Review of Poverty Diagnostics in African PRSPs by keralaguest


									                  Review of Poverty Diagnostics in African PRSPs
                                       Lionel Demery, AFTPM

      I.      Introduction

There are two essential strengths to the PRSP process. First, it emphasizes country ownership,
with ‘country’ being interpreted to include government and the wider civil society. Second, it is
organized around achieving poverty reduction goals, and the needed actions to achieve them.
Both are challenging features, and can in many cases involve a major departure from past
practice. Setting goals and identifying effective actions is particularly difficult, requiring as it
does a great deal of information and knowledge about the determinants of poverty across its many
complex dimensions. More specifically it calls for an understanding of the impact on poverty of
public action. Such knowledge can only be a reliable guide to policy if it has a solid empirical
basis, and has emerged from a careful poverty diagnosis. This note investigates the extent to
which the completed and approved Poverty Reduction Strategy Papers (PRSPs) in the Africa
Region have such a basis.

    The review relies entirely on the PRSP documentation (the PRSP, the PRSP Progress Report
where applicable, and the Joint Staff Assessments). This is a significant limitation, since the
documentation rarely provides technical detail, or even cross references to the technical work that
underpinned the document. Because of this, our ability to assess the quality of the underlying
poverty diagnostic analysis is constrained. The review focuses on those products that have
cleared due process, both within the country and at the Bank and Fund Boards. In all, seven
PRSPs (and their related JSAs) are covered: Burkina Faso, Mauritania, Mozambique, Niger,
Tanzania, Uganda and Zambia. The review therefore suffers to some extent from selectivity bias,
in that the countries with completed PRSPs are most likely to have been better prepared with
poverty data and poverty diagnostic that those that have yet to complete their first PRSPs. In part
motivated by this consideration, the review adds a post-script on the evidence from a selection of
Interim PRSPs.1 The Annex table gives some of the detail that is drawn upon in this note.

      The approach we have taken in this review is to ask three basic questions:

       Is the information base of the PRSP. (i.e. the available data) relevant and timely?

       Is the knowledge base sufficient: are the determinants of poverty well understood?

       Has policy impact been assessed empirically?

    The focus is primarily on income poverty, and the economic actions and policies
governments are adopting to reduce it. Other dimensions of poverty (illiteracy, ill-health) are
dealt with in other reviews, though reference is made to some of them. We find that the
information base is relevant and up-to-date in most African PRSP countries. But there are
significant data problems, which were not always recognized in the PRSP documentation. We
were surprised not to find greater commitment to serious participatory poverty assessment. Most
of the strategies set forth in the PRSPs could not be traced to robust empirical analysis, though
there were exceptions. In general, we find that PRSPs tended to neglect the link between growth
and poverty, by failing to identify the transmission mechanisms linking groups of poor people to

    These include Chad, Ethiopia, Kenya, Malawi, Rwanda, and Senegal.

the benefits of growth. The relatively weak (and sometimes hasty) poverty analysis in the PRSPs
has meant that the policy prescriptions comprising the poverty reduction strategy have typically
not emerged from an understanding of the empirical determinants of poverty.

    II.        The information base

    The information needed for an effective poverty reduction strategy cannot be defined in any
universal sense. In part, it depends on the nature of poverty. If there are only confined pockets of
poverty, the information called for would be mostly limited to the living conditions of those
groups affected. On the other hand, if poverty is extensive, affecting large geographic areas and
several social and economic groups, the information demands can be humungous. Whichever
best describes poverty in a particular country, the information base must include data at the
household (if not individual) level, covering all the main dimensions of deprivation that
characterize poverty in the particular setting.

    Most African countries can be characterized as facing mass poverty, with very high incidence
of income poverty, spread widely across the countries, and with the majority of the populations
facing hardships such as ill-health, illiteracy, and insecure livelihoods.          Under such
circumstances, a nationally representative household survey would appear to be a minimum data
requirement to make any useful assessment of poverty. All of the African countries which had
completed PRSPs at the time of this review could draw on nationally representative household
surveys from which income poverty estimates were derived (Table 1). But two countries (Niger
and Tanzania) could only draw on data from the early 1990s, which clearly have limited
relevance to current policy issues. For income poverty the ‘relevance gap’ between the
formulation of the poverty reduction strategy and the latest survey varied from two (Burkina
Faso) to nine (Niger) years. The situation was similar for other types of household survey,
including those designed to measure progress in human development. For example, the latest
Demographic and Health Survey (or similar) drawn on in the PRSPs was as recent as 1999 in
Burkina Faso and 2000 in Niger. Uganda and Tanzania, however, had DHS data only for 1995
and 1996 (respectively).

    Table 1: Household survey data availability for African PRSPs
Country            Date of PRSP       Latest available income-   Latest available human
                                      expenditure household      development/welfare household
                                      survey                     survey

Burkina Faso       June 2000          1998 PS                    1999 DHS
Mauritania         February 2001      1996 EPCV survey           1998 DHS
Mozambique         August 2001        1996/97 HCS                1997 DHS
Niger              January 2002       1993 ENBC                  2000 MICS; 1998 DHS
Tanzania           November 2000      1991/92 HBS                1996 DHS
Uganda             March 2000         1997 WMS                   1995 DHS
Zambia              March 2002        1998 LSMS                  1997 DHS

    Four further questions arise in the context of the poverty information base for the PRSPs.
The first concerns whether data allowed tracking of recent poverty trends. Only Niger and
Tanzania failed to incorporate some poverty trend analysis in the PRSP. Uganda was the most
advanced here, with six surveys tracking living conditions over the period 1989-1997 (at the time
the PRSP was prepared). Burkina Faso, Mauritania and Zambia drew on surveys in (respectively)

1994/1998, 1990/1996 and 1991/1998). There is some suggestion (discussed below) that these
surveys (in Burkina Faso, Mauritania and Zambia) do not yield robust estimates of poverty time
trends. The Tanzania PRSP reported only projected poverty rates to 2000 (based presumably on
the 1991/92 survey and national accounts estimates of private consumption growth—though the
procedure was not made explicit in the paper).

