty (durations and transitions), with fewer studies on multidimensional poverty dynamics. Existing multidimensional studies often analyze adjacent-year transitions or chronic poverty indices but tend to separate duration from transition analysis, missing cumulative duration effects. Studies focusing specifically on older adults are rare and largely static. This paper fills these gaps by examining multidimensional poverty dynamics over multiple waves for mid-aged and older adults, integrating duration with transitions and assessing determinants.

Conceptual framework and hypotheses: Drawing on human capital (Schultz), life-cycle poverty (Rowntree & Bradshaw), social capital (Bourdieu), and spatial poverty trap theory, the study proposes that individual characteristics, family structure and living arrangements, social capital, and living areas affect poverty dynamics, and that durations reduce transition probabilities. Hypotheses: H1 (gender, age, marital status), H2 (household size, number of workers, living with children), H3 (economic transfer and social activities), H4 (rural/western disadvantages), H5 (longer durations reduce transition probability). Data: Four CHARLS waves (2011, 2013, 2015, 2018) for respondents aged ≥45. Included those participating in all four waves; excluded those <45 in 2011. Final analytic sample n=11,566. Measurement—MPI: Constructed using the Alkire-Foster method with dual cutoffs. Dimensions and indicators (14 total) across five dimensions: education (schooling), health (chronic diseases, ADL limitations, depression), living standards (running water, durable goods, housing structure, flushable toilets, bathing facilities, cooking fuel), social security (health insurance, pension insurance), employment and income (employment, per-capita household income). Normative weights assigned as listed in the paper; poverty cutoff k=2 (poor if deprived in at least two dimensions). Computed headcount (H), intensity (A), and MPI (M0=H×A). In education, deprivation defined as not completing primary school. Health included both physical and mental indicators; social security covered social/commercial insurance; income used per-capita net household income. Determinants: Five groups: durations (poverty duration until exit; nonpoverty duration until return), personal characteristics (gender, age, marital status), family structure and living arrangements (household size; nonworking population ratio; proportion aged <18 or >60; living with/near children), social capital (economic transfers with close/distant relatives; participation in social activities), living areas (urban/rural; region: East/Middle/West). Analytical methods: (1) Descriptive statistics for trends and decomposition. (2) Poverty transition matrices for adjacent waves (four states: poor→poor, poor→nonpoor, nonpoor→poor, nonpoor→nonpoor) on full sample (n=11,566). (3) Kaplan–Meier survival analysis to estimate exit and reentry: exit analysis included 3,155 individuals poor in 2011; reentry analysis included 1,616 who exited poverty in 2013; estimated survivor and hazard functions over time. (4) Discrete-time proportional hazards models (complementary log-log link) to assess determinants of exit and reentry; after excluding observations with extensive missing data on covariates, final model sample n=11,315. Duration dummies set relative to longest spells (8 years for exit; 6 years for reentry).

- Overall MPI trend: From 2011 to 2018, headcount ratio (H) fell from 27.3% to 13.2%, while intensity (A) stayed roughly stable (59.4% to 56.5%); MPI (M0) declined from 0.162 to 0.075. H dropped notably in 2011–2013 and 2015–2018, with a slight increase in 2013–2015 (Table 3). - Decomposition: Education consistently contributed most to MPI; health’s contribution rose annually, indicating growing importance of physical and mental health. Living standards and social security contributions trended downward overall. Employment and income contribution rose then fell. In 2018, top indicators were schooling (0.348), employment (0.134), ADL (0.079), chronic diseases (0.077), and depression (0.067). Low contributions persisted for running water, housing structure, medical insurance, durable goods. Income and pension insurance contributions dropped sharply from 2015 to 2018 (Table 4). - Transitions (adjacent years): Persistent poverty decreased: 13.31% (2011–2013) → 11.42% (2013–2015) → 8.01% (2015–2018); persistent nonpoverty increased: 67.72% → 72.16% → 73.84%. New entries into poverty persisted in every interval (Table 5). - Kaplan–Meier exits: Of 3,155 poor in 2011, 48.78% remained poor by 2013; 17% remained poor by 2018 (7-year spells). Exit hazard generally declined with duration, with some rebound 2015–2018, coinciding with intensified poverty alleviation (Table 6). - Kaplan–Meier reentry: Of 1,616 who exited poverty in 2013, 64% remained nonpoor by 2015 and 55% by 2018 (36% and 9% reentered by those times). Reentry hazard decreased over time, indicating longer nonpoverty spells reduce reentry risk (Table 6). - Discrete-time hazards (Table 7): Longer durations substantially reduced transition likelihoods (supporting H5). Exit from poverty was more likely for men, married individuals, larger households, those with economic transfers (close or distant relatives), those participating in social activities, and urban residents; less likely for age ≥65 and for residents in the Middle vs East. Reentry into poverty was less likely for men, married individuals, larger households, those living with/near children, those with economic transfers with distant relatives, and those participating in social activities. Urban residence did not significantly reduce reentry. Middle and West regions showed higher (but not always significant) reentry relative to East.

The findings demonstrate that while multidimensional poverty incidence fell, intensity among the poor remained stable, implying limited welfare gains for those who remained poor. Health factors became increasingly central to overall deprivation among mid-aged and older adults, highlighting the need to prioritize chronic disease management, functional support (ADL), and mental health to further reduce poverty. The negative association between transition probabilities and durations indicates strong state dependence: long-poverty spells hinder exit, and sustained nonpoverty reduces reentry, guiding timing and targeting of interventions. Individual and household characteristics shape dynamics: men and married individuals fare better in both exiting and avoiding reentry, suggesting gendered and social support advantages; larger household size and co-residence with children likely provide care, economic, and psychosocial support that facilitate exit and prevent reentry. Social capital—economic exchanges and social activity participation—supports exits and limits reentries, affirming the role of social networks. Spatially, urban residence facilitates exits, likely via better access to services (especially health care), and the Middle region lags the East, aligning with macroeconomic disparities. These results address the research question by clarifying who exits or reenters poverty and when, underscoring policy relevance for dynamic monitoring and targeted, duration-sensitive support.

- From 2011–2018, multidimensional poverty incidence among mid-aged and older adults declined markedly, while average intensity remained stable. - Health emerged as a leading contributor to deprivation; by 2018, schooling, employment, ADL, chronic diseases, and depression dominated contributions. - Most poverty spells were transient; exit and reentry coexisted across adjacent waves. - Duration strongly shaped dynamics: longer poverty reduced exit rates; longer nonpoverty reduced reentry. Individual, family, social capital, and spatial factors significantly influenced both exits and reentries, often with opposite effects across the two transitions. Policy implications include strengthening health monitoring and services for older adults; early, targeted assistance for new poor; sustained monitoring of recent exiters to prevent reentry; bolstering family and community-based supports; and addressing regional and urban–rural disparities. Future research should leverage more waves to refine reentry analysis and examine COVID-19 impacts on multidimensional poverty dynamics.

The analysis of reentry into poverty was constrained by limited rounds of data, which may affect precision in modeling reentry dynamics. The study also relied on respondents present in all four waves and excluded cases with substantial covariate missingness in hazard models, which may limit generalizability.