! ! town households only. Comparing dynamics of The ESS uses a stratified, two-stage sampling scheme. multidimensional and consumption- Enumeration areas (EAs) were randomly selected in based poverty in Ethiopia proportion to population size; 290 and 43 EAs were selected from rural and small town areas, respectively, Poverty can be viewed as taking many different forms, and twelve households were chosen from each EA. ranging widely over a set of monetary (consumption or Tracking between waves was done at the household income) and nonmonetary dimensions (health and level-- with a low attrition rate of 4.9%-- leading to a education). Recent literature documents that people panel sample of 3,776 households. We further restrict who are identified as poor in the consumption space the final analytical sample to exclude households that are often different from those who are multi- are missing information on any of the nine dimensionally poor (MDP). However, less is known deprivations or real consumption per adult equivalent, about whether the dynamics of MDP are similar to the for a final balanced sample of 3,197 households. dynamics of a relative consumption-based measure of poverty. Using two waves of panel data from the Methods!! Ethiopia Socioeconomic Survey (ESS), we explore the We used the OPHI methodology as a guide when correlation between monetary and nonmonetary constructing the underlying weighted deprivation measures of poverty and wellbeing in the cross-section index (k) used to define MDP. We incorporate three and dynamically in rural and small town Ethiopia dimensions of wellbeing-- education, health, and living between 2012 and 2014.! standards-- with each dimension weighted to represent one-third of the deprivation index, and each individual Background!! indicator weighted equally within a given dimension. K While the body of literature on poverty dynamics is takes some value between 0 and 1, with 0 indicating no extensive, the majority of studies draw conclusions deprivations and 1 signifying deprivation in every about the dynamics of income- or consumption-based indicator. poverty only; there is a growing, but still relatively young, literature base on the dynamics of MDP. Even In order to classify a household as poor or non-poor, more elusive is the correlation between the dynamics a minimum number of weighted dimensions are of MDP and consumption-based poverty; is there established and only those who are deprived in signal between changes in multidimensional wellbeing dimensions exceeding this value are considered poor. and changes in consumption? In this brief we use a value of k in each wave such that the proportion of individuals experiencing MDP Data-- matches the proportion of individuals facing relative We analyze panel data from two waves of the ESS, a consumption-based poverty (approximately 30% collaboration between the Central Statistics Agency of among rural and small town areas). By allowing k to Ethiopia (CSA) and the World Bank’s Living Standards change each year, this estimate (hereafter referred to as Measurement Study- Integrated Surveys of Agriculture MDEP) can similarly be thought of as a relative non- (LSMS-ISA) project that collects multi-topic panel data monetary estimate of poverty. at the household level. The ESS began in 2011 (ESS1), with 3,969 rural and small town households. In 2013, a Results-- second wave (ESS2) was administered, revisiting the Despite defining both measures of poverty to capture ESS1 households and an additional 1,500 urban the bottom 30% of their underlying distributions, we households; the panel sample includes rural and small find that only 27% of individuals that are poor in either dimension, are poor in both dimensions. We also find stories, we find evidence suggesting that changes in the little overlap between quintiles of annual consumption two underlying values of k and consumption are in fact per adult equivalent and k in 2014 (see Table 1). Only not linked; i.e., knowing what happens to an 25% of the rural and small town population fall in the individual’s k between waves does not help us know same quintile of both distributions, 35% of individuals what happens to that individual’s consumption over are one quintile apart when comparing the two the same period, and vice versa. Approximately 58% indicators, and 40% are two or more quintiles apart. of individuals whose k worsened between waves also This shows that whether we use a monetary or non- experienced a decline in consumption; the other 42% monetary measure of poverty has a meaningful impact saw an improvement in their consumption (see Table on who will be identified as poor at a given point in 2). Similarly, nearly 53% of individuals who improved time. In fact, 75% of individuals would be placed in a in k actually experienced a worsening in consumption. different quintile depending on whether or not we In fact, using Pearson’s chi-squared test of viewed wellbeing as being defined by consumption or independence, we fail to reject the null hypothesis that deprivations in non-monetary dimensions. the two distributions are independent (p=0.234). Table 1. Crosstab of consumption and k quintiles, 2014 Table 2. Contrasting changes in k and consumption k Consumption Quintiles of k Real consumption Worsened Stayed Improved Total quintiles Poorest 2nd 3rd 4th Top per adult equiv. the same Poorest 5.68 3.96 5.59 3.31 2.36 Worsened 0.193 0.108 0.247 0.547 2nd 4.91 3.28 4.44 3.63 3.73 Improved 0.140 0.094 0.219 0.453 3rd 3.65 3.27 5.33 4.49 3.96 4th Total 0.333 0.202 0.466 1.000 2.12 3.45 4.57 4.00 5.35 Note: In a Pearson’s chi-squared test of independence, we fail to reject Top 1.24 2.33 4.90 3.92 6.56 the null hypothesis that the two variables are independent of each other, at p=0.234. Observations are weighted to make results representative of all rural and small town individuals in Ethiopia. Balanced panel sample When comparing the dynamics of the two poverty size includes 3,197 households. indicators, separately, we observe similar levels of movement in and out of poverty. Eighteen percent of Discussion-&-Policy-Implications- rural and small town Ethiopians face chronic MDEP, Our finding that k and consumption are not necessarily which is only slightly higher than the 15% identified as co-moving, has important implications for how we chronically poor using traditional consumption-based assess individuals’ progress in improving wellbeing estimates (see Figure 1). There is slightly elevated over time. Until more is learned about precisely what movement in and out of consumption-based poverty, each of these measures is picking up, our analysis with nearly 31% changing status between 2012 and indicates that a policymaker could be missing 2014; only 25% of individuals transitioned between important changes in wellbeing by focusing only on multi-dimensionally poor and non-poor states. monetary or non-monetary measures of wellbeing or Figure 1. Dynamics of MDEP and consumption- poverty. Until further evidence provides more based poverty understanding of what each of these indicators is MDEP Consumption-based capturing, both should be tracked. poverty Wave 2 Wave 2 Generous funding assistance for this research came from the Poor Not poor Poor Not poor UK Department for International Development Ethiopia. Poor The findings outlined in this brief are drawn from: Seff, I. & 17.5 12.5 14.5 14.6 Wave 1 Jolliffe, D. (forthcoming) “Multidimensional poverty dynamics Not in Ethiopia: How do they differ from consumption-based 12.0 57.6 16.1 54.8 poor poverty dynamics?” However, even though the dynamics of MDEP and To access the ESS data: hhtp://go.wordlbank.org/ZK2ZDZYDD0! relative consumption-based poverty seem to tell similar !!!!!!!!!!!! ! !