     The second question concerns how responsive governments have been to update the poverty
data—and the related question of whether the PRSP process evoked this response. In two
countries (Tanzania and Uganda) the timing of the PRSP preparation was unfortunate, since these
countries were implementing large living standards surveys at the time it was being prepared.
Uganda was able (as a result of a concerted data cleaning and analysis effort) to draw on the 2000
Uganda National Household Survey (UNHS) in its PRSP Progress Report (March 2001), but not
in the 2000 PRSP. Tanzania’s PRSP Progress Report was able to draw from the 1999 DHS and
an incomplete sample of the 2000 Household Budget Survey. It could not, however, estimate
income poverty from the 2000 survey—the consumption aggregates could not be computed
meaningfully for the incomplete sample that was available at the time. The Mauritania PRSP
refers to an ‘ongoing’ household living standards survey which will inform future rounds of the
PRSP process. This issue of governments’ responses to the need for data is also discussed in the
post-script on I-PRSPs.

    The third issue concerns the integration of income poverty data (typically from household
surveys) and the wider set of economic information—particularly the system of national
accounts. There is no evidence in the PRSPs reviewed here that there has been a systematic
attempt to ensure some broad consistency between the information on economic performance
(generally from the national accounts) and that on household economic activity and well-being
(from the surveys). This is particularly important for these countries, where accelerated
economic growth is the cornerstone of their poverty reduction strategies. PRSPs must inevitably
link economic growth to household livelihoods and well-being. This is discussed further in

    Finally, what of the timeliness of participatory poverty data for the PRSP (in addition to the
specific participatory processes which were set up to prepare the strategy)? The Uganda
Participatory Poverty Assessment Project was fielded in 1999, and provided potentially relevant
data for the PRSP. However, its results contradicted the quantitative (survey) data. In the
ongoing review of the strategy, the current round of the UPPAP is to be conducted in the
sampling clusters that were the basis of the UHS sample, so that direct comparisons can be made
between participatory and structured survey findings. The participatory poverty assessment
which the Tanzanian PRSP drew from was somewhat more dated, being conducted in 1995. No
participatory surveys were in evidence in the PRSPs for Burkina Faso and Mauritania.

    In sum, the quantitative information base that these African countries could draw on was
relevant and (apart from two cases) reasonably up-to-date. And the situation is rapidly
improving. For future PRSPs and PRSP Progress Reports, these countries will be well placed
with up-to-date, relevant household survey data on poverty. Progress with participatory data
appears to be more patchy. In Uganda the second UPPAP is currently being fielded.
Mozambique will soon have access to the latest participatory assessments from 2001. And in
Mauritania there is mention in the PRSP of a participatory monitoring strategy (though no details
are given). In both Burkina Faso and Tanzania, very little reference (if any) is given to
strengthening participatory monitoring processes.

    III.     The knowledge base

    An even greater challenge for the PRSP task is to translate the available data into knowledge
about poverty—its various dimensions and underlying causes. Without such knowledge the
poverty reduction strategy has to rely on casual empiricism or just guesswork. What evidence is
there in the PRSP documentation that the strategies prescribed were derived from such a
convincing knowledge base?

    Poverty measurement

    A useful starting point is an assessment of the robustness of the poverty measure derived
from survey data and used in the PRSPs. This is important for two reasons. First, the
measurement of poverty is an essential step in gaining an understanding of its causes, dynamics,
and possible cures. Second, poverty measures are needed to establish a base line in the PRSP,
which can then be used to monitor progress accurately over time. Measuring income poverty
involves three steps: calibrating a welfare measure (such as real household consumption per
capita); identifying a poverty benchmark on the welfare measure, below which people are to
considered as poor; and computing an aggregate poverty measure or index, such as the poverty
headcount. Each is examined in turn.

     All PRSP poverty estimates are based on real household consumption. In most cases,
consumption is normalized on per adult equivalence. To what extent are estimates of real
consumption robust, and consistent over time? The Uganda PRSP incorporates the most
thorough empirical assessment of income poverty and its recent dynamics. Careful comparisons
of all six surveys available to the PRSP confirmed a sharp growth in real consumption and
downward trend in poverty. As surveys were all nationally representative and their designs
basically comparable the PRSP was confident that the poverty estimates which were used were a
reliable basis for policy prescription. The consumption aggregates on which they were based
took account of cost-of-living variations both across regions and over time.

     Three countries reported trends in poverty which were less certain or robust. Burkina Faso
used the 1994 and 1998 priority surveys, which though nationally representative, were not
conducted through the year—and in fact covered different phases of the agricultural cycle. It is
not clear whether the adjustments to the data to deal with this were well founded empirically.
The consumption aggregates also appear to have been defined in current prices, with an
adjustment made to the absolute poverty line to assess changes over time. This suggests that
regional price variations (including rural-urban differences) were not taken into account in the
welfare (consumption) measure. In Mauritania, the PRSP refers to a fall in measured inequality
(the Gini ratio falling from 50 percent to 38 percent in just 6 years). A fall in the Gini of this
magnitude is questionable, and is more likely to be a result of measurement error in one or both
of the surveys. Yet neither the PRSP nor the JSA express doubts about the underlying data
analysis.2 And the time trends for poverty in Zambia are also open to question, given problems
with data comparability not mentioned in the PRSP (or JSA). The 1993 survey was fielded at the
peak of the hungry season (April to June), while all the other surveys were fielded during
October-December. The 1991 survey also has weaknesses which have led analysts to conclude
that only the 1996 and 1998 are strictly comparable. The PRSP conclusion that rural poverty has
fallen and urban poverty has increased is therefore open to question, since it only comes from a

  Analysis of the same data set by IDS in Sussex UK gives a much smaller change in the Gini, which
suggests that the problem lies in the analysis underpinning the PRSP.

comparison with the 1991/1993 surveys. Taking just the 1996-1998 evidence, rural poverty is
unchanged (and urban poverty has increased).

    An issue, which seems to be common across most of the African PRSPs reviewed here, is the
absence of any deep discussion of the poverty line. The Burkina Faso, Mozambique and Zambia
PRSPs provide the most information on the poverty lines used. The Mozambique poverty line is
constructed as the sum of a food poverty line based on a daily nutritional standard plus a modest
amount of non-food expenditures. We are also told that the line is fixed after ‘weighting the
provincial poverty lines and adjusting them to reflect variations in the cost of living’ (P. 13).
‘Abject’ poverty or destitution describes those whose consumption was less than 60 percent of the
poverty line.3 The Burkina Faso poverty line is also anchored around a nutritionally based
concept of poverty. However, it is not stated clearly how the line was re-drawn for the later
period to account for price changes (i.e. what inflation procedure was used). The Zambia PRSP
contains an interesting discussion on the meaning of poverty in the Zambian context. The
Mauritania PRSP uses the dollar-a-day measure, without further justification, and without
providing details of whether conversion to the local currency equivalent was on the basis of
purchasing power parity. An ‘extreme poverty threshold’ was also used (being lower than the
$1/day line), but no information is provided. Niger adopts a relative poverty line, generated from
the 1993 income distribution. Both the Uganda and Tanzanian PRSPs do not report details of the
poverty lines used in the PRSP.

