During the economic crisis in Indonesia, the headcount rate on poverty changed relatively quickly over short periods of time. Poverty increased rapidly when the crisis worsened and, likewise, decreased rapidly when the economy stabilized. This implies that large numbers of households were moving in and out of poverty relatively frequently. It also implies that a significant number of households experienced relatively short periods of poverty, that is, just a fraction of a year.
Generally, the movement of households in and out of poverty is assessed on a yearly basis (for example, Bane and Ellwood 1983; Baulch and Hoddinott 2000; Jalan and Ravallion 1999a, 2000). These studies utilize panel data of households with a year as the basic time unit. According to these data, a household deemed not poor in two consecutive surveys will be considered as having never been poor during the whole period between the two surveys. In reality, however, the household could have experienced a period of poverty in between the two surveys. Such a situation could occur, for example, if each year the survey was conducted in the harvest season, a period when rural households are generally better off.
To understand the short-term dynamics of poverty, it is necessary to have panel data that rely upon a time unit that is less than a year. Muller (1997), for example, uses quarterly panel data in a one-year period between 1982 and 1983 to estimate the transient seasonal and chronic poverty of peasants in rural Rwanda. He finds that the worst poverty occurs after the dry season at the end of the year. Severe poverty is generally the result of a seasonal, transient component of annual poverty, where the seasonal component of the incidence of poverty is much smaller. Hence he concludes that the actual differences in the severity of poverty, either between developing and developed countries or between rural and urban areas in developing countries, may be much worse than shown by the usual chronic annual poverty measures or by measures of the seasonal incidence of poverty.
Similarly, Dercon and Krishnan (2000) use a panel data set of households in rural Ethiopia that were visited three times over an eighteen-month period.