Poverty remains a persistent structural problem in Indonesia, particularly in provinces with weak economic structures and high dependence on low-productivity sectors, where economic growth has not always translated into equitable welfare improvements. This study aims to examine the effect of Gross Regional Domestic Product (GRDP), Farmer Exchange Rate (NTP), Farmer Labor Force (TKP), Expected Years of Schooling (EYS), and Open Unemployment Rate (OUR) on poverty levels in the 20 provinces with the highest poverty rates in Indonesia. This research employs a quantitative approach using dynamic panel data analysis. The population consists of the 20 poorest provinces in Indonesia, with annual observations from 2015 to 2024. The data are secondary data obtained from the Central Statistics Agency (BPS). Data were collected through documentation techniques and analyzed using the Generalized Method of Moments (GMM) to address endogeneity, autocorrelation, and unobserved heterogeneity, with model validity tested using Sargan and Arellano–Bond tests. The estimation results indicate that individually GRDP, NTP, TKP, EYS, and OUR do not have a statistically significant effect on poverty in both the 2015–2019 and 2015–2024 periods. However, the Wald test shows that all variables simultaneously have a significant effect on poverty, indicating strong interdependence among economic, labor, and human development factors. These findings imply that poverty reduction in the poorest provinces cannot rely solely on economic growth or employment expansion, but requires improvements in job quality, labor productivity, human capital, and inclusive economic transformation to ensure that growth benefits reach low-income populations.