    None of the PRSPs reviewed entered into a discussion of consensus building around the
poverty line. For an absolute poverty line to have any real meaning, it must have general
acceptance in the society. While the Tanzanian PRSP refers to ‘the absence of official poverty
lines’ (p. 6), it does not make any proposal to generate such lines. What all countries need for
future poverty analysis and monitoring is a process through which the society at large
(government, research institutions, non-government organizations, civil society organizations)
comes to a consensus on what it means to be poor in the specific country context. Only then can
absolute poverty measures and goals have genuine meaning.

    Finally, most of the PRSPs relied heavily on headcount poverty indices. Only Burkina Faso
reported and used other indices (such as the depth and severity indices), though they were
mentioned (but not analyzed) in the Niger, Mozambique and Mauritania PRSPs. Surprisingly, the
Uganda PRSP only mentions headcount poverty, even though severity indices fell sharply,
indicating that the poorest of the poor appeared to benefit from recent growth.

    At the very least, poverty reduction strategies must be based on a robust poverty profile, in
which the location and main activities of the poor are identified. Because they are sample-based,
most household surveys are limited in the degree of detail or disaggregation that they can sustain.
Yet, for practical purposes, governments need quite specific information about poor people and
their livelihoods if poverty reduction strategies are to be at all relevant and effective. How
successful are the PRSPs in identifying specific poverty groups as the basis for their strategies?

    All the completed PRSPs we reviewed highlighted geography as a fundamental factor behind
persistent poverty, and reported regional disaggregations of poverty. The PRSPs in Burkina Faso,
Mauritania and Tanzania provide poverty maps. However, the degree of disaggregation is limited
by the sample designs in most cases. The Mauritania and Uganda PRSPs were only able to report
poverty rates confidently for four regions, while Burkina Faso could distinguish ten planning

 It can probably be assumed that the Mozambique poverty line is constructed from the IAF survey of
1996/97 (used for the poverty analysis) as it is the only survey mentioned in the PRSP.

regions. The Mozambique PRSP reported poverty rates at the provincial level, and also has a
poverty mapping exercise underway at the district level. Tanzania provided a poverty ranking for
twenty regions, though the data used for this, and their statistical accuracy, were not discussed.

     Geography is only one factor behind the persistence of poverty. Economic activity and social
characteristics (household size and composition, ethnicity, gender) are also important. Most
African PRSPs disaggregated poverty data by socio-economic grouping (the exception was
Tanzania). The discussion was limited in the Mauritanian PRSP, which referred to ‘insufficient
comparability and precision of nomenclatures’ (p.6) as constraining such disaggregations. The
Mozambique PRSP presented the principal sectors of employment (public sector, private sector,
self-employed) for the poor and non-poor populations.4 Partitioning households by economic
activity was perhaps most telling and insightful in the Burkina Faso and Uganda PRSPs. Both
highlighted the fact that cash crop producers benefited significantly from recent economic
developments, but not food producers. Interestingly, the story was revised in Uganda’s PRSP
Progress Report, which used more recent data. These showed gains now emerging even among
food producers. In Niger, the identification of groups (in high and low ‘cost areas’) was not
helpful or easily understood.

    Two issues emerge from the above discussion of disaggregation of poverty data. First, the
need for PRSPs to develop more relevant groupings of households, particularly for the rural
population. The ‘food-crop farmers’ category is too large and heterogeneous to be useful for
policy. What is needed is a taxonomy of this group which is more helpful for policy analysis. A
serious neglect in the African PRSPs is of the role that gender inequality plays in constraining
agricultural growth and rural poverty reduction. Women are typically food farmers in rural
Africa, but have unequal access to productive assets and public services. Yet this issue is hardly
mentioned in the PRSPs. A similar data need would apply to other large groups—female-headed
households and ethnic groups. At the other extreme, some small (and largely homogenous)
groups—such as the handicapped, the infirm, the homeless and street children—can be missed
from the analysis altogether. This is because they are either not in the sampling frame of a
household survey, or because they are simply too few in number to be picked up by a national
random sample. Either way, there is an obvious need for the PRSP process to ensure that
knowledge about such groups is generated in the poverty information strategy.

    Second, the PRSPs we reviewed largely paid little attention to poverty trends for
socioeconomic groups, nor to the transmission mechanisms linking the groups to the growth and
development process. Why food farmers in Uganda now seem to be benefiting from growth is a
central issue for the poverty reduction strategy. What are the transmission mechanisms that have
led to this favorable outcome? Have labor markets tightened, and real wages increased in rural
areas; or has the demand for food increased? The Burkina Faso PRSP referred to the multiplier
effect arising from the fact that rural incomes are spent mainly on local goods—the product
market therefore acting as an effective conduit in spreading the benefits of growth. But apart
from this, there was little coverage of this issue. Yet it is central to the formulation of effective
poverty reduction strategies, and especially those anchored to accelerated economic growth.

    The Burkina Faso document is perhaps the most detailed in its microeconomic diagnosis of
poverty. It explored in some depth the reasons why agricultural productivity and farm incomes
are low (low levels of education, technology, access to land, markets and infrastructure). The
Mozambique PRSP presented information on the characteristics of the poor—demographic
characteristics (large households, high dependency ratios), low levels of education and access to

    However, the it did not show poverty rates by sector of employment.

health services, and land ownership. Multivariate regressions were then used to identify several
determinants (or at least correlates) of poverty.5 Although both the Mauritanian and Uganda
PRSPs specified actions to help expand agricultural incomes, they did not report any supporting
microeconomic empirical analysis. Overall, the African PRSPs reviewed here provided very little
empirical microeconomic analysis to underpin the strategy, particularly for rural/agricultural

    To what extent do the PRSPs draw on participatory approaches to poverty diagnostic
analysis? There is no evidence in the PRSPs of Burkina Faso and Mauritania of any participatory
poverty study. Uganda, on the other hand, had fielded its UPPAP in 1999, just prior to the PRSP.
The issue here is that the UPPAP concluded that poverty was increasing, contradicting the
quantitative evidence of the structured surveys. But apart from this concern, the PRSP appears to
draw very little from the UPPAP.6 Similarly Tanzania conducted a participatory assessment in
1995, but does not draw much from it in the PRSP (there is only a passing reference to women
perceiving themselves as poorer). However, the Tanzanian PRSP process also involved a series
of ‘Zonal Workshops’ which were designed to elicit the views of ‘grassroots stakeholders’.
These consultations were drawn upon in the PRSP. However, there is no assurance that the
results of this exercise are robust from a research standpoint. Both quantitative and participatory
methods of social enquiry must observe rigorous methodological disciplines before their results
can be accepted as a basis for policy. It is unlikely that the Zonal Workshops observed such
discipline. Zambia utilized a recent (2002) qualitative study on poverty in selected urban and
rural communities. This highlighted economic poverty (money, employment, food), but also
deeply important psychological dimensions (family harmony and so on).

