Report No: 61393-NE
NIGER
Investing for Prosperity
A Poverty Assessment
October 15, 2012
PREM 4
Africa Region
Document of the World Bank


Fiscal Year
January 1 - December 1
Currency Equivalents
Currency unit: Franc CFA (FCFA)
US$1 = 510.00 (Exchange rate effective September 2012)
Weight and Measures
Metric system
Abbreviations and Acronyms
BCEAO            Banque centrale des Etats de I'Afrique de I'Ouest (Central Bank of the West
African States)
CAADP            Comprehensive African Agricultural Development Program
CGE              Computable General Equilibrium
CV               Coefficient of Variation
DHS              Demographic and Health Survey
ENVAM            Enqu6te sur Ila vuln6rabilit6 alimentaire des M6nages
ENBC             Enqu6te nationale sur le Budget et la Consommation des M6nages
FEW     ;j       Family Early Warning
GDP              Gross Domestic Product
HOI              Human Opportunity Index
MDGs             Millennium Development Goals
NGAC             Niger General Agricultural Census
NGO              Non-Governmental Organization
QUI BB           Questionnaire des Indicateur de Base de Bien-6tre
R&D              Research and Development
SDR              Stratigie pour le D&veloppement Rural (Country's Rural Development
Strategy)
SICCLA           Seeds Certification and Legislation Agency
SSA              Sub-Sahara Africa
TFP              Total Factor Productivity
U51VR            Under-Five (Child) Mortality Rate
UNICEF           United Nation Children Funds
WFP              World Food Program
Vice President : Makhtar Diop
Sector Director : Marcelo Giugale
Sector Manager : Miria Pigato
Task Manager(s) : Andrew L. Dabalen,
Janet M. Owens


NIGER: A Poverty Assessment
Table of Contents
ACKNOWLEDGEMENT                                          ................................................... vi
EXECUTIVE SUMMARY                                       ................................................... vii
Chapter 1: Poverty Trends And Profile    ..........................       .....  .............1
A.   Economic Performance            ........................................................ 1
B.   Poverty Trends and comparisons        .............................     .......... 3
C.   Minimizing Comparability Problems        ....................................... 5
D.   Poverty, Growth, and Inequality      ........................................... 10
E.   Sources of Change in the Distribution of Economic Welfare  ....................... 14
F.   Correlates of Poverty.......................                ...................  23
Chapter 2: High Vulnerability and Lack of Resilience    .......................   ......27
B.   Drought and Price Shocks Dominate....................................28
C.   Shocks Impose High Welfare Costs.....................................34
D.   Coping Strategies Suggest Weak Resilience...............................41
E.   Existing Safety Net programs are not well-targeted..........................43
F.  Therefore chronic poverty is high......................           ................45
G.   Conclusion:...................................................... 49
Chapter 3: Children's Opportunities In Niger..................................50
A.   Introduction..................................................... 50
B.   Poverty and Education.............................................. 52
C.   Correlates of Enrollment............................................ 55
D.   Poverty and Nutrition.........................................           ..... 56
E.   Poverty and Mortality       in.Ni..................................... ......... 58
F.   Evolution of opportunities.................50........            ................59
G.   Conclusion ...................................................... 64
Chapter 4: Agriculture, Income, and Rural Poverty     .......................  ....... 66
A.   Overview of theChapter            ......................................   ...... 66
B.   Overview of Agricultural Production...............................         ..... 67
C.   Farm Level Production: Stylized Facts.............................      ...... 70
D.   Participation in agricultural sector and household poverty ............... ...... 79
E.   Determinants of farm-level agricultural production and efficiency ............... 82
F.   Government spending and long term growth and poverty targets.     ............... 92
G.   Policies for improving agricultural production and increasing returns to the Poor. 101
References                             ................................................................ 103
Annex 1:   Figures and Tables: Vulnerability and Resilience ............... .....   .... 109
Annex 2    Children's Opportunities in Niger  ......................................... 120
Annex 3:   Figure s and Tables - Agriculture, Income and Rural Poverty  ............... 127
Annex 4:   Stochastic Production Frontier Model    .................................. 130
ii


List of Boxes
Box 2.1:  ............................................................. 46
List of Figures
Figure 1.1: GDP Growth and changes in GDP per capita, 2000-2008......................... 1
Figure 1.2: GDP Composition and Primary Sector Growth, 2001-2007 .................. 2
Figure 1.3: Distribution of Welfare and Adjusted Welfare, 2005 and 2007/08  . ............. 7
Figure 1.4: Dominance curve in 2005-2007/08   ........................    .......... 9
Figure 2.1: Household perceptions of shocks and difficulties during the last 12 months .......... 29
Figure 2.2: Rural household perceptions of shocks and difficulties during last 12 months........ 30
Figure 2.3: Urban household perceptions of shocks and difficulties during last 12 months...... 30
Figure 2.4.: Department-level annual rainfall means (2007, 2008, and 2009 vs. 20 year mean) 31
Figure 2.5: Mean and CV of rainfall, 1989-2009, by department........... .............. 32
Figure 2.6: Reported shocks among households reporting inferior harvests relative to previous
season...................................................... 32
Figure 2.7: Monthly millet prices and cumulative monthly rainfall during the rainy season (F
CFA), 2000-2010           ........................................ ...... 33
Figure 2.8: Perceived impact of food price increases during the preceding 12 months on
household food security                        ......................................... 36
Figure 2.9: Perceived impact of food price increase on nutrition of children under 5 years...... 36
Figure 2.10: Exposure of rural households to rainfall irregularities: absolute (left) and standard
deviations (right) from 20 year mean rainfall during the rainy season, June-
September, at department level ......................    ..... ......... 38
Figure 2.11: Sources of household food consumption/expenditures................... 40
Figure 2.12: Households reporting fewer meals than usual consumed by households and
children.................................................... 40
Figure 2.13: Households reporting fewer meals than usual consumed by households and
children.................................................... 41
Figure 2.14: Number of days (out of last 7 days) in which households decreased daily rations. 42
Figure 2.15: Number of days (out of last 7 days) in which households resorted to consuming
food "de p6nurie"        ...................................  .......... 42
Figure 2.16: Household reporting the sale of breeding animals to meet food needs in last 30
days....................................................... 42
Figure 2.17: Number of days (out of last 7 days) in which households consumed seeds due to
food insecurity.......................              ..................  43
Figure 2.18: Fraction of population receiving any transfer programs, 2007-2010. . ......... 44
Figure 2.19: Benefit incidence of subsidized cereal program, 2008 and 2010............. 45
Figure 3.1: Enrollment Rate for Children aged 7-13 by Consumption Quintile (Poorest to
Richest) .................................................... 53
Figure 3.2: Enrollment by location and poverty status  ................. ................ 54
Figure 3.3: Enrollment rate by gender, location and poverty status.................... 54
Figure 3.4: Opportunities in Education, Niger compared to neighbors, circa. 2006................ 61
Figure 3.5: Opportunities in Health, Niger compared to neighbors, circa. 2006..................... 63
Figure 3.6: Dissimilarity Index using one by one each circumstance (education) . ......... 64
Figure 3.7: Dissimilarity Index using one by one each circumstance (health)  . ............. 64
iii


Figure 4.1: Shares of Rain-Fed and Irrigated Production in the Agricultural sector (%) ......... 67
Figure 4.2: Distribution of Cultivated Land by Region (percent) ........... ................ 71
Figure 4.3: Average cultivated land by regions and departments (ha) ......................... 72
Figure 4.4: Per capita Cultivated Land (hectares)  ....................... .......... 72
Figure 4.5: Land Size Farmed by Region      ................................. ..... 73
Figure 4.6: Average Cultivated Land by Region and Gender (hectares)................. 73
Figure 4.7: Main Activity by Gender of Household Head (percent)  .........  ........... 74
Figure 4.8: Cultivated land size by crops and gender (ha)  ..................... ..... 76
Figure 4.9: Input Usage and Access through Market (%)                .................. 77
Figure 4.10: Sources of Household Income (%) ......................... 80
Figure 4.11: Share of Agricultural Income by Departments and Poverty Status (%) ............... 81
Figure 4.12: Agricultural Participation and Poverty Incidence       ................... 81
Figure 4.13: Output elasticities with respect to inputs.......................   85
Figure 4.14: Access through Market ........................................... 85
Figure 4.15: Distribution of farming efficiency.................................... 86
Figure 4.16: Efficiency by gender and education                 ................................. 88
Figure 4.17: Efficiency by gender and use of mechanization          ...................  88
Figure 4.18: Efficiency and market distance                   .................................... 89
Figure 4.19: Efficiency and rainfall ............................................ 90
Figure 4.20: Sorghum yield and rainfall......................................... 90
Figure 4.21: Millet yield and rainfall........................................... 91
Figure 4.22: Relationship between farming efficiency and on-farm investments................... 91
Figure 4.23: Spending in agricultural sector ........................................ 93
Figure 4.24: Disbursement of projects funds by sectors (%) ......... ................ 94
Figure 4.25: Public agricultural R&D spending (Billion 2005 FCFA) ........... ........... 95
Figure 4.26: NAIP, CAADP and MDG1 Budgets (billions of FCFA)  ........................... 96
Figure 4.27: Growth rates by scenarios.  ........................................ 97
List of Tables
Table 1.1: Distribution of employment and productivity, by sectors...................... 3
Table 1.2 : Survey coverage and date in the field  ........................ ......... 4
Table 1.3: Comparison of Recall periods for some items in 2005 and 2007/08... ............ 5
Table 1.4: Poverty Indicators, 2005 and 2007/08  ..................................... 8
Table 1.5 : Poverty by Geographic Regions...................................... 10
Table 2.1: Reported household impacts of shocks ranked as most important (2007) ........... 35
Table 2.2 : Impact of shocks on per capita consumption, food expenditures, and millet
consumption: regression coefficients for various self reported shocks................ 37
Table 2.3: Impact of rainfall shocks on per capita consumption (total, food, millet): regression
results ....................    .      ...................... ........ 38
Table 2.4 : Price/rainfall elasticity of per capita household consumption: regression results .... 39
Table 2.5 : Movement In and Out of Poverty (% of population)  ................. ..... 47
Table 2.6 : Characteristics of the Chronic Poor .....................  ............. 48
Table 3.1: Differences in nutritional status of children across regions. ....... ........... 57
Table 3.2: Human Opportunity index in education and health  .................. ..... 60
Table 3.3: Decomposing total change in Human Opportunity Index.............. ..... 61
Table 4.1: Main exports by Value, 2001-2005    ............................  ...... 68
iv


Table 4.2 : Comparison of Niger's yields of cereal production with Sahelian countries Average
yields, in Kg/Hectare (2003-2007)   ............................  ...... 68
Table 4.3 : Growth accounting of agricultural sector (%)  ............... .......... 70
Table 4.4: Cultivated Land in Pure Stand by Regions and Crops (%) .......... ... ................ 71
Table 4.5 : Land Management by Gender (%) .......................          ...... 75
Table 4.6 : Main sources of agricultural inputs (%) .......................... 78
Table 4.7 : Main causes of production change.................................. 79
Table 4.8 : Use of External Funding by Farmers (%) ...................... 79
Table 4.9 : Average livestock income (millions, FCFA)  ........................ ..... 82
Table 4.10: Average crop income (millions, FCFA)  ................................. 82
Table 4.11: Estimation results........................           ................  84
Table 4.12: Potential drivers of efficiency gap               ................................... 86
Table 4.13: Government budget allocation (%) ....... ...................... 94
Table 4.14: Key agricultural investments (2001-2008)  ........................ ..... 95
Table 4.15: Composition of Niger National Agricultural Investment Program ...... ....... 97
Table 4.16: Key simulation parameters........................................ 99
Table 4.17: Annual average TFP growth (%) ............................... 99
Table 4.18: Impact of irrigation expansion on growth rates (%) ........ ............ 100
v


