Based on a rural and urban data set from Ethiopia, exiting from or re-entering poverty were found to depend on the time spent in or out of poverty. In comparison to urban areas, exiting rural poverty was easier than re-entering it. However, exiting poverty was extremely difficult the longer households were in that state, even more in urban than rural areas. In addition, the average time spent in poverty following a poverty spell is quite long for a typical household. Time-varying and other household characteristics were examined in the context of exiting and re-entering into poverty. Features of chronic poverty and vulnerability were also analyzed and the policy implications discussed.
Frequently used aggregate measures of poverty such as the headcount ratio, do not account for past experiences of poverty. Some might have already spent many years in persistent poverty, others might have just fallen into poverty, and still others might have just escaped poverty but have a high probability of falling back in. The fist category represent the chronically poor, the second (hopefully) the transient poor and the third the vulnerable. The distinction of these features of poverty, along with the time-varying and individual-specific determinants is very important for policy purposes.
Recent literature on the dynamics of poverty focuses on the mobility across a given income threshold or poverty line, and attempts to distinguish chronic from transient poverty1. A household’s consumption level at a specific time depends on its assets, and its ability to smooth consumption. If the household is credit constrained it may find it hard to cope with negative shocks. Chronic poverty can thus depend not only on current income but also on the household’s lack of assets or its limited ability to translate assets into incomes. Incomes change over time by asset accumulation, changes in returns driven by savings behaviour or exogenous shocks.2 Household income depends on the gender, education and other characteristics of its members, the changing size of the household due to fertility and migration decisions, as well as the state of the labour market. Part of the exercise in poverty dynamics is to investigate how these factors influence the persistence of poverty.
The dynamics of poverty has generally been assessed in two ways, the spells-approach focusing on transitions in and out of poverty, and the components approach, separating the chronic from transient component of poverty (Hulme, Shepherd, 2003, Jalan and Rvallion, 2000). To identify the chronic component of poverty, one can use average consumption over several periods (Rodgers and Rodgers, 1991). The spells approach is a powerful tool for understanding how the transient poor can emerge again from poverty if the analysis can clearly identify the factors that underlay their falling. But, to understand chronic poverty one needs to analyse social structures and mobility, or rather immobility, within them.
The discussion of transient poverty leads quite naturally to the discussion of vulnerability, which is not necessarily captured by current income estimates. What one would like to know is the extent to which households near the poverty line have assets that can serve as buffers against shocks. The shocks can be of several kinds, from droughts affecting agricultural output, to unemployment, illness or death of members of the household. Liquid assets (monetary assets or livestock, although in a general crisis the prices of livestock can collapse) can help protect households against these shocks. Households may also be able to incur debt, sell other assets than livestock, or pull children out of school. They may also draw on their social networks or in the end rely on support from government or other institutions.
There have been few empirical studies on the dynamics of poverty. Bane and Ellwood (1986, p. 2-4) classified approaches to the study of poverty dynamics into tabulations of poverty over some fixed periods, methods using spell-durations and exit-probabilities and statistical methods which model the level of some variable such as income, allowing for complex lag error-structure.
McKay and Lawson (2003) reviewed the evidence on chronic and transient poverty noting that many studies had concluded that transient was more important than chronic poverty, though they themselves were sceptical. They belive that sometimes too stringent conditions had been imposed for a household to be classified as chronically poor, and also there were measurement errors that might explain why a household at some point in time seemed to escape from poverty. Yaqub (2003) reports evidence from 23 countries on factors that explain upward mobility, which was correlated with more land, and more education, while downward mobility was correlated with increased household size and the number of dependents. Dercon and Krishnan (2000) explored short-term vulnerability of rural households in Ethiopia finding that poverty rates were very similar over the 18 months over three surveys, although consumption variability and transition in and out of poverty were high.
This paper examines poverty persistence, chronic poverty and vulnerability using both the spells and components approach on a rich panel data set that covers approximately six years in four waves. To our knowledge such empirical work, notably one based on the spell approach, is rare for less-developed countries, and non-existence for Africa.
The next section outlines the methods used to capture poverty transitions, chronic poverty and vulnerability, section 3 describes the data and report exit and re-entry probabilities for various household types and separating the transient from the chronic components of poverty. Section 4 reports the determinants of chronic poverty and vulnerability and discusses the policy implications. Section 5 summarizes and draws conclusions.
See surveys in Baulch and Hoddinott (2000), Hulme and Shepherd (2003), McKay and Lawson (2003), and Yaqub (2003).
Gunning et al (2000) have investigated the income dynamics in the resettlement areas of Zimbabwe. They had data on asset accumulation over time and combined this with estimates of changes in asset returns in an interesting analysis of a process of income convergence. There is little evidence in the literature on the cumulative income of shocks to households.