    Interestingly, although all studies utilized data on the broader (especially health) dimensions
of household wellbeing (such as the DHS, MICS and CWIQ), none chose to disaggregate the data
by economic status (despite recent advances in the techniques for this). Sahn et al (1999) have
shown the advantage of such disaggregation, highlighting the different trends in outcomes across
the income/wealth quintiles. Yet this was missed in the PRSPs. In a number of countries,
changes in child and infant mortality were quite uneven across the wealth quintiles, with
increases (in Senegal, Zambia and Zimbabwe) occurring most among the richest. Such dynamics
have significant policy messages.

    IV.      Policy impact

     Given the time that most countries had to prepare the PRSPs which are reviewed here, it
would be expecting too much for a systematic empirical (ex ante) analysis of the future impact of
all the actions taken on poverty. Nevertheless it is important to assess whether the PRSPs in our
sample did include some analysis of the connection between policies and public actions on the
one hand, and poverty outcomes on the other. Our review examines two broad aspects of this
issue: Did the PRSPs make any ex post assessment of the impact of past policies on poverty?
Did they utilize knowledge about the underlying determinants of poverty in identifying the
proposed actions to be take to reduce poverty in the future?

  One of the determinants identified is slow economic growth. However, since only one survey (1996/97)
is available, it is unclear how ‘growth’ could have been an independent variable in the regression.
   The UPPAP provided substantial insight into the importance of water in the lives of the poor, and the
PRSP does in fact emphasize water supply in the strategy. But it makes no reference to the UPPAP in this

    Past policies

     All the PRSPs covered in this review have accelerated economic growth at the heart of the
strategy. But few analyzed the relationship between growth, household economic well-being, and
poverty in the historical context. Tanzania ignored it altogether, and in fact provided very few
data on past economic growth. The Mozambique PRSP presented the results of both backward
and forward simulations using the 1996/97 data, while noting that these ‘suffer from well-known
methodological and statistical limitations’ (P. 19). The Burkina Faso PRSP juxtaposed relatively
favorable GDP growth during the 1990s (sparked by the 1994 devaluation) against an unchanging
incidence of consumption poverty.7 But this was not explored or explained. Is the conflict due to
differences in average consumption in the household survey compared with the national accounts,
or to a change in income distribution? An explanation may well lie in the high degree of income
inequality, but this was not sufficiently explored in the PRSP. Both the Mauritania and Uganda
PRSPs included useful discussions of poverty reducing effects of past growth. Recovery and
growth in both countries (evidenced in the national accounts) seem to be supported by the
household survey data, and the observed trends in household consumption and poverty.

    Two PRSPs (Burkina Faso and Uganda) emphasized the fact that past reforms (the CFAF
devaluation in Burkina Faso and the liberalization of coffee marketing in Uganda) have raised
incomes and reduced poverty among cash-crop producers. But both PRSPs did not go into any
depth, neither did they refer to background empirical analysis. The Mauritania PRSP admitted to
the ‘weakness of the capacity to manage economic and sectoral policies’ (p. 12), but it did not
specify which policy interventions might have aggravated poverty. There was very little
discussion in the Tanzanian and Mozambique PRSPs of past policies, and how they have affected
poverty. There was presumably little scope to do this with just one year of household survey data

    Thus, the evidence suggests that most (though not all) the PRSPs we reviewed entered into a
discussion of how past economic reforms have affected growth and poverty. But few offered
convincing empirical evidence (from quantitative analysis) of the links between past policies and
poverty outcomes. Perhaps the most striking feature of these PRSPs is their general neglect of
the transmission mechanisms between growth (and pro-growth policies) on the one hand, and
household economic welfare and poverty on the other. This is surely a high priority issue for
future PRSP-oriented analytical work in all the countries, and perhaps especially the African
countries where the knowledge base is the weakest.

    Current actions and future outcomes

    Looking forward in time, all the PRSPs assume an acceleration in the pace of growth, and his
is seen as central to poverty reduction. Some countries (notably Tanzania and Uganda) report
simulations to estimate likely poverty reductions from growth scenarios. The Uganda work is
reported in more detail, and is clearly more advanced, showing the conditions under which the
assumed 7 percent growth rate of GDP would achieve the poverty target. Growth targets in
Burkina Faso and Mauritania were overly ambitious, and no attempt was made in the PRSP to
analyze the possible impact on poverty. In fact neither of these PRSPs report income poverty

  A similar point is made in Senegal’s I-PRSP, but is not substantiated with evidence. The growth
recovery occurred during the post devaluation period (after 1995) while the poverty increase was only
reported for the 1991-1995 period.

    A serious shortcoming of the PRSPs is the failure to analyze the growth process itself, and to
assess the sources of long term growth empirically. The exception is Uganda, which draws on a
study by the World Bank assessing the constraints to accelerated growth. It states explicitly that
the growth target is unlikely to be met without public action in enhancing educational attainment,
deepening financial intermediation, and enforcing rural of law. This is one of the most
challenging tasks of the PRSP, dealing as it does with the longer term. And it is not surprising to
find little attention to it in most PRSPs. The tendency to assume a growth path (usually implying
a marked acceleration in economic progress) is not appropriate in the PRSP context.

    All the PRSPs reviewed here failed to present what might be termed a coherent framework in
analyzing how public actions would lead to the stated targeted poverty outcomes. Ideally, the
actions that the PRSP prescribes should emerge from an empirical understanding of the main
determinants of the poverty problem. This was generally lacking. Even when there is a link
between PRSP actions and the poverty analysis, it does not appear that there is rigorous empirical
understanding of how the actions will bring about outcomes.