ACKNOWLEDGEMENT
This report was prepared by a core team consisting of Andrew Dabalen and Janet M. Owens
(Co-TTLs), Nazmul Chaudhury, Roy Katayama, Prospere Backiny-Yetna, and John Ulimwengu
(IFPRI, consultant). Siobhan Murray estimated local area rainfall and developed rainfall maps.
Paula J. White provided excellent assistance with document preparation and editing often on
very short notice. Salifou Noma delivered critical administrative and logistic support in Niamey.
The report was prepared under the guidance of Marcelo Giugale (Sector Director, AFTPM) and
Jan Walliser (Sector Manager, AFTP3).
We are grateful to the following for very helpful suggestions and overall guidance: Ousmane
Diagana (Country Director, AFCW2), Nestor Coffi (Country Manager, AFMNE), and Katrina
Sharkey (Country Program Coordinator, AFCF1 and AFCF2). The Team benefited from comments
provided by Harold Alderman (Peer Reviewer and Consultant, DECRG), Madhur Gautam (Peer
Reviewer and Lead economist, SASDA), Robert Johann Utz (Senior country economist, Niger)
and Jan Walliser (Sector Manager, AFTP4). We are grateful to the staff of the Institute of
Statistics of Niger (INS) and the Ministry of Agriculture for sharing their survey and macro data
sets which we used heavily in each of the chapters. We would also like to express our gratitude
to the Cellule Crises Alimentaires (CCA), or the food security unit, which kindly shared the price
data.  Many thanks, too, to the meteorological department for the rainfall data and
consultations on rainfall estimation, and to the Ministry of Hydraulics and the Environment, for
providing commune-level map data.
VI



EXECUTIVE SUMMARY
1.   Income per capita in Niger has not changed in nearly a decade. Looked at over
the past 30 years, income per capita has in fact declined, so that in 2009, it stood at
about 30 percent less than in 1980 (Figure 1). This has profound implications for the
well-being of the population. This report examines poverty trends and distribution of
the poor in this larger context, paying particular attention to the most recent past. The
report contributes to our understanding of the progress made in combating poverty in
three ways. First, it updates our knowledge of poverty outcomes by examining the
trends in poverty and vulnerability, as well as the profile and distribution of the poor
and vulnerable across the country. Second it looks at the most common shocks, and
their scale and influence on the welfare of the population. Third, it highlights the
progress the country has made in improving opportunities for acquiring human capital
and increasing incomes in rural areas. In this respect the report examines changes in
access to education and health and improvements in productivity and income in small
holder agriculture. It also explores the potential impact of public investments in
agriculture. On the basis of the findings, the report makes modest recommendations on
how to improve welfare outcomes.
Figure 1: GDP per capita: 1960-2009
1000
900
700
7   600
Soo
6 E400
300
200
100
0
Source: World Development Indicators
2.    The report finds that, the biggest achievement in the last decade has been the
substantive improvement in opportunities to acquire human capital. To be sure
income poverty appears to have declined as evidenced by a decrease in the fraction of
the poor below the poverty line by about 4 percentage points from 2005 to 2007.
Furthermore this decline has been slightly faster in rural areas. Inequality, whether
measured by the Gini index or consumption gap between the top and bottom deciles
also declined. However, the most noticeable changes happened to be the expansion of
vii


opportunities in health and education. The human opportunity index (HOI) - which is a
coverage ratio that takes into account the equity of access to basic services - for
children's education and health, rose sharply between 1998 and 2006. A detailed probe
into the reasons for these large changes reveals that both expansion of coverage to all
groups (scale effects), while at the same time equalizing gaps in coverage between
groups differentiated by wealth, location (urban/rural), regions, ethnicity and so on
(equalization effects), played key roles (Figure 2).
Figure 2: Decomposition of Total Change in HOI, 1998 - 2006
H Composition  U Scale  8 Equalization
14
12
10
6
4
2
0
-2              -0)       0)                                       ZF0
-o                 E)~        C-
o                                       E
Opportunity
Source: World Bank staff estimates from survey data
3.    Niger has made considerable gains in increasing coverage of education and
improving the health of children, but it can do more.  Primary school attendance
among rural children lags behind their urban peers. Coverage may be affected by both
factors of demand (family-based economic considerations, social norms that promote
early marriage among girls, and lack of belief in the efficacy of a basic education) and
supply (school placement, school quality, gender- sensitive content and infrastructure
considerations) which are especially sensitive in rural areas. Thus, to overcome urban-
rural differences in educational achievement, parents, teachers, administrators, and
policy makers should adopt strategies to promote changes in attitudes and incentives
towards school attendance for girls and boys in rural areas.
viii


4.    Niger remains an environment characterized by high vulnerability to shocks
and poverty. The shocks are frequent and often catastrophic. Droughts, which are
remembered for the hunger they bring, are the most prominent. But just as common
are the correlated price shocks. As an example, in the past couple of years, Niger has
experienced a severe drought which brought food insecurity to 2.7 million people, a
political crisis, floods and a global financial crisis which added to the hardships by
shrinking demand for commodity exports which Niger depends on disproportionately
for revenues. Even in areas, such as opportunities for children in education and health,
where progress has been much higher, children's circumstances continue to matter a
lot. In particular, opportunities remain very low and lag behind their neighbors. Human
opportunity index for school attendance is barely higher than 50 percent and just about
60 percent for child nutrition. And the ratio for finishing primary school on time was a
mere 3 percent in 2006. By comparison, HOI for school attendance and child nutrition in
Ghana was 80 percent.
5.    As a result, the economy, but especially the rural farm households, experience
huge income volatility and welfare losses. In a period of four years, spanning 2004 and
2008, Niger's GDP has dipped to negative territory in one year and grown by 10 percent.
And this pattern of positive growth for a few years, followed by (often large) negative
growth, has been going on for a decade. Agriculture, which provides a livelihood for
almost 80 percent of the population, is stuck in a low-productivity equilibrium, unable to
sustain a few years of normal growth. With the exception of rice, yields of other staples
(millet, sorghum and groundnuts) are at most half of the West African average. They are
also lower than Sahel countries facing similar agricultural endowments and risks. For
instance, yields of millet, the main staple, are twice as high in Burkina Faso and 63
percent higher in Mali.
6.    The welfare costs of these shocks are large. First, there is the enormity of the
uncertainty of livelihoods which is psychologically taxing. Then the shocks themselves
create a cycle of busts and recoveries in most of rural incomes. The effects of the busts
are potentially long lasting. As an example, households which receive rainfall that is 100
mm less than long term (20 years) departmental average would see consumption
decline by 7 to 13 percent, which is not easy to recover in one year. Our estimates also
show that a 10 percent reduction in rainfall could lead to a 10 percent reduction in the
main staple - millet. Health shocks are estimated to lead to similar magnitudes of
consumption loss. Besides immediate welfare losses, the risks and shocks also induce
risk-averse behavior. They partly explain low levels of uptake of modern agricultural
inputs, which in turn explains low agricultural productivity and consequently high levels
ix


of impoverishment. It is this risk environment and multiplicity of shocks that greases the
wheels of the low productivity traps.
7.    Furthermore, households coping   strategies suggest weak   resilience.
Households use a long list of strategies to cope with the shocks. These range from
reducing consumption to depleting assets. Most of the strategies are inadequate for
covariate shocks and potentially harmful. They are inadequate because when a
covariate shock strikes a household, mutual insurance will not be available, and there is
a limit to how long a household can reduce consumption of its members. They are
harmful because forgoing nutrition, health and depleting assets will undermine long
term productivity, and potentially entrap these households and their children in poverty
over a long time.
Building a Resilient Society
8.    Given the risk environment and levels of vulnerability, Niger needs to build
resilience. There are two interconnected paths to achieving resilience. One is to
improve incomes overall and the other is to have instruments to insure income
volatility.
Increase incomes by transforming agriculture
9.    To improve overall incomes, Niger will have to focus on improving incomes in the
sectors where most of the population earns its livelihood, which is primarily agriculture.
Just improving agricultural productivity to reach the Sahelien average, would improve
yields by 50 percent, which would be a huge boost to incomes and nutrition of the
population. It is a well-known fact that Niger agriculture is characterized by minimal use
of modern inputs, which in part explains its low yields. As we discussed, the risk levels
that Nigerien farmers face explains, partially, the reluctance to adopt high-yielding but
risky inputs. This calls for four policy considerations to improve agricultural incomes.
10.   Expand irrigation infrastructure. Agricultural productivity in Niger is low in part
because it is too dependent on low and erratic rainfall. Irrigation provides one solution
to minimize such weather risk to production. At present irrigation is confined principally
to areas around the Niger River. Irrigated crops include rice, onions and some high
valued fruits and vegetables. However, there is potential for increasing the use of
irrigation to boost agricultural productivity and food security. Past studies have shown
that productivity increases from irrigated export and food crops are likely to generate
the greatest gains in household welfare. The government is planning to invest heavily in
irrigation, as indicated by the outlay on the multi-purpose Kandadji dam which is
x


purported to expand irrigated land by 30,000 hectares over the next 30 years But there
is widespread consensus that even more areas could be brought under irrigation
through expanding use of River Niger and what is believed to be substantial
underground water resources.
11.   Provide insurance for agricultural production, especially weather insurance.
Within the context of Niger's risk environment, it is understandable that farmers may be
reluctant to adopt modern inputs because they face huge losses in the event of failure.
To overcome this hurdle the farmers need insurance. A few experiments being tried in
other parts of the world hold promise for Niger. One suggestion would be to offer
farmers weather-indexed income insurance - paid through a small premium - which is
paid out in the event of a severe drought. Since the payout will be conditioned on an
exogenous event, the transaction costs will be lower as are the agency problems on the
part of the farmer. Rather than insure individuals, another consideration would be to try
community insurance, where payments will be to a community in the event that a major
shock (again indexed to an exogenous event) wipes out the harvest for that community.
Another idea would be to offer insurance for modern seed purchases as is being tried in
Kenya. In this case, the farmers pay a small premium on seed purchases at the time they
buy the seed, which automatically enrolls them in an insurance scheme. In the event of
harvest failure, they receive a cash payout to buy more seeds for the next planting
season. In all cases, the programs can be transacted through mobile phones, which can
be a huge advantage in a country like Niger, where populations may be geographically
dispersed.
12.   Re-invent gender-sensitive extension services through mobile technologies
with complementary investments in agricultural research. Extension services have
long been used to promote adoption of new technologies to boost agricultural
productivity. The traditional system relies on face to face interaction, but because of a
host of problems - underfunding, understaffing, low skills and low motivation of staff -
it has not proven effective in many contexts. However, the opportunity exists now to
use the new technologies to reach many more farmers at once and in a timely manner,
accounting for differences in literacy competencies. Given the high penetration of
mobile phones, there is a possibility to use this medium to deliver useful information to
the farmers at the time they need it In this regard, a highly functioning agricultural
extension service that promotes state-of the art practices stands as a necessary
complement to cell phone technologies that can be used to disseminate concise
information quickly. Certainly, extension and agricultural research activities will need to
be coordinated to improve the management of farm level production - that is improve
technical efficiency - among rural producers and reduce the gaps between actual and
xi


