Crises like pandemics cause severe socioeconomic losses, disproportionately impacting low-income groups. The COVID-19 pandemic, with its lockdowns and restrictions, significantly affected livelihoods, particularly for those lacking savings and relying on mobility for work. The study focuses on Hubei Province, China, a region heavily affected by the pandemic and actively involved in poverty reduction efforts. Understanding the pandemic's impact on recently de-proverished households in Hubei can provide valuable insights for policy development regarding future crisis response and poverty alleviation. The research investigates two primary mechanisms through which individuals might return to poverty: direct health impacts from COVID-19 leading to job loss and increased healthcare costs, and indirect impacts on income due to job losses in vulnerable sectors like casual labor, characterized by single income sources and limited savings. The study utilizes a dataset of 78,931 government-identified poor households across 10 counties in Hubei, reflecting geographical and socioeconomic diversity. The analysis considers three scenarios simulating 1, 2, and 3-month lockdowns, assessing the risk of returning to poverty under each scenario and identifying associated risk factors.

The study reviewed existing literature on the effects of COVID-19 on poverty, highlighting the mechanisms through which the pandemic could lead to a return to poverty. The review informed the identification of key risk factors and helped contextualize the study's focus on vulnerable groups recently lifted out of poverty. It specifically mentions studies that have examined the impact of COVID-19 on various aspects of life, such as global economic impacts, effects on health, and the role of non-pharmaceutical interventions.

The study employed a dataset comprising information on 78,931 government-identified poor households in 100 towns across 10 counties in Hubei Province. Data included total income, income structure (wage, operating, property, and transfer income), and household characteristics (Shannon-Weiner diversity index, proportion of wage income, working hours of migrant workers, family size, family head gender, land ownership, distance to village road, access to power and support from enterprises). Three scenarios, simulating 1, 2, and 3-month lockdowns, were used to estimate the impact on household income, assuming complete wage loss for migrant workers during the lockdown period. The threshold for returning to poverty was defined as an annual household income below 5000 CNY. Two logistic regression models were employed to identify household characteristics associated with the risk of returning to poverty. The first model focused on income characteristics across all counties, and the second incorporated 11 additional household characteristics using data from Yingshan County. The Lindeman, Merenda and Gold (LMG) method and the proportional marginal variance decomposition (PMVD) were used to assess variable importance. A semi-parametric generalized additive model (GAM) was used to predict the risk of returning to poverty at the town level, considering selected regional characteristics, such as distance to railway stations. The GAM's steps included variable selection to address multicollinearity, selecting a beta family link function, model fitting through comparison of different models, and evaluating the model through residual distribution analysis and assessment of the adjusted R² and Akaike information criterion.

The study revealed a significant increase in the percentage of households at risk of returning to poverty under the 3-month lockdown scenario (from 5.6% to 22%). Households with single income sources, shorter working hours, and more family members were identified as high-risk. The proportion of wage income was consistently and significantly correlated with the risk of returning to poverty across all scenarios. Proximity to railway stations emerged as a crucial geographic factor influencing the risk. A 10 km decrease in distance to the nearest railway station increased the risk by approximately 2%. The spatial analysis showed a stable pattern of high-risk regions clustered around railway stations, while low-risk regions were located further away. Interestingly, some areas near railway stations demonstrated lower vulnerability, possibly due to more diversified economic activities and reduced reliance on migrant worker income. The LMG and PMVD methods consistently identified distance to the nearest railway station as the most important factor related to poverty risk.

The findings highlight the vulnerability of households reliant on single income sources, particularly those employed in sectors significantly affected by lockdowns. The concentration of high-risk regions around railway stations underscores the impact of mobility restrictions on migrant workers, a significant segment of the low-income population. The study's findings provide crucial evidence for targeted policy interventions focusing on vulnerable households and regions. The consistent importance of proximity to railway stations suggests that policies should consider transportation access and potentially support diversification of income sources in these areas. The contrasting results for areas near railway stations – some highly vulnerable and others less so – suggest the importance of considering local economic diversity. This research has significant implications for policymakers, offering insights to better target assistance and mitigate the socioeconomic impacts of future crises.

This study provides valuable insights into the specific household characteristics and geographic locations most vulnerable to returning to poverty during the initial phases of the COVID-19 pandemic. It emphasizes the need for targeted interventions focusing on households with limited income diversity, particularly those near transportation hubs. Future research should investigate the longer-term impacts of the pandemic, expand the sample to include a broader population, and explore other dimensions of poverty affected by the pandemic. Further studies focusing on the combined risk of multiple shocks and broader implications for the Sustainable Development Goals (SDGs) are also necessary.

The study’s analysis was limited to the 3-month lockdown period, potentially underestimating the long-term impacts. The focus was on households previously identified as poor, excluding those who may have fallen into poverty during the pandemic. The analysis primarily focused on income loss, neglecting other impacts of the pandemic, such as impacts on healthcare, education, and housing. Further, data on household characteristics was only fully available for Yingshan County, limiting the generalizability of some findings.