An understanding of the dynamics of persistent poverty that is not static, contrary to its portrayal in the literature, is important for policy‐makers. We therefore performed multinomial logistic regression and binary logistic regression on panel data to capture rural poverty dynamics in Nepal's Far‐Western Hills. The former revealed that the risk of experiencing chronic poverty was significantly higher for households in the more remote village development committee of Melauli, those headed by women or older men with less education, and those with higher dependency ratios and lower participation levels in community‐based organisations. The latter, entailing disaggregation of transient poverty, revealed a higher risk of moving into poverty for members of occupational castes (OCs) and housevholds in Melauli because of their limited socioeconomic assets. The occurrence of natural disasters also increased households’ risks of moving into poverty. Conversely, more years of schooling among household heads facilitated movements out of poverty. Thus, poverty reduction programmes focusing on relatively remote villages and OCs that promote educational access and employment generation are necessary to address chronic and transient poverty. Further, the introduction of temporary relief programmes following natural disasters, and of insurance schemes for covering associated losses, will alleviate transient poverty.