potential efficiency. As the effort to restructure currently ineffective extension service
continues, it would be necessary to ensure that services respond to needs of producers
in a gender-sensitive manner accounting for differences in crops farmed and production
technologies utilized.
13.   Finally, improve efficiency of output and input markets. Input usage continues
to vary widely across regions, but there are signs of growing competence of institutions
for input distribution, including the seed certification and legislation agency, seed
producer farmers and community managed input stores. Yet, much still remains to be
done to improve the efficiency of these critical markets. For instance, fertilizer usage
depends on several factors including having physical access on a timely basis and cash or
credit to finance purchases. Moreover, the capacity to access markets may be
differentiated by gender. Therefore, the timing of fertilizer usage may be especially
critical in high value crop production technologies. The improvement of information
flows is especially needed for grain markets which are crucial for the food security needs
of Nigerien households. As in extension, there are opportunities to use mobile
technologies to improve information and efficiencies in the grain markets. Two such
opportunities include:
* Sub-national weather and price monitoring. The government currently
monitors rainfall data through a series of stations through standard methods.
The stations have expanded over time and there appears to be better data
collection. Similarly, prices of main staples from multiple (mostly urban)
markets are monitored. But rainfall and price data exhibit strong correlation
and huge variability. Therefore, it is important to expand the number of
stations and markets from which this information is collected. This could be
done relatively easily now using mobile technology. There is no obstacle to
obtaining this information on a regular basis from many more villages than
are currently covered.
* Sub-regional grain price monitoring. Niger's grain supplies and prices are
highly correlated with the status of markets in the neighboring countries,
especially those in Northern Nigeria. In one study, price changes in Benin and
northern Nigeria was found to influence price changes in 75 percent of the
markets in Niger (Aker, 2008). Therefore, a monitoring system that is focused
only on the Nigerien markets will not be able to predict changes in the food
market very well. But such regional market monitoring may be easier to do
now than in the past. It can be done through gathering prices from a group
of inter-connected regionally influential grain merchants or government-to-
xii


government agencies sharing information. While some of this activity
currently exists it could be scaled up on a more formal basis.
Reduce vulnerability through a robust safety net system
14.   In addition to increasing incomes Niger needs to improve its resilience to shocks
by building an effective safety net system. Ideally, a good safety net system will provide
income support to the extremely poor during normal and bad times, but also scale up to
cover, only temporarily, those affected by large and covariate shocks. At present there
exist several safety net programs ranging from cash transfers to school feeding. Most
are designed to respond to emergency assistance, and while they may provide relief in
such difficult times, their impact on reducing chronic food insecurity remains limited.
Therefore, an effective safety net system is needed to provide real security on a
permanent basis.
15.   But this will require knowing the poor and the vulnerable in ways that are far
better than currently exists. As noted above a monitoring system to obtain better price
and weather information will help provide a better early warning system regarding the
localities that are badly affected by a severe shock. However, this in itself would not be
enough. To respond more quickly and protect the population, it is important that Niger
collect better information about who the poor are. As this report has shown, it has not
been easy to determine the size of the poor and the progress made because the surveys
that underpin such an exercise have been conducted too infrequently and
inconsistently. Similarly, the safety net programs are not well-targeted, in some cases
even regressive, in part because the programs have limited knowledge of the truly poor
people. Therefore, it would be helpful to put in place a system that better identifies the
poor through (a) a modern register of the safety net eligible recipients and (b) a better
monitoring and evaluation of the programs' impacts.
16.   The government is putting together the building blocks of such a system, which
includes a mix of cash transfers and programs conditioned on work - e.g. cash for work
- and which can be scaled up during emergency situations. The planned safety net
program is expected to contain many innovations, such as building a modern
management information system and targeting and payment systems. A key proposal in
implementing effective payment systems is the possible use of mobile technologies to
transfer the benefits to beneficiaries. Proper execution of the planned safety net
program holds enormous potential for reducing chronic food insecurity, building the
assets of poor and strengthening resilience of the population.
xiii


Strengthen the foundation of poverty monitoring through integrated and
timely surveys
17.   Increasing prosperity and resiliency would entail substantial investments and
trade-offs in spending. Therefore, it is essential to be transparent about the real impact
of the choices made. In particular, monitoring of interventions should provide the basis
for assessing their geographic impact and measuring overall performance. This is
especially important for Niger which operates in a resource-constrained environment
where most investments are financed through external assistance. Past evidence
suggests that resources allocated at national level don't always reach local level service
providers.
18.   As this report has pointed out, it has been difficult to assess the poverty impacts
of past investments because the available survey data has been both insufficient and
inconsistent. In the past 6 years, there has been several surveys - a core welfare survey,
a household budget survey, and 4 food security and vulnerability surveys, to mention a
few. Still, even with all these surveys, there are several problems with the data
available. First, comparable surveys, meaning surveys conducted in the same way over
time, are rare. This means that surveys that would allow us to learn about the real
impact of development interventions are infrequent. Second, many of the surveys are
not comprehensive enough and miss crucial information that would help us understand
the reach and performance of many government programs. For instance, not all surveys
collect information on safety net programs, or key agricultural investment programs.
This information could provide basic knowledge on the depth of coverage of key
programs and whether project resources are reaching beneficiaries.
19.   Yet, to learn how these programs affect the lives of intended beneficiaries, it is
absolutely essential to collect frequent and timely information that is accurate,
consistent and comprehensive. The Living Standard Measurement Survey - Integrated
Survey in Agriculture (LSMS-ISA) provides the foundation for such a system. First it is
integrated and comprehensive, meaning that it is designed to collect information from
multiple sectors with one instrument. For example, as the title suggests, it will provide
detailed information about agricultural investment and production at the household
level, and sometimes try to account for intra-household actions. Second, it will be
conducted in a consistent and comparable manner over time. Third, it is designed as a
panel that is repeated every three years, so it will improve the frequency of comparable
data available. Furthermore, once it is accepted as a core of the household survey
program, it provides opportunities for innovation such as selecting a few households in
the survey to "listen to" - i.e. collect very timely (possibly monthly) and relevant
xiv


information - using mobile phones. However, at present the survey relies on donor
support, so sustainability is an issue. We view the LSMS-ISA as a key tool for identifying
and knowing the poor and thereby enabling the effectiveness of the planned safety net
system, to monitor vulnerability across Niger, and more generally to monitor and
evaluate the development impact of investments. Therefore, we strongly recommend
that the Government of Niger ensure its sustainability by taking responsibility for its
future financing.
xv



Chapter 1:           Poverty Trends And Profile
1.1   This assessment of poverty and inequality comes at an important juncture for
Niger. After 10 years of relative stability, the country experienced a political crisis in
2009 when its then incumbent president, Mamadou Tandja, refused to abide by term
limits established in the constitution and relinquish power after two 5-year terms in
office. The crisis led in early 2010 to a military coup d'etat. The military government
proceeded by empowering technocrats to make institutional reforms and indicated that
it would re-instate democratic processes by establishing presidential and parliamentary
elections within a one-year time period. This transition was concluded in February
2011, when a new president was elected after a second round vote. In contrast to the
Tandja regime, the military government began its rule with open dialogue, including
acknowledging that the country was confronting a severe food deficit after unusually
low rainfall in 2009 undermined crop and forage production, and therefore reduced the
incomes of both crop and livestock producers.
A.    Economic Performance
1.2   Prior to the political crises, the Tandja regime had been credited with fostering
economic growth and improving poverty reduction by ramping up foreign investment in
the mining and oil sectors and targeting government resources to improving access to
and the provision of quality public services.  Despite relative political stability and
increased investment, household prosperity improved little over the last decade. The
graph below (Figure 1.1) indicates that GDP per capita grew only slightly more than 11
percent over the period or an average annual growth rate of a little more than 1.2
percent. During the same period, economic growth was highly variable and the
fluctuation has been detrimental for overall wealth accumulation.
Figure 1.1: GDP Growth and changes in GDP per capita, 2000-2008
G
C                                                        AG
D
a                                                        n r
P
p                                                        n o
I                                                        U W
t                                                        a  t
E
a                                                        IMb=
GDP Growth (Annual %)  --GDP per capita (constant 2000 US$)
Source: World Development Indicators (WDI)
-1-


1.3   Figure 1.2 below presents the major sectoral components of Niger's economy for
2000-2007. Clearly, the agricultural sector dominates the economy: it accounts for over
40 percent of GDP and its relative share has remained unchanged over the period.
However, agricultural sector growth has fluctuated widely and these fluctuations
correspond strongly to GDP per capita trends presented above.
Figure 1.2 GDP Composition and Primary Sector Growth, 2001-2007
1.4   The dominant role of agriculture in GDP is also evident from the vast majority of
the Nigerien labor force employed in it. Table 1.1, below shows the distribution of
employment by sector and their productivity contribution. The employment structure is
based on annual household survey data but it is likely to be representative of historical
trends. Sectoral composition has not changed over time, nor has the share of rural
population. Based on household survey estimates (ENBC-2007/08), over 80 percent of
working-age adults are employed in agriculture, yet this sector has the lowest level of
productivity in the economy. Thus, most of the Niger labor force remains trapped in the
least productive sector of the economy. However, to make progress on poverty
reduction the sector must become more productive given the close relationship
between employment and agriculture. Enhanced productivity requires improvements in
the level of technical and managerial skills used in the sector and an orientation towards
high value crop production. Investments made in the agricultural sector are likely to
have pro-poor growth impacts since the sector accounts for the largest share of
employment.
1.5   By comparison, the mining sector is the most productive but it employs only a
fraction of the population - less than one percent. The anticipated mining boom which
stems from expected price increases in extractive commodity prices and increased
extraction of those commodities in Niger will generate significant government revenues.
If properly captured within the national budget and managed with transparent
processes, the revenue could be used to ramp up the quality and quantity of public
services and improve development. Thus the second round effects associated with the
-2-


increased returns to the mining sector could have positive investment effects in the
public sector but they won't necessarily have pro-poor employment effects.
Table 1.1: Distribution of employment and productivity, by sectors
% GDP     Growth        %           %         Relative
2000-2007  employment    non-ag    productivity
(ag + non-ag)  employment   1
Agriculture              46.1        7.9       83.2          --         0.6
Mining                    4.5       -1.0        0.2         0.8        19.3
Industry excl construction  6.1     -0.9        1.4         5.0         4.4
Manufacturing             4.8       -0.2        1.1         4.0         4.3
Utilities                 1.3       -3.1        0.3         1.0         4.6
Construction              2.3        7.8        2.0         7.3         1.1
Trade, transport, and other  31.8    4.1        9.9         35.4        3.2
services
Trade                    17.2        3.8        1.7         6.1        10.1
Transports                5.4        6.8        1.7         6.0         3.2
Other services            9.2        3.4        6.5        23.1         1.4
Government                9.2        5.9        3.2         11.4        2.9
Source: Skills Development, World Bank 1  Relative productivity is obtained by dividing each sector's output share
with the employment share.
1.6   The observed changes in GDP per capita growth and sectoral composition of GDP
growth have strong implications for poverty outcomes during the last 5 years. In
particular, weak and variable performance in the agriculture sector over the past several
years - especially in 2002, 2004, and 2009 is likely to have had a negative impact on
agricultural incomes. Droughts in these years combined with significant food price
increases, provoked subsequent food crises and dramatic spikes in child malnutrition
and other health indicators. Therefore, we would expect that overall changes in poverty,
at best, would have been minimal. However, as we explain below, measurement of
poverty performance over time in Niger is imprecise because of lack of comparable
data.
B.     Poverty Trends and comparisons
1.7    Between 2005 and 2010, Niger's National Institute of Statistics (INS, French
acronym) and supporting partners have collected a lot of data using multiple household
surveys.  In 2005 INS conducted a household survey using a relatively short
questionnaire (QUIBB) over a 3-month period, which was used to produce poverty and
other welfare estimates for the Niger PRSP. Then in 2007/08, INS fielded a multi-
module budget and consumption questionnaire (ENBC, French acronym). At around the
same time, together with World Food Program (WFP), annual vulnerability surveys
(ECVAM, French acronym) were conducted starting in 2006.
1.8   All these surveys were conducted to help us understand the living conditions of
households, the shocks they dealt with, their use of social services (e.g. education,
health, water and so on), and their sources of income. Taken individually, there is much
-3-