     Some examples might serve to illustrate this limitation. The Mauritania PRSP listed a series
of actions to enhance the growth prospects of sectors, such as mining, iron, industrial fishing,
tourism, road transport, marine transport, and so on. And yet there was no reference to how the
expansion of these activities would directly or indirectly raise the incomes of the poor. Even
actions to raise food production (provision of technology packages, pest control, extension
services, and improved marketing services) are prescribed without reference to any underlying
empirical analysis of food-producer output and incomes. Similar considerations apply to the
Zambia PRSP, which highlights the development of mining and tourism, but nowhere discusses
how this might (either directly or indirectly) benefit the incomes of the poor. The Tanzanian
PRSP listed a dozen or so actions to improve agricultural incomes, but their empirical basis was
not made clear. Burkina Faso makes recommendation for several structural reforms in its
strategy (improving competitiveness, reducing transport margins, reduce minimum wages,
disengage government from the marketplace), but there is no reference to any analysis of the
likely impact on different poor groups. The several actions prescribed to enhance rural incomes
reveal very little empirical analysis of their likely effects, and their impact specifically on the
lives of the poor. The six actions proposed in the Uganda PRSP under the Plan for the
Modernization of Agriculture were stated without reference to any empirical analysis.

    None of the PRSPs reviewed includes a specific ex ante assessment of the poverty and social
impact of macroeconomic and structural policy reforms being proposed in the PRSP. The
reasons for this are not clear. The time limit in the preparation of the PRSPs might have been the
binding constraint. It might alternatively be due to limited analytical capacity in the countries.

    One of the key instruments in the fight against poverty is public spending, and subsidized
services—especially health, education, water and infrastructure. Yet none of the PRSPs
contained any assessment of whether these important services were being used by the poor. One
useful shorthand would be estimates of the benefit incidence of public spending (in say health,
education and water). Yet not one of the seven full PRSPs reviewed here reported such analysis.

    V. Post script on some IPRSPs

    The ‘selection bias’ in our review of completed PRSPs counsels caution in applying the
findings to the rest of the region. To check on this, we reviewed also a selection of (six) I-PRSPs

approved in 2000.8 Most of these did have recent data on economic living standards to hand.
PRSPs in Ethiopia, Kenya, and Malawi had household survey information for the late 1990s
(1997-1998). But Rwanda and Chad had almost no nationally representative data to work with.
Senegal only had a 1995 survey to go on. Most of the I-PRSPs, however, refer to plans to
implement surveys in the near future. Rwanda is perhaps the most striking (along with Tanzania)
of the data situation improving markedly in the early phase of the PRSP process. New data are
programmed also for Chad, Ethiopia and Malawi. The Chad experience is an interesting one.
Because of the absence of any nationally representative survey for the I-PRSP, the authorities
decided to implement a large and complex household survey. On the advice of Bank staff, they
have reviewed this decision, and will instead, carry out a much lighter survey, more in line with
national capabilities. The Ethiopian Household Income and Consumption Expenditure Survey
was in the field at the time that the I-PRSP was being prepared.

    Although one cannot expect in depth analysis in the I-PRSPs, the poverty diagnosis in the I-
PRSPs is disappointing. In some countries (Chad and Malawi for example) the analytical content
of the I-PRSPs is very thin. In some (Senegal) it is confusing. In others (notable Rwanda and
Kenya), it is insightful and helpful from a policy perspective. As with the full PRSPs, the I-PRSP
documentation we have examined shows no systematic policy analysis or coherent analytical
framework. But preliminary indications are that the draft PRSPs emerging in former data-poor
countries have benefited enormously from the improved data situation, and have a much deeper
analytical basis than might have been predicted given the I-PRSP. The draft PRSPs for Malawi
and Rwanda contains a much-improved poverty diagnostic (compared with the I-PRSP).

      V.      Concluding observations

    This review has focused on the seven completed full African PRSPs in FY00, FY01 and
FY02 (as well as a more cursory look at six I-PRSPs). It has relied almost entirely on the PRSP
documentation, and has not researched the empirical work undertaken as background to the
documentation. Nonetheless, a number of important observations can be made about the poverty
analysis underpinning the poverty reduction strategies in these countries.

    First, the poverty information base was reasonably up-to-date in most in most African PRSPs.
The key issue is the timeliness of the data, and the priority countries are now giving to ensure that
data gaps do not persist. Uganda in many respects represents good practice for Africa. The
commitment it showed in completing, processing and analyzing the 2000 Uganda National
Household Survey in time for the PRSP Progress Report in 2001 shows what can be achieved,
even in the context of a typical low-income African country. Most countries covered in this
review appear to be committed to a poverty information strategy, featuring repeated future
household surveys. Some countries have made these intentions explicit in the PRSP
documentation, whereas others offer only vague statements of intent. A similar picture emerges
from the review of participatory/qualitative information. The PRSP in Uganda was exceptionally
well informed by recent participatory poverty studies, though the results were underplayed in the
final Ugandan document. There is some tension between the findings of participatory
assessments and quantitative surveys, the former being more pessimistic about poverty trends.
The current work in Uganda to link qualitative and quantitative studies represents a major
development. For the other countries it was difficult to discern concrete plans for future
participatory/qualitative assessment.

    The I-PRSPs for Chad, Ethiopia, Kenya, Malawi, Rwanda and Senegal were reviewed.

    An essential feature of the poverty knowledge base is the accurate measurement of income
(or consumption) poverty. This is needed for both poverty diagnostic work, and for future
monitoring of progress. There is some evidence of careless poverty measurement in these PRSPs.
And it was not clear that even the most basic requirements of an economic welfare measure (such
as controlling for differences in prices or household needs) were met in the underlying data work.
PRSPs need to be more transparent in their description of poverty measures and its monitoring.
None of the PRSPs referred to a consensus among civil society as to what it means to be poor in
the context of the country concerned.

     All the PRSPs included discussion of the microeconomic determinants of poverty. Most
relied on descriptive approaches (often appearing to draw from casual empiricism or ‘common
knowledge’).9 Most of the PRSPs neglected gender inequality as a determinant of persistent
poverty. This would seem to be a particularly important omission in African PRSPs, given their
emphasis on agricultural growth. All referred to the geographical constraints leading to persistent
poverty, though it was not always clear whether this diagnosis was reflected in the actions
proposed in the strategy. Regional issues are seen as a major cause of poverty, but the poverty
reduction strategies did not appear to incorporate any fundamental regional perspective.
Although most PRSPs also were able to present poverty data for different social groups, their
capacity to do so was often severely restricted by the data. Limited sample size meant that
poverty estimates and trends could not be sufficiently specific for most practical policy purposes.
But perhaps the more important issue emerging from our review is the failure of PRSPs to forge
convincing empirical links between the livelihoods of these poverty groups and the overall
process of growth and development. There was simply not enough analysis of the transmission
mechanisms linking different groups of poor people through markets and services to the benefits
of growth. The links between economic growth and the livelihoods of the poor were typically not
articulated in the PRSP documentation.

