Poverty remains a persistent development challenge in East Nusa Tenggara (NTT), reflecting structural constraints in an archipelagic region with high inter-district heterogeneity. Most previous studies on poverty determinants in NTT rely on static models and therefore have limited capacity to capture poverty persistence and potential endogeneity between poverty and its determinants. This study aims to analyze the determinants of district- and city-level poverty in NTT from 2010 to 2024 using a dynamic panel approach. The dataset comprises 22 districts and cities observed over 15 years, yielding 330 observations, and is sourced from Statistics Indonesia (BPS). Estimation is conducted using the First Difference Generalized Method of Moments (FD GMM) with one-step robust standard errors to address dynamic panel bias and endogeneity concerns. The results indicate that poverty is persistent, as shown by a positive and significant lagged poverty coefficient. Economic growth, proxied by lnGRDP, significantly reduces poverty, while human capital indicators, such as years of schooling (MYS) and life expectancy (LE), also exert significant poverty-reducing effects. In contrast, the open unemployment rate (OUR) and population size (lnPOP) are not statistically significant. Notably, the minimum wage (lnMW) is positively and significantly associated with poverty, suggesting potential employment or adjustment pressures when wage policy is misaligned with productivity. Policy implications emphasize the need for inclusive growth, strengthened investments in education and health, and a minimum wage design consistent with local productivity and labor-market structure.