we can learn on these topics from each of the surveys. However, to understand a
country's progress on multiple dimensions of living standards, we would often be
interested in changes over time. Unfortunately, the numerous surveys in Niger, all done
within the last five years (2005-2010), do not allow easy tracking of the evolution of
living standards. There are several reasons for lack of comparability.
1.9   Seasonality in consumption: The ENBC covered 12 months of data collection to
minimize the impact of welfare changes arising from seasonal variation in consumption.
By contrast, the QUIBB was conducted for 3 months in 2005. To complicate matters,
the ECVAM surveys were conducted at different times as well (see Table 1.2).
Therefore, to the extent that seasonal patterns of consumption matter in Niger, as we
believe they do, comparability of all these surveys will be undermined.
Table 1.2: Survey coverage and date in the field
Survey                    Begins           Ends           Coverage
QUIBB 2005              April 2005       July 2005   national
ECVAM 2006               Nov 2006        Dec 2006    national
ECVAM 2007               Dec 2007        Jan 2008    excludes rural Agadez
ENBC 2007               April 2007      March 2008   excludes rural Agadez
ECVAM 2008               Nov 2008        Dec 2008    excludes rural Agadez
"Rural" ECVAM 2010      April 2010      April 2010   excludes Agadez
"Urban" ECVAM 2010      April 2010      April 2010   regional capitals
1.10  Diary versus prospective recall: Additional problems with comparability arise
from how households were asked to remember their consumption patterns. In the
ENBC, a daily recording for a period of 7 days, of key common consumption items such
as food was used. The enumerators involved in the collection weighed all food items
and measured the volume of all containers used to conduct transactions of foodstuffs.
The diary is considered the most reliable and accurate method; it is also the most costly.
The QUIBB, on the other hand, utilized multiple prospective recall periods to collect
foodstuffs. Enumerators asked respondents to make average monthly calculations
based on whether the respondent had reported consuming the mentioned item any
time during the last 12 months. Prospective recall method was also the method of
choice for collecting household consumption for the ECVAM surveys. Past survey
research has indicated that the length of recall will affect responses. Higher reported
expenditures on food correlate positively with shorter recall periods. Longer recall
periods are likely to result in under-reported consumption-especially for small,
frequent purchases. Reports on durable goods expenditures are also likely to vary with
length of recall. Table 1.3, presented below compares length of recall for all
commodities between the QUIBB and the ENBC.
1.11  Expansion of the list of consumption goods: A third source of potential non-
comparability is the expansion of some food items in the ENBC. As a result, the choice
of goods represented in the consumption basket is more detailed compared with the
QUIBB basket. For example, the list of food items to report consumption is about 80 for
-4-


the QUIBB and 250 for the ENBC. In the ECVAM           surveys, the list is shorter than both
QUIBB and ENBC. All else being equal, we assume that the longer list will correlate
positively with reported food expenditures, and thus negatively with the estimates of
poverty. Similar differences exist among non-food items as well. We expect that
changes in the share of food to overall expenditures across the two survey periods, from
more than 67 percent in 2005 to less than 54 percent in 2007-08, may have resulted
from an artifact of the questionnaire design. The dramatic decrease in the share of
foodstuffs relative to durables is likely to lead to a decrease in the estimated poverty
rate.
Table 1.3: Comparison of Recall periods for some items in 2005 and 2007/08
Recall period                Technique of data collection
2005             2007/08            2005           2007/08
Common recall items
Housing renting            30 days             1 month         Actual expenses   Actual expenses
Electricity                30 days             1 month         Actual expenses   Actual expenses
Water                      30 days             1 month         Actual expenses   Actual expenses
Education                  Current school year  Last 12 months  Actual expenses  Actual expenses
Non common recall items
Food, alcohol, tobacco     30 days             7 days diary    Monthly average   Actual expenses
and number of
month of
consumption
Clothes                    6 months            12 months       Actual expenses   Actual expenses
Health, consultation, drugs  30 days           6 months        Actual expenses   Actual expenses
Health, hospitalization    30 days             12 months       Actual expenses   Actual expenses
Transportation (rural      30 days             1 month, 3       Monthly average  Actual expenses
transportation, minor repairs,                 months or 6     and number of
etc.)                                          months          month of
consumption
Transportation (urban      12 months           1 month, 3      Actual expenses   Actual expenses
transportation, other repairs)                 months or 6
months
Transportation (oversee)   12 months           6 months        Actual expenses   Actual expenses
Restaurant                 30 days            3 months         Monthly average   Actual expenses
and number of
months of
consumption
Source: Author's compilation using 2005 and 2007/08 questionnaires
1.12    Finally, INS modified the ENBC data to improve comparability across 2005 and
2007/08. The modifications included additional corrections of outlier observations in
three regions, beyond the usual adjustments for outliers that were done for all regions.
It is not clear that such selective editing, beyond what is standard practice, does not
itself introduce considerable errors.
C.      Minimizing Comparability Problems
1.13    We suspect that the multiple ways in which the survey differences arise could
introduce non-trivial errors in the estimation of poverty indicators across time.
Without making adjustments for these changes, one is likely to estimate incorrect
-5-


trends and changes in poverty rates. Therefore, below we discuss some steps we take to
minimize the impact of these differences. We first describe how we reconcile the
differences between QUIBB and ENBC, and then explain how we use the ECVAM in
informing our knowledge of poverty trends.
1.14  First, we create a common consumption basket between the QUIBB and ENBC to
minimize the impact of consumer basket differences on respondent recall and thus
reported consumption. Effectively, we exclude the consumption goods or services that
were not common in all years. After the adjustment, consumption stays relatively
similar.
1.15  Second, we re-weight the QUIBB data relative to the base year data (ENBC) to
minimize the impact of differences in recall periods across the two surveys and their
effect on reported consumption. We use a method proposed by Tarozzi (2007). The
method relies on a reweighting of the consumption aggregate in all years relative to a
base year. The weights are obtained from a procedure that relies on using parts of the
consumption aggregate for which the recall period has not changed over time together
with household characteristics that have been collected the same way across years.
These observable household characteristics and the set of expenditure items with
identical recall periods over time are used to generate propensity scores, or the
conditional probabilities that an observation belongs to a particular survey. Using the
propensity scores we then construct inverse probability weights to correct the original
welfare indicator of the target survey (QUIBB) and to generate a comparable poverty
trend.
1.16  As part of the reweighting procedure we must identify one of the surveys as the
auxiliary, or base survey, the one which is assumed to generate more precise
information for generating the consumption aggregate, and the other survey as the
target survey. Between the QUIBB and the ENBC surveys, the ENBC is better suited to
measure consumption, and thus measure poverty, although it also has some drawbacks.
Moreover in the future, poverty monitoring is more likely to be based on the design of
survey methodologies more closely resembling the ENBC. Thus, we maintain the ENBC
as the base survey and the QUIBB as the target.
1.17  We estimate the propensity score using a probit model, and we include the
following explanatory variables that are common to both surveys: household
composition (the number of household members by age and gender, and their squared
values), the education of the head, the characteristics of the dwelling, the possession of
some durable goods and the per capita expenditure of goods collected with the same
recall periods. The inverse probability weight derived from this model is used to adjust
the welfare indicator. Figure 1.3 is a plot of kernel densities which shows how the
distribution of welfare compares across survey periods, before and after the
adjustments are made.
-6-


Figure 1.3: Distribution of Welfare and Adjusted Welfare, 2005 and 2007/08
Distribution of welfare - National
08
8     9     10    11    12    13    14     15    16    17
Logaritin of per capita expenditure
-4-2005oRiginal  --+2005 conunon    2005comnon_e
-#-2007original  ---2007conuon
Source: Author's calculation using 2005 and 2007/08 surveys
1.18  Third, we also address the issue of data editing targeting three regions by
reverting to using the ENBC data files that treated the outliers in the same way in all
regions and did not do any additional treatment of outliers for three regions. This would
lead to a revision of the poverty line from 144, 750 F CFA to 156,172 F CFA per capita
per year 2005 and from 150,933 F CFA to 162,842 F CFA in 2007/08, using a caloric
intake of 2100 Kcal per day per person (Backiny et al., 2008). The lower poverty lines
are those obtained with heavily edited data, while the higher poverty lines are obtained
with unedited data. The welfare measure we use is consumption, which INS (2006,
2008) constructed using food and non-food spending, consumption from own
production and the shadow value of rent attributed to houses in which the occupants
are the owners of their residence and, thus, do not pay rent. We use consumption per
capita in all years to be consistent with the practice started by INS in 2005.
Welfare Gains
1.19  In Table 1.4 below, we present poverty estimates for 2005 and 2008. We include
the INS estimates (before adjustments) and our own (after adjustments). Our estimates
confirm the downward trend in poverty that INS reported (2008). Poverty decreased
from 64.4 percent in 2005 to 60.8 percent in 2007/08-a decrease of 3.6 percentage
points. We estimate a slightly larger reduction in poverty than the 2.7 percentage point
difference of INS. Our estimates diverge further from INS at disaggregated urban and
rural levels. We find that poverty decreased by 2.8 and 4.1 percentage points in urban
and rural areas, respectively. In contrast, INS attributes the greatest decrease in poverty
incidence of 7.5 percentage points to urban areas and only a 2 point reduction to rural
areas. Our rural-urban estimates would seem to be more realistic when accounting for
the fact that the QUIBB was conducted after widespread post-harvest losses resulting
from drought of 2005. Thus, we would expect to find the greatest gains made after the
2005 drought in rural areas.
-7-


1.20    The gains made between 2005 and 2007/08 are further corroborated by the
declines estimated among other poverty indicators. Both the national poverty gap and
squared poverty gap have fallen sharply, which means that the depth of poverty has
decreased and that average income among the poor has grown closer to the poverty
line and that the poorest are better-off. However, while we note here that gains were
made, we caution that the large changes between these two years are still sensitive to
differences in survey design methodologies. In particular, the ENBC survey collected
many more food items, which is one of the differences that can't be perfectly resolved.
Table 1.4: Poverty Indicators, 2005 and 2007/08
2005                                    2007/08
Poverty  Poverty  Squared  Share   Share  Poverty  Poverty  Squared  Share   Share
head    gap    poverty     of     ofthe   head    gap     poverty    of     ofthe
count            gap    population  poor  count            gap    population  poor
New estimates
All            64.4    28.9     16.3     100.0   100.0   60.8    20.8     9.2      100.0   100.0
Urban          44.2    18.4     10.0     17.0     11.7   41.4    12.6     5.6      16.0     10.9
Rural          68.6    31.0     17.6     83.0     88.3   64.5    22.3     9.9      84.0     89.1
Former estimates
All            62.2    24.3     12.4     100.0   100.0   59.5    19.6     8.4      100.0   100.0
Urban          44.1    15.3     7.3      17.0     12.1   36.7    11.3     4.9      16.0     9.9
Rural          65.9    26.1     13.4     83.0     87.9   63.9    21.2     9.1      84.0     90.1
Source: Author's calculation using 2005 and 2007/08 surveys
1.21    We view the poverty trends with another caveat. Although the incidence of
poverty is estimated to have decreased in 2007/08 the absolute numbers of the poor
have not changed. Accounting for population growth, we estimate that there were
8.159 million poor in 2007/08 compared to 8.142 million in 2005. Thus, the rapid
increase in population growth essentially cancels out gains made in poverty reduction:
the number of poor has not declined in Niger. Second, we use dominance curves
normalized by the poverty line to compare the two distributions of per capital
expenditure. Figure 1.4 indicates that these two distributions cross and there is no strict
dominance. Thus, at 1, per capita expenditure is equal to the poverty line. In fact, by
increasing the poverty line by only 15 percent, the poverty rate would have increased in
2007/08 relative to 2005.
-8-