     Finally, it should come as no surprise that there was little analysis of the impact of the policy
actions on the lives of the poor. It would be fair to conclude that most PRSPs failed to properly
apply the diagnostic analysis of poverty to the formulation of the poverty reduction strategy. The
neglect of regional issues is one example. Many PRSPs listed policy actions which had no clear
connection to the poverty diagnosis; when there was a connection, it was not apparent that
analytical work had been undertaken to connect actions with outcomes. Some had only a limited
perspective of the historical record—of the evolution of poverty in the recent past (though
Uganda is an exception to this). This constrained their ability to assess how past policies and
strategies affected poverty. Another major limitation was the general neglect of the incidence of
public spending—a central instrument in the fight against poverty. The techniques are well
understood and readily applied to survey data, yet none of the PRSPs we covered reported the
results of benefit incidence work. None of these PRSPs incorporated any structured attempt at an
ex ante assessment of the poverty and social impact of the policies and actions prescribed. In our
view, both the ex post assessment of past policies drawing on history, and the ex ante poverty and
social impact analysis of policy interventions need to be considerably strengthened in future
rounds of the PRSP.

    Reference: Sahn, D., D. Stifel and S. Younger (1999) ‘Inter-temporal Changes in Welfare:
Preliminary Results from Ten African Countries.’ Cornell Food and Nutrition Policy Program,
Working Paper No. 94, Cornell University: Ithaca, NY

 As noted elsewhere in this review, it was also simply not possible at times to glean whether rigorous
analytical work had been undertaken.

                                       Annex Table 1: Assessment of the poverty analysis in approved Africa PRSPs

                    Adequacy of poverty data                          How well have the nature and determinants of poverty outcomes           To what extent have the growth and
                                                                      (income and non-monetary dimensions) been identified? Have              distributional impact of past policies and
                                                                      trends in key poverty outcomes and determinants been presented?         programs been assessed?

Tanzania       PRSP hampered by no recent household              No attempt at a quantitative analysis of the determinants of poverty.        Little discussion (and no analysis) of past
(PRSP          survey (last quantitative survey in 1991/2;       No trends in poverty available for the PRSP. Some new data emerge in         policies and poverty. Some reference to
Nov. 2000)     and participatory surveys in 1995 and 1997).      the PRSP Progress Report, but the data are sketchy. Conclusion drawn         production and income gains from the
               PRSP baseline taken to be the 1991/2 survey       is that there has been little improvement in either income poverty or        reform-led adoption of market-oriented
               estimate—an obvious limitation. Household         human development indicators during the 1990s. Some use of Zonal             approaches, but there is reported
               budget survey and DHS are currently being         workshops to build a picture of main poverty determinants. Also some         dissatisfaction expressed at Zonal
               conducted. The PRSP utilized the 1996 DHS.        attempt to generate regional welfare ranking, identifying most deprived      workshops (market orientation means less
               The PRSP Progress Report rejects poverty          regions by welfare criterion. No benefit incidence analysis.                 credit and extension service). No social and
               estimates from 2000 HBS(based on 3 months                                                                                      poverty impact analysis. Gender issues not
               of data) because of small sample. Some                                                                                         addressed (e.g. no targets for school
               attempt to relate views at Zonal workshops to                                                                                  enrollment of girls).
               measured poverty outcomes, but time
               disconnect a problem. Discussion of HD
               outcomes also based mainly on evidence in
               the first half of the 1990s.
Burkina Faso   Two nationally representative ‘Priority           The analysis of the determinants of poverty is well done. Good               Although real GDP grew by 5% pa between
(PRSP          Surveys’ in 1994 and 1998 were the basis of       analysis of low productivity among crop producers (especially food),         1994 and 1999, the surveys show no change
June 2000)     the poverty assessment. The poverty line is       and their vulnerability to price fluctuations. Sound analysis of human       in rural poverty and an increase in urban
               nutritionally based, and is re-calibrated to      development indicators, with DHS data on outcomes and service use            poverty (between 1994 and 1998). The
               1998 prices (though the details of this are not   disaggregated by wealth quintile. There is a separate section on women       failure of growth to deliver poverty
               given). This re-calibration gives an increase     and poverty, but this needed integrating with the main diagnosis (e.g.       reduction is a key question that is not
               in the urban headcount. Good poverty maps         no reference to gender in the discussion on water). There is some            properly addressed in the PRSP. Did mean
               are provided, and poverty data are reported by    discussion of land and credit, but no direct empirical links to poverty or   household consumption fail to grow, or was
               socioeconomic group. Poverty fell among           household outcomes are made. No benefit incidence analysis.                  there a sharp deterioration in the
               cash-croppers after CFAF devaluation, but has                                                                                  distribution?
               risen among food producers. Income
               inequality appears to be a big problem (richest                                                                                There is some reference to the CFAF
               20% get >60% of income), but not much                                                                                          devaluation and the fall in poverty among
               discussed.                                                                                                                     export crop producers, but otherwise no real
                                                                                                                                              analysis of past macro/structural policies.
               The PRSP reports the results of a participatory                                                                                Lots of ‘structural’ policies proposed
               assessment-—how the poor view themselves.                                                                                      (improving competitiveness, reducing
               Not linked to quantitative data..The PRSP                                                                                      transport costs, reduced minimum wage,
               draws from the 1999 DHS.                                                                                                       disengagement from marketing, etc), but no
                                                                                                                                              analysis on social or poverty impact.

                                      Annex Table 1: Assessment of the poverty analysis in approved Africa PRSPs

                   Adequacy of poverty data                           How well have the nature and determinants of poverty outcomes            To what extent have the growth and
                                                                      (income and non-monetary dimensions) been identified? Have               distributional impact of past policies and
                                                                      trends in key poverty outcomes and determinants been presented?          programs been assessed?

