1. A poor life? Chronic
poverty and downward
mobility in rural
Ethiopia, 1994-2004
Stefan Dercon and Catherine Porter
University of Oxford
2. An applied analysis of rural Ethiopia
• Using a (by now well known) panel dataset
from rural Ethiopia we explore poverty
dynamics between 1994 and 2004
• Looking `trajectories’ of the poor over time
• Identification (in comparison with snapshot)
• What are the determinants of chronic poverty
and downward mobility?
3. Summary of findings
• 1984 Famine is still an important predictor of
chronic poverty
• Chronic poor have fewer assets in 1994, but do
actually have an upward growth path over time
• Illness and agricultural shocks are important in
pushing people down to a point from which they
cannot recover
• Suggests different and complementary policy
approaches for the chronic and declining poor
4. Methodology
• Static poverty measurement (snapshot) is
fairly established in economics
• Chronic poverty literature has begun to bloom
with several competing measures
• No consensus as yet on how to aggregate
poverty over time
• We extend the well-known FGT poverty
measures by discounting the past to identify
different trajectories of poverty
5. ERHS Data
• Collected by IFPRI with Oxford, Addis Unis
• Use data from rounds 1-6 of ERHS (1994-2004,
excluding R2)
• 15 villages in Ethiopia, 1100 households in this
analysis
• Relatively representative though small
• Use consumption as our welfare indicator (per
adult equivalent, adjusted for inflation)
6. Static Poverty over time
Squared Mean
Poverty poverty consumption
Year Headcount gap gap among poor (Birr)
1994 0.39 0.16 0.09 26.84
1995 0.44 0.19 0.10 25.72
1997 0.24 0.08 0.04 30.31
1999 0.28 0.09 0.04 31.13
2004 0.22 0.07 0.03 30.29
7. Who are the poor 10 years later?
• Of the households who were classified as poor in
1994:
– just under a third were still poor in 2004
– comprising 9.7 per cent of all households.
• A higher percentage of households (24.8 per cent)
moved out of poverty than fell into it (10.2 per cent)
• Just over half (55.3 per cent) of all households in the
sample were non-poor in both 1994 and 2004
• This helps explain the sharp decline in the headcount
poverty rate noted above
8. Number of periods in poverty
Number of
Times in Poverty households Per cent Cum. %
Never 420 36.02 36.02
Once 291 24.96 61.07
Twice 202 17.32 78.51
Thrice 142 12.18 90.27
Four times 81 6.95 97.43
Always 30 2.57 100
Total 1,166 100
10. Characteristics of the chronic poor in
rural Ethiopia
• Geographically concentrated with pockets of
deprivation.
• Lowest rates amongst the two biggest ethnic groups
• Average number of ill hh members>1
(non chronic poor=0.5)
• 40% have low BMI
(compared to 22% of non-chronic poor)
• Famine affected areas in 1984 are a very strong
predictor of chronic poverty
• Road improvement, cultivating coffee and having
assets (livestock, schooling) in 1994 reduce
probability of chronic poverty
11. Are the Chronically Poor and Declining
Poor Different?
• Some similarities but the declining poor have
more assets than the chronic poor)
• Rainfall and agricultural shocks matter more in
pushing people into poverty
• High share of agricultural income in 1994 reduces
the probability of declining into poverty
• Schooling and gender of household head not
predictors of declining into poverty
12. Summary & Policy Implications
• In 1994, 39% of EHRS households were poor
• Of these, only a few would stay in poverty, and many
others would fall into poverty
• 1984 famine plus in 1994 having low assets, low
schooling and female head predict chronic poverty in
the coming decade
• Having higher share agricultural income, experiencing
shocks or poor rainfall predict downward trajectories
• Suggests asset building is the key to reducing chronic
poverty, including education, as well as employment
opportunities.
• And more comprehensive safety nets are needed to
complement this for chronic poor and for vulnerable
Notas do Editor
Lots of conceptual work going on in this area- ours is a pragmatic, applied approach looking at the empirical trends in this well known dataset.
NB poverty line is 2000 calories plus some basic non-food expenditure (e.g. firewood)
Identification of chronic poor then: if 3 periods of poverty or more, then 22% poor, 4 periods then 10% poor etc. NB Jalan Ravallion – 13% of households have average consumption below the poverty line.
[Gara Godo (over a third), Geblen, Doma’a, Imdibir and Korodegaga. ] Coffee producers not chronic poor (enset yes)
A few surprises here compared to chronic poor. Adelekeke is the one with declining poor. And only 10% in this case
Some suggestion that multiplicative shocks also occur. The chronic poor some are getting out… but some are more vulnerable to shock