Figure 1.4: Dominance curve in 2005-2007/08
Dominance curve 2005 - 2007 / 08 - National
O -
O-5                  1           1.-52
Normalized per capita expenditure
-     Year 2005 -     Year_2007
Source: Author's calculation using 2005 and 2007/08 surveys
1.22  So far we have presented the evolution of poverty between 2005 and 2007 from
survey data collected through the QUIBB and ENBC surveys. But as we stated above,
Niger has conducted additional surveys (ECVAM) between 2007 and 2010, and a natural
question would be whether we could learn of any changes to poverty since 2007.
Unfortunately, establishing comparability between ECVAM and the two prior surveys is
even more difficult. Therefore, we make no attempt to arrive at comparability between
ECVAM and the QUIBB and ENBC surveys. However, we can still use the ECVAM surveys
to say how poverty may have evolved since 2007. The first step is to make ECVAM
series as comparable as possible over time. As is evident from Table 1.2, comparability
of the ECVAM series itself is questionable due to differences in questionnaire design,
sampling, and seasonality. In particular, the 2006 data is clearly not comparable to the
rest, so we do not use it in any of the analyses. Moreover, the 2010 data is probably
different from the 2007 and 2008 surveys because it was conducted at a month that
was too close to the "lean" season. So one effort was to make sure that the
consumption baskets were common and rural and urban areas are defined uniformly
across surveys. The second step is to obtain a trend building on the QUIBB and ENBC.
To do so, we take as a given that the more credible data was collected from the ENBC in
2007 for a number of reasons. First, the consumption modules were more complete.
Second, the survey was done throughout the year, so problems of seasonality do not
arise. Third, 2007 (unlike 2005) was a normal year, as defined by average rainfall and by
consequence, grain harvest. Therefore, we assume that the poverty rates estimated in
2007 using ENBC are more likely to be closer to the true living conditions in 2007.
1.23  To obtain the trends since 2007 using ECVAM surveys, we first calibrated the
poverty rates using the ECVAM 2007 data to be exactly the same as the rate obtained
using ENBC 2007. We then use the implied poverty line in ECVAM 2007, adjusted for
inflation, to obtain poverty rates in 2008 and 2010. We use this simple back of the
envelope method to focus solely on trends in poverty since 2007. Table 1.5 shows the
poverty outcomes using this simple procedure. Three observations are worth noting.
First, ECVAM surveys appear to have oversampled the poor areas in some regions. This
-9-


would seem to be the case for Niamey for instance, and possibly Agadez. Therefore, it is
important not to read too much into the poverty ratios for these regions. Furthermore,
remember that our main interest in introducing the ECVAM survey data is not to
establish the accuracy of the poverty ratios, but to gauge the pattern of change in
poverty, assuming a starting level of poverty in 2007. With this in mind, the second,
observation is that poverty appears to have declined in 2008 and then rose in 2010. The
decline in 2008 could be explained by the exceptionally good grain harvest in 2008,
which was 25 percent higher than the average in the past 20 years. The 2008 is second
only to the 2010 harvest, which stands so far as the best in the last 20 years, and 41
percent higher than the average in the last 5 years (Reuters, Jan 4, 2011). Similarly, the
rise in 2010 could be explained by the very severe drought of 2009, which affected half
the population. Finally, while the decrease in poverty was widespread in 2008, the rise
in poverty thereafter was far worse in rural areas than in urban areas. Urban poverty
rose but not to the rates in 2007, while rural poverty rose to a rate that is higher than
the 2007 ratio. For the rest of the chapter we focus on the data from QUIBB and ENBC.
Table 1.5 : Poverty by Geographic Regions
Poverty Headcount     Distribution of the     Distribution of
Rate                  Poor               Population
2007   2008  2010     2007   2008  2010     2007  2008   2010
Area of residence
Urban             43.5   31.1  38.9     11.4   7.4    8.4     15.9   11.8  13.6
Rural             64.1   52.3  66.4     88.6   92.6  91.6     84.1  88.2   86.4
Region
Agadez            51.6   26.1  35.2      0.7   0.9    0.7     0.8    1.8    1.2
Diffa             46.6   41.7  51.5     2.6    2.8    2.6     3.4    3.4    3.2
Dosso             67.1   57.7  56.2     16.4   17.0  12.4     14.9   14.7  13.8
Maradi            63.6   52.5  78.1     23.1   23.1  25.4     22.1   21.9  20.4
Tahoua            51.8   46.3  64.8     16.5   17.9  19.2     19.3   19.3  18.5
Tillab6ry         73.7   60.4  72.0     23.3   22.2  19.8     19.2   18.3  17.2
Zinder            52.0   38.6  51.0     17.3   15.9  16.5     20.2   20.5  20.2
Niamey            71.4   36.4  40.1     0.2    0.1    3.5     0.2    0.2    5.5
Total             60.8   49.8  62.7     100.0  100.0  100.0   100.0  100.0  100.0
Source: World Bank staff estimates using ECVAM survey data
D.     Poverty, Growth, and Inequality
1.24   Poverty trends are dependent upon the rate of economic growth and the
manner in which the growth is distributed between groups in the population, e.g.
between the non-poor and the poor. In Niger, we would expect that economic growth
resulting from increased returns to the agricultural sector would disproportionately
benefit the poor because that is where the majority seek their livelihoods. Gains in
agriculture may result from pro-agricultural policies, including increased agricultural
-10-


investment, favorable climatic conditions, and changes in returns to factors of
production.
1.25  Both the poverty measures and the dominance curves presented above suggest
that economic growth which occurred between 2005 and 2007/08 also reduced
inequality.  The dominance curves show that the largest change in per capita
consumption occurred below the poverty line. Above the poverty line there is little to
no change in per capita consumption. This visual inspection is reinforced by the changes
in the poverty gap and squared poverty gap measures presented in Table 1.4 above. The
poverty gap index which indicates the mean income shortfall required to reach the
poverty line may reflect better average changes in living standards among the poor. In
2007/08 the poverty gap decreased by a little over 8 percentage points. The squared
poverty gap which measures the level of inequality among the poor also dropped
precipitously. It decreased by 7 percentage points over the period.
1.26  In Table 1.6, we examine how per capita expenditure is distributed across
income deciles in 2005 and 2007/08. According to the two surveys, mean per capita
expenditure has decreased by 2.2 percent between 2005 and 2007/08; but at the
median, there is a huge increase of nearly 14 percent of per capita expenditure. The
decrease in average per capita expenditures in 2007/08 suggests that the country has
grown poorer, but the distribution of consumption shows that the changes have
benefited the poor groups. For example, in 2005 the poorest 30 percent of the
population shared 9.3 percent of the total household consumption and the wealthiest
30 percent shared 63.3 percent; in 2007/08 the poorest 30 percent of the population's
share of consumption increased to 13.7 percent and the wealthiest 30 percent share of
consumption decreased to 54.8 percent.
Table 1.6 : Distribution of Per Capita Expenditure, 2005 and 2007/08
Per capita expenditure         Share of expenditure by decile
Mean    Median   1    2   3   4    5   6    7    8     9     10
2005     175530   117573  1.9 3.2 4.2 5.1 6.1 7.4 9.0    11.1  14.9  37.1
2007/08  171713   133825  3.4 4.8 5.5 6.5 7.3 8.5 9.7    11.5  14.9  28.4
Source: World Bank staff estimates using 2005 and 2007/08 surveys
1.27  We decompose the changes in poverty between 2005 and 2007/08 to identify
the shares we can attribute to economic growth holding inequality constant and to
income redistribution holding economic growth constant (Datt & Ravallion, 1991).
Table 1.7 decomposes poverty into growth and redistribution effects for national, urban
and rural areas and for each poverty measure. At the national level, both growth and
redistribution have contributed to the reduction of poverty, but the contribution from
redistribution seems more important. For example, of the 3.7 percentage point
decrease in the poverty headcount, only one percentage point is due to growth.
Economic growth has climbed slowly during the period, and thus most of the reduction
in poverty can be attributed to the redistribution effect.
-11-


1.28  The results of the decomposition patterns differ for urban and rural areas.
According to the two household surveys, growth has been negative in urban areas and it
could have led to an increase in poverty; but the changes in inequality which benefited
the poor disproportionately have offset the growth effect, resulting in net poverty
reduction. In contrast, both growth and redistribution effects move in the same
direction in rural areas, but growth dominates the redistribution effect in poverty
reduction. Overall, the redistribution effect has had the greatest impact on reducing the
severity of poverty in all areas.
-12-


Table 1.7: Poverty Decomposition, 2005 and 2007/08
2005     2007/08     Total  Growth  Distribution  Residual
variation   Effect     effect  Effect
National
Poverty headcount  64.487   60.761     -3.726    -1.080     -2.646      0
Poverty gap        28.835   20.778     -8.057    -0.751     -7.307      0
Squared poverty gap  16.286  9.243     -7.043    -0.479     -6.564      0
Urban
Poverty headcount  44.16    41.352     -2.808    4.008      -6.815      0
Poverty gap        18.415    12.643    -5.772    1.645      -7.417      0
Squared poverty gap  10.001  5.568     -4.433    0.944      -5.377      0
Rural
Poverty headcount  68.579   64.47      -4.109    -3.271     -0.838      0
Poverty gap        30.933   22.333     -8.600    -2.200     -6.400      0
Squared poverty gap  17.552  9.945     -7.606    -1.428     -6.178      0
Source: Author's calculation using 2005 and 2007/08 surveys
1.29   Both our estimates and the estimates obtained by INS suggest that poverty has
declined, but the size and the spatial distribution of the changes differ. According to our
estimates, poverty has dropped from 64.4 percent in 2005 to 60.8 percent in 2007/08,
or by 3.6 percentage points. This is more than the previous estimated decrease of 2.7
percentage points made by INS. If at the national level the estimated changes are close,
the story is different at the urban/rural levels. Our estimates imply that poverty has
dropped from 2.8 percentage points in urban areas and 4.1 percentage points in rural
areas. By comparison, INS estimated a 7.5 percentage point drop in urban areas and
only a 2 percentage point in rural areas. The fall in the headcount ratio is confirmed by
the decline of the other poverty indicators. At the national level for example, the
poverty gap and the squared poverty gap seem to fall sharply, which means that the
income of the poor, on average, has grown closer to the poverty line and that inequality
among the poor has declined, and both results confirm that the poor are better-off.
1.30  Overall, agro-ecological conditions during the 2005-2007/08 survey periods offer
important insights on the pattern of growth and resulting trends in poverty reduction.
The 2005 QUIBB data were collected from April -mid July 2005. The source of income
generating consumption for most rural households would correspond to the preceding
agricultural rainy season, June-September, 2004. During this period, Niger experienced a
severe drought which caused large shortfalls in domestic grain production and spikes in
the purchase price of grains in domestic markets. Thus, rural producer purchasing
power would have been dramatically affected by both a negative income shock and a
positive retail price shock. Since consumption of grains represents approximately 70
percent of the rural diet, there would have been limited prospects for substitution to
other commodities. Thus, we would expect to find a large decline in consumption
measured during this post-harvest period. In particular, the months of May to August
which precede the normal harvest season are considered the most lean-even following
normal production years -- since this is when harvest stocks from the preceding
production year for the majority of rural producers have been depleted.
-13-