Mauritania    Two nationally representative household            Poverty trends are presented, but the analysis is superficial. First, there   There was no attempt to link the survey
(PRSP         surveys (EPCV) in 1990 and 1996 were the           are obviously problems with the underlying data which are ignored             evidence on poverty to the national
Feb. 2000)    basis of the poverty assessment. $1/day            (and not mentioned in the JSA). Second, there is no interest in defining      accounts growth story. The PRSP fails to
              poverty line and an undefined extreme poverty      poverty in the Mauritania context. The PRSP rightly points to low             explain why past growth reduced income
              line were used. The data show sharp fall in        levels of agricultural productivity, precarious geophysical environment,      inequality (if indeed it did). References to
              poverty rates, but there are doubts about this.    small land holdings, lack of infrastructure, as major problems, but the       past macro and structural reforms are only
              Using the same data the Gini is estimated to       direct links to poverty (linking these factors to household income/           cursory. Again no analysis of exclusion—
              have fallen from 0.50 to 0.38 in just 6 years      consumption) are not made. Overall the PRSP makes a reasonable                or gender inequality.
              which is very unlikely. Whereas the surveys        assessment of the deep-seated roots of poverty, but it does not report
              report regional poverty rates, the PRSP states     any rigorous empirical diagnosis relating these factors to household-
              that coverage of both surveys was                  level outcomes. There is little gender disaggregation in the diagnosis—
              incomplete—a weakness corrected in the             apart from school enrollments. Even the discussion of access to
              current survey (started in July 2000). The data    potable water made no reference to gender inequality. No discussion of
              permitted limited analysis of poverty across       whether the poor use publicly subsidized services.
              socioeconomic groups. The PRSP used the
              same surveys to assess education levels, but
              the sources for the health data re not reported.
              In sum, there are data, but the results suggest
              real problems with them. Possibly the
              2000/01 survey should be the base line for the

Uganda        Consistent sequence of household surveys an        Sharp fall in poverty between 1992 and 1997 discussed, and attributed         Some discussion of positive effect of
(PRSP         excellent data set for poverty diagnosis. PRSP     to growth recovery, agricultural market liberalization, and favorable         agricultural market liberalization on poverty
March 2000)   had household survey data to 1997. PRSP            coffee prices. PRSP Progress Report confirms the poverty decline,             among export crop producers. Significant
              Progress Report drew on 2000 Uganda                based on updated trend analysis (1997-2000). No multivariate analysis         discussion of growth path and poverty
              National Household Survey. All data                of poverty. Poverty diagnostic is based mainly on poverty projections,        reduction, including use of international
              disaggregated by four main regions and six         taking the 1997 survey as base. Experiments show that the poverty             cross sections. Limited discussion of
              socio-economic groups, and generally               target is attainable, but requires higher agricultural growth. Trends in      gender inequalities, and no ex ante social
              considered of high quality. Discussion of          child mortality disaggregated by wealth quintile. Cross section analysis      and poverty impact analysis.
              Uganda Participatory Poverty Assessment            of child mortality shows that halving the rate by 2017 is feasible, but
              Program apparently contradicting survey data.      will call for energetic effort. Reference to research on land and
              Resolved by reference to sample selectivity        women’s rights, but no detail provided. Gender inequalities largely
              and different (i.e. non economic) welfare          neglected. No reference to gender in discussion of water (coverage of
              measures.                                          water weak compared with UPPAP findings. No benefit incidence

                                      Annex Table 1: Assessment of the poverty analysis in approved Africa PRSPs

                   Adequacy of poverty data                        How well have the nature and determinants of poverty outcomes           To what extent have the growth and
                                                                   (income and non-monetary dimensions) been identified? Have              distributional impact of past policies and
                                                                   trends in key poverty outcomes and determinants been presented?         programs been assessed?

Mozambique     PRSP uses a household survey (IAF)             Headcount ratios, poverty gap, and square of the poverty gap                 Assessment of the impact of past policies
(PRSP          conducted in 1996/97; a population census      disaggregated by rural/urban, region, and province. ‘Human poverty’          on poverty reduction was not attempted as
August 2001)   (1997), and a DHS (1997). A new household      index also presented by rural/urban and by province. While time series       there is just one year of data; there is some
               consumption survey is underway.                data are available for certain social indicators, trends in poverty          description of past macro performance.
               Poverty indices are disaggregated by           incidence are not presented due to a lack of data. Poverty profile           Due to lack of data, the results of some
               rural/urban, region, and province. A poverty   provides information on the characteristics of the poor and non-poor,        fairly crude simulations of the impact of
               mapping exercise is underway which will help   disaggregated by rural/urban. The characteristics examined are:              growth on poverty rates (using the 1996/97
               to target district-level pockets of poverty.   demographics (household size, dependency rate, age of first child);          IAF data) are shown.
               Gender disaggregation is not a priority.       education (head, male and female children); health and nutrition             The PRSP also notes that research is
                                                              (utilization of services, malnutrition); agriculture (land ownership, use    underway to analyze the incidence of health
               A Participatory Poverty Appraisal (APP) was    of inputs); type of employment; and access to services. (Agricultural
               conducted in 1995-96. In 2001, a series of                                                                                  and education benefits. Such analysis of the
                                                              data are poor; the results of the Agro-Animal Husbandry Survey is            distributional impacts of recent growth and
               participatory rural assessments was carried    awaited).
               out. Data appears to be accessible outside                                                                                  sectoral policies will help to inform PRSP
               government.                                    The gender dimension of poverty is somewhat glossed over; the profile        updates.
                                                              mentions that female-headed households are not disproportionately
                                                              poor. On this point, the JSA notes that there is little in the PRSP that
                                                              addresses gender inequality, which is a research topic that donors have
                                                              studied. Issues of vulnerability are also not given a high priority in the
                                                              poverty diagnostics – there is no analysis of the impact of natural
                                                              disasters, HIV, etc.
                                                              The PRSP indicates that regressions were used to identify the following
                                                              determinants of poverty: slow economic growth; poor levels of
                                                              education; high dependency rates; low agricultural productivity; lack of
                                                              employment opportunities within and outside agriculture; and poor
                                                              infrastructure, particularly in rural areas. However, it should be noted
                                                              that there was only one cross-section of household data for this

Zambia         Data are good, with household surveys in       Useful, but possibly flawed discussion of poverty trends, including the      Good discussion on failure of past reforms,

                                       Annex Table 1: Assessment of the poverty analysis in approved Africa PRSPs

                    Adequacy of poverty data                         How well have the nature and determinants of poverty outcomes           To what extent have the growth and
                                                                     (income and non-monetary dimensions) been identified? Have              distributional impact of past policies and
                                                                     trends in key poverty outcomes and determinants been presented?         programs been assessed?

