Despite its importance (for example in order to distinguish the nature of chronic and transient poverty), good studies on poverty dynamics in Indonesia is quite rare. One important constraint is the need of panel or longitudinal data where observations of the living conditions of the same individuals or households are made at several points in time. In developing countries, the existence of this type survey, of course, is not a norm. In Indonesia, longitudinal household survey was not present until in 1993, Rand cooperation and Demographic Institute of University of Indonesia, conducted Indonesian Family Survey (IFLS), in which around 7,600 household in 13 province in Indonesia were survey and re-surveyed in 1997. This two set of data (to our knowledge) resulted in the first dataset that was specifically designed to be a panel/longitudinal data. In this paper, we would utilize this rich set of data, to study poverty dynamics in Indonesia, by dividing poverty into transient and chronic poverty. We would investigate socioeconomic characteristics of those households that belongs to transiently-poor and chronically-poor group to see any distinctive characteristics between the two. In addition to that, we would investigate whether regional factors i.e. urban-rural location, and geographical difference (such as across province or localities) would also reveal distinctive patterns.