1.31  Following the severe drought experienced in 2004, rainfall increased during the
subsequent agricultural rainy season and production returned to normal patterns. The
2005-2006 harvest was considered satisfactory and agricultural production increased
20.2 percent relative to 2004. This sector achieved strong growth rates of 11.6 percent
and 7.7 percent, respectively, in 2005 and 2006. During the corresponding period of
ENBC data collection, rural consumption expenditures would have depended largely on
the proceeds from the 2007 rainy season harvest. Agricultural growth in 2007 fell to
only one percent. Although performance was weak it was not affected by drought.
Agricultural growth is likely to have produced upstream and downstream effects in off-
farm production activities. Thus, agricultural growth could positively affect growth in
urban areas, but the largest impact will occur in rural areas where most of the poor live.
Table 1.8 : Inequality indicators, 2005 and 2007/08
2005                           2007/08
p90/plO   Gini    Atkinson 1    p90/plo    Gini    Atkinson 1
National      7.2     48.0       33.0          4.5     36.6      19.7
Urban         10.4    53.4       39.9          6.3     42.7      26.2
Rural         6.3     43.2       27.7          4.1     32.9      16.1
Source: Author's calculation using 2005 and 2007/08 surveys
1.32  The previous developments tend to show that the recent period has been pro-
poor since the poor have benefitted from a greater share of growth bounce-back than
the non-poor. This pattern is consistent with the fall in inequality. The most widely
used inequality indicator, the Gini index, fell from 48 percent to 36.6 percent at the
national level. Average per capita consumption ratio from the 10th richest to the 10th
poorest was 7.2 in 2005 and it dropped to 4.5 in 2007/08. All of the inequality indices
presented above in Table 1.8 have fallen across national, urban, and rural areas.
However, while the decrease in inequality is promising, it is difficult to believe that the
magnitude of the inequality reduction is also not responsive to artifacts of survey
methodologies. When looking at the estimates, it is important to note that despite our
efforts to achieve comparability, there remains a lingering voice for caution. In
particular, although we have tried to harmonize the consumption baskets and correct
for the difference in recall periods, the differences in level of detail between the two
baskets will have an impact on consumption estimates.
E.    Sources of Change in the Distribution of Economic Welfare
1.33  In this section we analyze the micro determinants of poverty in 2007/08 and we
try to determine the sources of change in distribution of economic welfare between
2005 and 2007/08.
Geography, Natural Resources and Poverty
1.34  Poverty is pervasive in Niger but it dominates the lives of rural households. A
little more than 65 percent of rural households are poor compared to a little over 41
-14-


percent of urban households - and most of the population, more precisely 84 percent,
lives in rural areas, which means that almost 9 out of 10 of all the poor people live in
rural areas. Thus, it is important to understand the rural dimensions of poverty and how
the variance in regional attributes drives poverty.
1.35   Figure 1.5 plots regional dominance curves to rank the share of population by
poverty across regions. Although there is wide variation in the incidence of regional
poverty, one can discern three clusters of regions. There are the low poverty regions,
comprising Agadez', Diffa and Niamey. The regional share of poverty in Niamey tends to
be more diffuse and thus to exhibit greater inequality than other regions concentrated
in the first group. This is followed by the moderately poor region of Zinder with a
poverty headcount of 46.3 percent. Finally, we have the high poverty regions, which
include the remaining four regions (Tahoua, Dosso, Maradi, Tillabdri), where in each,
two-thirds of the population is poor, and whose poverty rates exceed Zinder's by almost
19 percentage points.
Figure 1.5 : Dominance Curve by Region in 2007/08
Dominance curve 2005 - 2007 /08 - by Region
0-
0             .5            1             1.5           2
Normalized per capita expenditure
Agadez          Diffa
Dosso           Maradi
Tahoua          Tillaberi
Zinder          Niamey
1.36  We explore the spatial dimension of poverty by examining how resource
endowments contribute to household welfare. The availability and use of water assets
in Niger is an important indicator of its current and potential agricultural performance.
Untapped reserves suggest the potential for new agricultural technologies and
improvement in returns to the sector. But the underlying risk factor in the Sahel - and in
particular, Niger- is the level and variation of expected annual rainfall. Cereal producers
depend on rainfall during the main cropping season, May - October, for the production
of millet, sorghum, and maize which account for a substantial portion of the Nigerien
food basket. Similarly, herders depend on the annual rainfall for the production of
1 The ENBC measured only urban poverty in Agadez because of the risk of regional insecurity. Thus,
poverty is likely to be under-estimated in the region.
-15-


livestock forage. Thus, expected distribution of rainfall across time and space may be
considered as an important resource endowment that will differentially affect
household welfare across these dimensions.
1.37   Figures 1.6 - 1.9 present the spatial distribution of poverty and water assets to
explore their underlying relationships. The poverty impacts of agro-climatic shock and
the capacity to adapt (ex ante) or cope (ex post) depends on having access to the means
to income diversification. Past research has indicated that diversification of economic
activities aimed at minimizing weather and environmental risk has led to different
income strategies across agro-ecological Sahelien zones.  Differential access to
household income diversification either by geography or by initial state of poverty has
the potential to affect inequality levels in the overall distribution. Some differences in
access across households could be driven by differences in capacity to make
2
investments in non-farm assets or to participate effectively in low-capital activities.
1.38  The maps, shown below in Figures 1.6 and 1.7, depict annual average rainfall and
annual variation (measured using coefficient of variation) of rainfall based on 30 years of
rainfall data collected through the ground-based rain gauge network. The maps were
estimated at the commune level using data only from rain gauges that covered the 30-
year time frame.    Commune-level values based on the geographic boundaries
delineated by the Ministry of Hydraulics and the Environment were then interpolated
using an inverse distance weighting procedure which also accounted for other
parameters such as radius and constraints used at each prediction location.
1.39   As one might expect both the level and stability of rainfall increases as one
moves in a southward direction. Gaya, the southern-most department of the Tillab6ri
region receives the highest annual average rainfall of about 800 millimeters. Southern-
most departments in Maradi Region receive approximately 600-700 millimeters of
annual rainfall. We would expect that these rainfall patterns correlate with spatial
differences in (rain-fed) agricultural productivity. Thus, areas benefiting most from
relatively high and stable rainfall levels should exhibit higher agricultural productivity
and lower poverty.
2 For an early discussion on the link between geography and diversification see Reardon et al, 1992. Later
work further explores the impact of diversification on income inequality (Reardon and Taylor, 1996).
Reardon et al, 2000, review the literature and household survey evidence from Africa, Asia, and Latin
America to discuss how evidence is mixed on non-farm diversification effects on rural income inequality.
See Ali, et al. "Rainfall Estimation in the Sahel. Part 1: Error Function", November, 2005, Journal of
Applied Meteorology, November 2005, for a discussion on evaluating the quality of rainfall estimates
through determining objective error functions.
-16-


Figure 1.6: Commune-level Average Annual Rainfall, 1980 - 2009
Mean annual total precipitation (mm)
10 50
1 51 - 100
fl101 - 150
[_] 151 - 200
_ 201 -,q0
g 301 - 400
a> 400               -
Source: World Bank Estimates based on GON Department of Meteorology rain station
observations and Ministry of Hydraulics and the Environment commune-level
boundaries
Figure 1.7: Commune-level Average Annual Rainfall Coefficient of Variation, 1980- 2009
CV of total annual precipitation (%)
[ 0- 0.25
F_1026-O4
'MP.41 -0.6
ggoel oa
g 101 -
AA
A A
\ A
A
Source: World Bank Estimates based on GON Department of Meteorology rain station
observations and Ministry of Hydraulics and the Environment commune-level
boundaries
-17-


Figure 1.8: Department-Level Poverty Incidence, 2005
0 .-0.05
0.05 - 0.1
0.1 -0.15
0.55.0.6
0.3-0.35
 0.35 -0.4
- 0.55 - 0.6
0.6 - 0.65
0.65 - 0.7
0.7 -0.75
JQiM1              0.75 . 0.8
0.8 -0.95
095-1
0.as5 .
Source: World Bank Staff Estimates based on QUIBB, 2005 and Niger Census, 2001
Figure 1.91.   Canton -Level Poverty Incidence, 2005
m__ 0.,0,05
0.55-0.1
0.6-0.6S
I0.19 -0.7
0.7 - 0.75
O-z0 8 0.5
-1 0.5 - 0.6
OA 0 0.9
O.0S0 - 0.5
small area poverty estimation conducted with the 2001 household census and the 2005
QUIBB survey data disaggregates poverty estimates (canton-level) at about the same
level of disaggregation as the rainfall data (commune-level.) In contrast to what we
would expect the southern-most areas possess some of the highest levels of poverty.
-18-


Cantons located in the southern area of Tillabdri have upwards of 75 percent poverty
rate; in the Maradi region, a group of southern cantons are estimated to have upwards
of 85 percent poverty ratio. By comparison, poverty incidence of 45 percent in some
Northern Regions of the country, where annual rainfall is less than 100 millimeters per
year is considerably lower.
1.41  Thus, the key observation from Figures 1.6-1.9 is that while annual rainfall
affects domestic agricultural production, annual growth, and poverty incidence, there is
no strong correlation between long run average rainfall and spatial distribution of
poverty. There are possibly several hypotheses for this outcome but two stand out in
the context of Niger. One is the possibility that Nigeriens have adapted to rainfall risk
through migration or relocation away from northern areas where rainfall is low and
variance is high. The least able to adapt to rainfall risk in the North could have moved to
the South where rain-fed agriculture is possible. An alternative explanation is the
possibility that rather than migrate the population has changed the composition of
economic activity and adopted the one most suited to the risk levels of North and South.
1.42  Overall, Nigeriens have not pursued an out-migration strategy to minimize
rainfall risk. Nigeriens who do not engage in irrigated agriculture- the second season
after rain-fed grains are produced- may exit their communities in search of work. Most,
however, return in time for the next agricultural season and overall net migration has
been negligible.4 In Table 1.9, we present the regional population structure covering
more than twenty years. Over this period the population in Niger has more than
doubled but the population structure across regions has remained relatively stable. The
last column in the table highlights gains and losses in population between 1977 and
2001. The southern regions have experienced small decreases in their share of
population but these losses have been minor.
1.43  The stable population structure suggests that Nigeriens have not migrated out of
areas where the lowest rainfall occurs to pursue rain-fed agriculture. Population growth
based on initial conditions over the past 24 years in the south and the east has resulted
in increased population density and greater dependence on rain-fed agriculture through
extensification. By comparison, the population in the areas of lowest rainfall in Niger
has remained stable and -and most likely less poor-- by not relying upon rain-fed
agriculture as a livelihood strategy. The relative dominance of rain-fed agriculture as an
income strategy is likely to expose households to high levels of risk and low levels of
income. Thus, without conditioning on other factors rainfall may in fact correlate
positively with poverty.
4 See Guengant, JP, "Comment B6n6ficier du dividende d6mographique ? " La d6mographie au centre des
trajectoires de d6veloppement dans les pays de I'UEMOA Analyse pays Niger F6vrier, 2011. The author
reports UN estimates indicating that between 1960 -2010, only 6000 net departures resulting from
migration occurred annually.
-19-


