Poverty remains a significant global challenge, and while China has eliminated absolute poverty, relative poverty persists. This study focuses on the dynamics of multidimensional poverty, acknowledging that poverty encompasses various dimensions beyond income. Existing research often employs static analyses or examines only income poverty dynamics. This study addresses these gaps by investigating the dynamics of multidimensional poverty among middle-aged and older adults in China, a population particularly vulnerable to poverty due to aging and potential physical and cognitive decline. The study utilizes the Alkire-Foster method to construct a Multidimensional Poverty Index (MPI) and incorporates dynamic analysis techniques to understand the transitions into and out of poverty. The theoretical framework integrates human capital theory, life cycle of poverty theory, social capital theory, and spatial poverty trap theory to examine the determinants of poverty dynamics. The study hypothesizes that individual characteristics (gender, age, marital status), family structure and living arrangements, social capital, and geographic location will significantly influence the probability of exiting and re-entering multidimensional poverty.

The literature review highlights the shift from a solely monetary perspective on poverty to a multidimensional approach, emphasizing Sen's capacity approach and the widely used Alkire-Foster method for MPI calculation. The review also notes a growing interest in dynamic poverty analysis, moving beyond static snapshots to understand poverty durations and transitions. While studies on income poverty dynamics are abundant, research on multidimensional poverty dynamics, particularly among older adults, is limited. The existing studies often analyze adjacent years or focus on the persistence of poverty, but lack a comprehensive, longitudinal analysis of multidimensional poverty dynamics considering the cumulative effect of durations. This study aims to fill this gap by analyzing the entire period from 2011 to 2018 and focusing on the older adult population.

This study utilizes four waves of data (2011, 2013, 2015, and 2018) from the CHARLS, a nationally representative longitudinal study of Chinese adults aged 45 and older. The study employs the Alkire-Foster method to construct the MPI, incorporating dimensions of education, health, living standards, social security, and employment and income, with indicators specified in Table 1. Normative weights are used, and a poverty cutoff of k=2 is applied. The analysis includes descriptive statistics, poverty transition matrices to analyze adjacent-year dynamics, Kaplan-Meier estimates to assess survival and hazard rates of poverty spells, and a discrete-time proportional hazards model to identify the determinants of poverty transitions. The determinants considered include poverty/non-poverty durations, individual characteristics (gender, age, marital status), family structure (household size, working population, living with children), social capital (economic transfers, social activities), and geographic location (urban/rural, region).

From 2011 to 2018, the MPI (M0) showed a downward trend primarily driven by a decrease in the headcount ratio (H), while the poverty intensity (A) remained relatively stable. Decomposition analysis revealed that education consistently had the highest contribution to poverty, followed by health, indicating the increasing importance of health in the overall welfare of older adults. Transition matrices showed a decreasing rate of persistent poverty and an increasing rate of persistent non-poverty. Kaplan-Meier estimates revealed that most individuals experienced transient poverty, with the probability of exiting or re-entering poverty decreasing with the duration of the respective state. The discrete-time proportional hazards model confirmed that longer poverty durations reduced the probability of exiting poverty, and longer non-poverty durations reduced the probability of returning to poverty. The model also showed that men, being married, larger household size, economic transfers, social activities, and urban residence increased the likelihood of exiting poverty, while these same factors (except for economic transfers with close relatives and living with children) decreased the likelihood of re-entering poverty. Older individuals (aged 65+) had a lower probability of exiting poverty than those aged 45-55.

The study's findings highlight the importance of addressing multidimensional poverty dynamics among older adults in China. The stable poverty intensity despite a decrease in the headcount ratio suggests that interventions need to focus on improving the welfare of those remaining in poverty. The increasing contribution of health to overall poverty underscores the need for improved healthcare access and support for chronic disease management. The negative association between poverty durations and transition probabilities supports the need for targeted interventions for those experiencing persistent poverty. The results also emphasize the role of family support and social capital in mitigating poverty risks. Policy implications include strengthening health monitoring systems, providing targeted support for persistent poverty, early interventions for newly impoverished individuals, and ongoing monitoring for those exiting poverty. Promoting social activities and diversified care models for older adults is also crucial.

This study provides valuable insights into multidimensional poverty dynamics among older adults in China, highlighting the importance of health and family support, and the need for targeted, dynamic interventions. Further research should include more longitudinal data to examine re-entry into poverty more thoroughly and investigate the impact of events like the COVID-19 pandemic on multidimensional poverty.

The study is limited by the available data, which spans only four time points, possibly limiting the precision of dynamic analysis, particularly regarding poverty re-entry. The reliance on self-reported data may also introduce biases. Future research with more frequent data collection and a wider range of determinants would enhance the findings.