(PRSP           Priority Surveys (1991, 1993) and Living        meaning of the poverty line. Comprehensive coverage of non-income            and on the importance of current
March 2002)     Standards Monitoring surveys (1996, 1998).      dimensions of poverty (especially health, education, access to               privatization policies. Specifies need for
                Household budget survey not used in I-PRSP.     infrastructure). Problem of orphans highlighted. Beneficiary                 structural reforms in agriculture and
                DHS data for 1992 and 1997.                     assessment highlighted these new dimensions of poverty. There are            education as important for poverty
                                                                problems with over time analysis, given lack of data comparability.          reduction. Target to reduce poverty to 50%
                                                                JSA does not pick this up. Limited data on socioeconomic groups, and         by 2004 looks arbitrary and unattainable –
                PPA in 1994. followed by agricultural project   geographic disaggregations not very helpful. PRSP attempts to                certainly not derived from serious analysis.
                beneficiary assessment. Also use of a 2002      demonstrate link between poverty profile and strategy, but this is not       Sections on past macro and structural
                case study of rural and urban communities,      very convincing. For example both tourism and mining are featured in         reforms make no mention of poverty.
                with rural respondents highlighting farming     the strategy, but there is no discussion of how this affects the poor.
                and food, and urban respondents emphasizing     (Note JSA disagrees with this, stating that the links are articulated.) No
                employment/money and food.                      benefit incidence analysis, nor any reference to distribution/inequality
                                                                aspects of human development outcomes and use of public services.
                                                                Interestingly deterioration in child mortality indicators is worse for top
                                                                quintile. This was not mentioned. JSA calls for more analytic work
                                                                with the data.

Níger           Many surveys are listed, including early        Analysis severely restricted by weak data. Relative poverty line based       Section on macroeconomic history makes
(PRSP           budget surveys and special surveys on           on 1993 distribution. No quantitative evidence on income/poverty             little mention of reforms (eg 1994
January 2002)   household security fielded by CARE              trends, and little disaggregation (say by region, socioeconomic group).      devaluation).
                International. But the main nationally          Use of human development data from DHS and MICS, but to attempt
                representative surveys of use for the PRSP      to highlight how outcomes and use of services are distributed across
                were the 1993 National Budget Consumption       population. In sum very little diagnostic analysis. No benefit
                Survey; the DHS (1992 and 1998) and the         incidence.
                MICS (1996-2000). But PRSP refers to the
                fact that most data are outdated and lack
                Qualitative survey conducted in June 2001
                was used as the main diagnostic tool.

                                       Annex table 2: Assessment of the poverty analysis in approved Africa I-PRSPs

                    Adequacy of poverty data                          How well have the nature and determinants of poverty outcomes           To what extent have the growth and
                                                                      (income and non-monetary dimensions) been identified? Have              distributional impact of past policies and
                                                                      trends in key poverty outcomes and determinants been presented?         programs been assessed?
Senegal        Priority Surveys (household) completed in         Analysis of income poverty confused. Argues that poverty has risen           Analysis of growth and poverty flawed. No
(I-PRSP        1991 and 1995. Not a good basis for PRSP in       despite growth, but the survey (poverty) evidence is for 1991-95, while      analysis of impact of devaluation (lack of
May 2000)      2002, and this is underscored in JSA.             the growth acceleration is 1996-99 (post CFA devaluation). Apparent          data). Statement that growth did not reduce
                                                                 increase in poverty (from 33% in 1991 to 65% in 1995) looks very             poverty must be questioned (but is ignored
                                                                 strange. One uses per capita and the other per adult equivalent              in JSA).
                                                                 consumption; poverty line is much higher in second year (CFAF110 
                                                                 392), but no reference to deflator. Little if any reference to non-income
Malawi         Latest survey is 1997/98 Integrated Household     Very little analysis presented—all promised in full PRSP. JSA refers         Quite a bit of discussion of past reforms in
(IPRSP         Survey. Although not entirely clear from the      to significant body of analytical work, but there is not evidence of it in   Malawi, but no mention of potential effects
August 2000)   I-PRSP and JSA, there may be another round        I-PRSP.                                                                      on the poor. Note that poverty outcomes
               of the IHS in the field to be used for the full                                                                                likely to be dominated by shocks (drought).
Chad           Very few data available. Informal sector          Survey not representative, but used as if it were. Analysis of the census    No mention of past macro and structural
(I-PRSP        survey of 1996/96 is the latest, but not          data very dated (1992-93). Data and analytical weaknesses of I-PRSP          policies, only current macro framework and
May 2000)      representative (covering only 4 out of 14         clearly stated in JSA.                                                       policies. No discussion of likely poverty
               prefectures).                                                                                                                  impact.
Rwanda         I-PRSP has no data to draw on – mostly dating     Poverty trends based on 1985 distribution – very questionable. 1985          Excellent discussion of reforms and the
(I-PRSP        back to 1985, and PPA of 1997. Draft PRSP         and 2001 household surveys not comparable. Analysis of 1997 PPA              growth record. Other than the use of the
Nov. 2000)     utilizes improved data: CWIQ 2001, DHS,           which highlighted emergence of new ‘very poor’ (as result of                 projections model, little attempt to link
(Draf t PRSP   2000, MICS 2001, and Household Living             genocide). I-PRSP contains exceptionally insightful analysis of              macro policies to poverty.
March 2002)    Conditions Survey (EICV), 2002. Almost too        poverty problem, bearing in mind the very weak data base. Regional
               many surveys.                                     differences in welfare are featured, but some faulty data work.
Kenya                                                            2000 I-PRSP refers to ‘preliminary results of 1997 WMS’ which is
               Main poverty data sources the DHS and the                                                                                      Mention of ‘stop-go’ macro policies, and of
(I-PRSP        Welfare Monitoring Surveys of 1994-1997.          strange. But the analysis is good, highlighting geography of poverty,        recent significant improvements in macro
Undated)       Also PPA done in 1996.                            landlessness, orphans, high fertility among poor. Good discussion of         and structural policies. No analysis of
                                                                 human development dimensions of poverty, and of the gender                   impact on the poor.
                                                                 dimensions of poverty among subsistence farmers.
Ethiopia       Household Income and Consumption                  No analysis of time trends using representative data. Reference to time      Good discussion of beneficial effect of past
(November      Expenditure survey (1995-96) dated. Welfare       trends from the panel data though questions about representativeness.        reforms (including liberalization and
2000)          Monitoring Surveys of 1996, 1997 and 1998 a       Panel studies should have been utilized more meaningfully in I-PRSP.         exchange rate policies), but the analysis
               useful source, but contain information mainly                                                                                  should have been taken further to reference
               on use of public services. Few measures of                                                                                     excellent panel studies. Little link between
               welfare outcomes at household level. HHICE                                                                                     policies and poverty analysis.
               survey repeated in 2000, but not utilized.


To top