Table 1.9: Regional Population Structure, 1977 - 2001
Structure                Structure               Structure
Region            1977      1977         1988        1988        2001        2001
Agadez          124,985     0.024      205,108       0.028     322,511       0.029
Diffa           167,389     0.033     186,792        0.026     347,534      0.031
Dosso           693,207     0.136    1,016,469       0.141    1,509,943      0.136
Maradi          949,747     0.186    1,385,170       0.192   2,241,805       0.202
Niamey         242,973      0.048     392,165        0.054     709,869      0.064
Tahoua          993,615     0.195    1,305,417       0.181    1,978,074      0.178
Tillabdri       928,849     0.182    1,322,025       0.183    1,894,634      0.171
Zinder        1,002,225     0.196    1,406,943       0.195    2,085,886      0.188
TOTAL         5,102,990     1.000    7,220,089       1.000   11,090,256      1.000
Source: INS
1.44   Although the population structure has changed little over 24 years, population
growth has been explosive-particularly in the poorest areas which are the most
densely populated. A significant portion of Niger's land mass is uninhabitable indicating
that there are wide disparities in population density across the country. The northern
and eastern regions contain substantial land areas where no human settlements exist
while the western and southern areas support the majority of the population. Thus,
approximately 90 percent of the country's population subsists on only one third of the
country's land. The Agadez region covers over 50 percent of the country's land mass
but it supports less than 3 percent of the country's population. Diffa accounts for a little
over 12 percent of the territory and only about 3 percent of national population. In
contrast, Maradi covers less than 4 percent of land but a little more than 20 percent of
the population lives in this region. Thus, population density varies widely across regions
in Niger and the build-up may suggest that the most densely populated areas are hitting
or approaching the upper limits of the carrying capacity of the land. s
1.45   Despite low annual rainfall, Niger possesses a water network- composed of both
surface water and ground water that may provide some opportunity for diversification
out of rain-fed agriculture. Figures 1.10-1.11 present these networks. The placement of
water reserves in Niger correspond to the well-known concentration of irrigated
agricultural crops such as rice in the Tillab6ri region and peppers in the Diffa region.
Figure 1.11 indicates that the southern regions of Maradi and Zinder have access to
underground water sources that are largely regarded as easily extractible depending
upon availability of irrigation infrastructure. These regions produce irrigated crops
during the second agricultural season, but this kind of irrigation-dependent production
may not be widely practiced.
s See World Bank (2005) for a more in-depth discussion on the impact of population growth and
agricultural land use.
-20-


Figure 1.10: Niger Surface Water Network
REPUBLIOQUE DU NIGER
Réseau lydrograpliique
In  deVI
Cnurs d'Eau P ipaur
Source: GON, Ministry of Hydraulics and the Environment
Figure 1.11: Niger Groundwater Network
REPUBLIQUE DU NIGER
Carte des systémes aquiféres
N
AIR          <     s
] k              r
Source: GON, Ministry of Hydraulics and the Environment
-21-


1.46  These water resources determine the potential for increasing opportunities for
households to participate in high-valued (irrigated) agriculture particularly among poor
households. The profitable utilization of irrigation technologies requires having access
to and being able to use water, but it also requires having human capital capacity to
manage the technology collectively with other producers and to establish
complementary institutions. Thus, both physical and human assets are required to
participate profitably in these promising agricultural production technologies but these
synergies are not necessarily common among the poorest communities. In the next
section we look at the links between socio-economic characteristics and poverty.
Socio-economic characteristics of poverty
1.47  In Niger, over the life cycle, the risk of being poor tends to be U-shaped, but
mostly this is the case in urban areas. In particular, the young and the elderly urban
residents have much higher propensities of falling into poverty than the middle-aged.
For instance, urban heads of households up to 20 years face a 40 percent chance of
falling into poverty. This risk is halved by the time such a person reaches 40 years, but
then it rises again thereafter. For rural areas, we find a steady rise in the risk of falling
into poverty throughout the lifecycle. Among rural households the risk of poverty
slopes steeply upwards to 65 percent for 40-year household heads and levels off to 60
percent for 60-year olds.
1.48  The steep positive poverty-age relationship may reflect the interplay of several
factors. Age may be indicative of a strong cohort effect in which the educational profile
of older household heads is likely to be extremely low compared with younger
household heads. Additionally, family size and dependency ratios in Niger are likely to
have grown faster than potential gains achieved from returns to employment or
accumulated wealth. Family size is an important factor weighing in on poverty. Rural
farm work is likely to contribute to demand for family size through endogenous demand
for farm labor. However as access to high-quality farm land deteriorates with
population growth, the marginal productivity of additional labor is likely to be small.
Other factors may influence family size decisions in rural areas where, on average, lower
investments in human capital are made in children and tradeoffs between the quality
and quantity of children have yet to be internalized among rural households. Rural
households, moreover, may be more sensitive to religious sentiments expressed against
controlling fertility and family size.
1.49  Poverty levels among female-headed households, for example, are less than
among male-headed households by about 7 percentage points. However, female
headed households are more likely to reside in urban areas and are on average more
than 35 percent smaller than male households. In a country where female educational
achievement is low, marriage is early, and fertility is high, women may possess relatively
few opportunities for economic security outside of marriage. However, marriage itself
may yield asymmetrical power relationships if women enter into unions with partners
-22-


who maintain control of economic assets leaving them with limited exit options in the
event of marital failure. While marriage contracts are supposedly protected under civil
law, they are often negotiated informally and they may be dissolved by male spouses
through simple public repudiation of their wives.6 These dissolutions may not lead to
any parsing of assets that were obtained during marriage or even joint custody
relationships of children. Thus, marital law as it is practiced in Niger engenders limited
protections for female partners. Women, therefore, are likely to confront economic
insecurity regardless of marital status.
1.50  Figure 1.12 plots female marital status and female poverty. We find that
relatively more female- headed households are likely to be poor if they reported being
divorced or separated. In particular, the state of being separated may suggest that a
woman's prior marriage ended without any formal decree-and, thus, without any
transfer of household assets to the female. Female heads of households who report a
polygamous union may be living separately with their children from the male spouse
who lives with another wife. In this situation it is possible that the female-headed
household shares the same level of welfare as the related male-headed household.
Figure 1.12: Poverty by female marital status
kemale Headed Household Poverty Incidence by
Marital Status
amous
Source: World Bank staff estimates using ENBC 2007
F.    Correlates of Poverty
1.51  So far the probabilities of falling into poverty by age, female headed status,
urban and rural residence and so on, have been raw estimates - meaning they have
been obtained without controlling for many other determinants of poverty status. In
See Cooper (1997), Marriage in Maradi, for an in-depth qualitative study on the evolution of intra-
household dynamics.
-23-


Table 1.10 below we estimate the consumption shortfall for households with certain
characteristics by regressing per capita consumption on household-level characteristics
and geographic indicators holding all the other determinants of poverty constant. This
provides the correlation between observed characteristics and household poverty. We
estimate linear regressions for the 2005 and 2007/08 household samples and separately
for urban and rural populations. The dependent variable is the logarithm of per capita
consumption and the determining variables can be grouped into five categories: (a)
household composition (number of young and school-age children, and adults and their
squared values); (b) demographic characteristics of the household head (gender of
household head, the age of the head and its square and whether the head has a
spouse); (c) human capital characteristics of the household head, (level of education
completed; employment status; whether the head works in the public sector); (d)
education and employment characteristics for the spouse of the household head; and
(e) region and urban/rural location dummies. We include variables that we consider
relatively comparable across the two surveys.
1.52  Household composition has a significant effect on the overall probability of
household poverty. In 2007, controlling for other variables, any household composition
category representing dependency on prime age adults, including all children below the
age of fifteen years and the elderly (65 years and older), has a negative impact on
consumption. On the other hand, the number of female and male adults does not affect
consumption probability. These results may indicate that composition of households,
not just its size, matters for welfare. Nevertheless, we cannot overlook the important
negative relationship between fertility outcomes and poverty. By reducing population
growth, Niger can achieve greater poverty reduction holding constant the level of
economic growth.
1.53  In 2005 the results show that female headed households had higher
consumption than male headed households, but this changed in 2007/08. The 2005
survey was implemented following a severe drought and food crisis. It is plausible that
female-headed households were more likely to have been targeted with food aid
following the crises. The 2007/08 results, which was a normal year without much aid
distribution, shows that female households are poorer than their male counterparts, all
else being equal.
1.54  As we expect, we find that the returns to education for both household heads
and their spouses are substantial, and these returns increase with each additional level
of schooling. Moreover, these effects are more pronounced when we focus on urban
households. Education has no poverty reduction effect on rural households. Given that
agriculture is practiced predominantly using low levels of inputs and technology in
settings where population densities have increased rapidly over time, the returns to
labor would be low. However, we expect that as the practice of agriculture becomes
more high-valued and successful producer strategies involve technology adoption,
education will have greater impact in the sector.
-24-


1.55     The gains from     a well-educated spouse are large, but below those for the head.
This effect could be attributed to the existence of gender discrimination in pay. The
large impact of education on per capita income and poverty justifies the
implementation of programs to promote education in the country where educational
achievement is low for both sexes and more so for girls living in rural areas. As the
demand for large families decrease over time, women may find greater incentives to
invest in the human capital of their children and their own market human capital.
1.56     As we have noted, employment variables have been measured with errors and
yield results which are not consistent. In general, we do find that households with a
head or spouse working in the public sector have higher per capita consumption. By
contrast, heads of households working in agriculture face a significantly larger
consumption shortfall and by implication higher probability of being poor.
1.57     Finally, controlling for other characteristics, geographic location has an impact
on per capita consumption. We find that differences exist across regions, but not across
urban and rural areas, at least in 2007/08. The absence of poverty variation across
urban and rural households once regions are taken into account suggest that localized
impoverishment, is probably more pronounced than rural and urban differences.
Table 1.10 : Correlates of Poverty
2005                                  2007
National       Urban         Rural    National       Urban        Rural
Household composition
N. of children below 5            -0.068***   -0.140***    -0.045**     -0.150***    -0.152***   -0.134***
N. of children below 5 squared    0.024***     0.041***    0.019***     0.015***     0.014**     0.012***
N. of children between 5 & 14     -0.173***   -0.197***    -0.156***    -0.172***    -0.211***   -0.144***
N. of children between 5 & 14 squared  0.017***  0.020***  0.016***     0.014***     0.018***    0.011***
N. of male adult                  -0.041**    -0.009       -0.058**     -0.028       -0.007      -0.058*
N. of male adult squared          0.007***     0.004       0.010        0.001        -0.001      0.005
N. of female adult                -0.029      -0.055       -0.005       0.023        0.049       -0.005
N. of female adult squared        0.007**      0.011**     0.005        0.001        -0.003      0.005
N. of people age 65 or more       0.077        0.191**     0.034        -0.145***    -0.209**    -0.098*
N. of people age 65 or more squared  0.085***  0.052       0.103**      0.039        0.072       0.016
Head demographic characteristics
Head is female (yes)              0.542***     0.459***    0.616***     -0.255***    -0.227***   -0.293***
Age of head                       0.029***     0.011        0.030***    -0.002       0.002        -0.005
Age of head squared               0.0003***    -0.0001      0.0003***   0.000        -0.000       0.000
Spouse in the hh (yes)            -0.112***    -0.254***   -0.011       -0.159***    -0.132       -0.195***
Head education
No education
Primary                           0.570***     0.591***    0.540***     0.082***     0.112***    0.019
Secondary 1                       0.643***     0.653***    0.606***     0.238***     0.263***    0.135**
-25-


2005                               2007
National      Urban        Rural   National      Urban        Rural
Secondary 2 and more              0.974***    0.983***    0.789***    0.560***   0.592***    0.318***
Head Activity
Has a job (yes)                   1.061***    1.107***    1.027***     -0.041     -0.140**      0.095
Head sector of activity
Own enterprise
Public administrat./corporation      0.057       0.055     0.221**    0.089**     0.093**      0.180*
Private enterprise                0.099***    0.109***    0.092***     -0.002      -0.029       0.123
Head type of industry
Non agricultural (yes)            0.107***    0.199***    0.077***      0.036     0.095**       0.001
Spouse education
No education
Primary                            0.069**     0.112**       0.044     0.053*       0.058       0.032
Secondary l and more              0.240***    0.288***       0.115    0.243***   0.261***       0.086
Spouse Activity
Has a job (yes)                      0.026      -0.038     0.076**   -0.062***   -0.159**      -0.024
Spouse sector of activity
Own enterprise
Public administration/corporation    0.081       0.156      -0.233    0.243***   0.217***       0.222
Private enterprise                   0.022   0.0518737      -0.009     -0.149      -0.145      -0.586
Spouse type of industry
Non agricultural (yes)               0.032       0.118       0.055    0.069**      0.148*       0.044
Area of residence
Rural (yes)                         -0.030                              0.006
Region
Niamey
Agadez                           -0.391***   -0.461***       0.029    0.408***   0.431***    (dropped)
Diffa                             0.124***      -0.103    0.568***    0.235***      0.056   (dropped)
Dosso                            -0.344***   -0.230***   (dropped)   -0.207***   -0.111**   -0.631***
Maradi                           -0.633***   -0.409***   -0.303***   -0.109***     -0.007   -0.598***
Tahoua                           -0.210***   -0.172***    0.163***   -0.130***  -0.127***    -0.502***
Tillab6ri                        -0.441***   -0.402***    -0.079**   -0.259***  -0.240***    -0.655***
Zinder                            -0.533***  -0.563***   -0.149***      0.031       0.070   -0.399***
Constant                         10.633***   11.062***   10.067***   12.749***  12.646***   13.136***
Statistics
R 2                                 0.4929      0.5533      0.3801     0.4231      0.4257      0.3425
N. of observations                   6690        2020        4670        4000        1916       2084
Source: World Bank staff estimates using 2005 and 2007/08 surveys
-26-


Chapter 2: High Vulnerability and Lack of
Resilience
2.1   As we have seen in chapter 1, poverty rates in Niger are high, and while they
have changed relatively little over the last 5 years of the 2000s, they have also exhibited
substantial swings from time to time. These observations are not inconsistent. An
explanation for why this is the case is the subject of the following chapters. Chapter 2
will explore the role that risks, especially uninsured risks, play in perpetuating poverty.
The rest of the chapters will look at other potential explanations for high levels of
poverty. Chapter 3 looks at unequal opportunities, especially in the provision of
services, while chapter 4 looks at determinants of agricultural productivity. Many
governments, including Niger's play a large role in the provision of services and often
proclaim to equalize opportunities for all the population using these services. The
Government of Niger also spends substantial resources in boosting rural productivity
and incomes. Our examination of determinants and outcomes in opportunities in
services and agriculture will start with outcomes at the household level, but the
systematic differences, indirectly implicate inadequacies of government policies.
2.2    Niger experiences numerous shocks, which strike frequently, sometimes with
catastrophic results. These shocks are driven by various natural, economic, and political
events. As an example, drought shocks in 2009 alone rendered about 2.7 million people
food insecure in 2010. Around the same time, the price of cowpeas, which is a major
source of protein for the poor, rose by between 60 to 170 percent above the nominal 5-
year average in many markets (FEWS Net, February 2010). Then in February 2010, after
a period of social tension, a coup d'6tat took place which temporarily suspended
international aid to the country. Finally, in July/August 2010, the Niger River rose to its
highest level in 80 years, and floods destroyed most of the adjoining rice fields,
remaining food reserves, and roads used to deliver food aid. This compounded the
already precarious food security situation. Meanwhile, health shocks, in particular
cholera arising from the floods broke out in the flooded regions.
2.3   There are many sources for the risks faced by Nigerien households. Some, like
weather and pest shocks, come as part of the country's geographic location and are
therefore completely exogenous. The same could be said about some of the
macroeconomic shocks such as price and financial instability. Yet, others such as social
conflict, to the extent that they are enabled by fragile social cohesion, are internally
induced. Moreover, whether shocks are exogenous or not, the depth of the suffering
they cause has a lot to do with the lack of resiliency of the country, where resiliency is
used to mean a robust mix of coping strategies ranging from mutual insurance, formal
insurance markets, and social protection programs.
2.4   In this chapter we look at the pattern and distribution of some of the major
shocks that affect Nigerien households. The first part of the chapter will be descriptive.
-27-


It will present the types of shocks that Nigerien households face, focusing on the events
in the past 5 years. The descriptive overview will look at how widespread the shocks
were and who was affected by them. Most of the evidence for this section will rely on
households' self-reported responses. Given that self-reporting tends to suffer from
recall error, we also make use of purposeful surveys such as price or rainfall data
collected by autonomous agencies to gauge the severity of these shocks. This is
followed by an examination of the welfare consequences of key shocks. In addition to
income or consumption, we look at reduced nutrition of children. We use this
information on welfare costs to infer the scope of vulnerability, and then conclude by
looking at the characteristics of the vulnerable, paying special attention to the chronic
poor.
B.    Drought and Price Shocks Dominate
2.5   The types of shocks reported by households can be obtained from two survey
series: the National Consumption and Expenditure Survey (or ENBC, French acronym)
conducted in 2007 and Household Consumption and Food Vulnerability Survey (or
ECVAM, French acronym) conducted each year from 2007-2010. Figures 2.1-2.3 provide
a list of the shocks that affected households between 2007 and 2009. As the graphs
show, there is a lot of overlap between the shocks asked in different surveys. The
reported shocks can be grouped into broad categories such as weather, natural disaster,
price, labor market, crime, and health. Weather shocks include droughts or irregular
rains and floods which we distinguish from natural disasters comprising fires and pests
and animal diseases. Price shocks also include input price hikes. Labor market shocks are
reported as job loss and unemployment, while health shocks include illnesses/disease or
exceptional health expenses and death of a member of the household. Finally, crime
shocks are captured by reports of insecurity or theft.
2.6   In the last 3 years, weather and price shocks were the most commonly
experienced by households. Both surveys asked households to report and rank the
shocks or difficulties they experienced during the preceding twelve months. In 2007,
most of the respondents appear to have chosen the manifestation of a drought, "poor
harvests" as the most severe shock. The most common sources of poor harvests in
Niger are drought, irregular rainfall, floods, and pest (especially locust) attack. During
this period, there were no floods or locusts, so that leaves the most likely causes as
drought and irregular rainfall. If we treat people who chose poor harvests and drought
as essentially reporting the same shock, then almost 35 percent of all households would
have ranked weather shocks at the top of the list in 2007. The fraction rises to over 40
percent if we add those who ranked the same shock as the second biggest. In the same
year, almost 15 percent of households ranked price shocks as their greatest shock, while
an additional 8 percent ranked it the second most severe shock that they faced. This
means that almost a quarter of the households ranked price shocks as first and second
in their rankings. These two shocks dominate the experiences of the households if we
accept that "lack of money" - ranked second in the graph - is probably a condition that
is brought about by the combination of multiple shocks such as drought (harvest
-28-


failure), price, and other shocks like health. A reasonable fraction of households also
report "drop in household income" as an important shock in 2008 and 2010. This is most
likely resulting from low rainfall/weak harvest, depending on how respondents see the
chain of events and ultimate impact.
2.7     In both 2008 and 2010, price shocks were identified by households as the most
prevalent, which is a reversal from 2007. Figure 2.2 shows that in both 2008 and 2010
almost two out of three households ranked price shocks as the most or second most
serious in their list in the preceding 12 months. In the same period, far fewer
households ranked weather shocks at the top. In fact in 2008, health shocks - illness or
exceptional health expenses - and drop in income were reported as more prevalent.
Weather shocks will begin to matter again in 2009, but still only about 15 percent
ranked irregular rains at the top of the list in the 2010 survey, suggesting that despite
the major drought households perceived the rise in food prices as more serious.
2.8     Negative effects of price shocks are stronger in urban areas. Although the
preponderance of weather and price shocks are pretty clear during the three years, the
distribution of these shocks in rural and urban areas was different. In 2007, almost 50
percent of rural households ranked weather shocks at the top of their list, compared to
less than 15 percent who reported the same in urban areas. By contrast, the price
shocks in 2008 and 2010 affected proportionately more urban households. In 2008,
over 50 percent of urban households ranked food price increase as the greatest shock
they suffered that year, but the rural fraction who reported the same was about 38
percent.    By 2010, the converse was observed, that is, a greater share of rural
households (39 percent) than urban households (27 percent) viewed food price
increases as the greatest shock, suggesting that the movement of domestic prices was
creating an additional strain on rural farmers and pastoralists already negatively
impacted by the 2009 drought.
Figure 2.1: Household perceptions of shocks and difficulties during the last 12 months
2007 - National                 2008 - National                2010 - National
Poor haest                       Food prie icrease              Food priceincreae
LackofoneyFodPo                      t~ft
Food pric,inrcsc                 Othersock/diffilct               Irreguarraot
Lackof.ter                       Dropin income                   Dropiom e  I
Ir,u, illnes/exceptionalealtepeses                                  Othr
Foodm ilal                          rregula.rin m         Iln,es /excep.onalelth expees  M
Marketacce  4                  Derthoffamily member            Dethoffamily member  I
Drought                          Ins urty/thft                    Jobs
t    adt  i                Natrlr asrophe(ood, f re, wind)  ANtural catatroph  (lood fire, wind)  I
Lackofanimal                    Longterm uremproymendt  1        Debtrm u   I
Flood                     Agricuraliput prireincrease       iecuril  hef I
roor y                       Detreimbursemert  I       Agricltualinput priceincrese I
Lackofpsture/ fodder  reatestdiffil       Job lo t Gr t difficulty  Longterm unemploymnrt
Livestockdises k d             Fuel/trasportpriceinc se  r r        Poweou  e "Greaetediffira
P. 2nd greate  difficlty       Poweroutage ndgreatestdiffiul   Fel/trasport priceincrease  2ndgreate tdifficuly
Lackof milk                     Rertlpriceincrease             Rentalprice ircrease
0   0  20  30  40  o50              0  10  20  30  40  50  60  70  0  10  20  30  40  50  60  70
Shref hou holds (%)              Shrea fhosehold( %)           Shr of huseholds (%)
Source: ENBC (2007)
-29-


ÿþFigure 2.2: Rural household perceptions of shocks and difficulties during Iast 12 months
2007 - Rural                                         2008-Rural                                         2010-Rural
0000 W  020000                                      1000 O2~~00~00
b 0k 0f o o ,oy
0000 P0~~000                                          Otflo,,,ookî d,ffooIw
LkOfOOOOO2  S                                         0,07,, 02000
S000.,0, 1,  o, 000000                           Il,,,, foroot,  IflooIîh  po,,,oo                                     0,flo,
DootehI
Food,o,,I,b,I,to
M,,k,t ,~~ooo                                    Othoffoor horbo,                                    OtfltfftO,t0000000
100k 0f ,,~I I,bo,                                         ~rrowltfloft                                            100
0000                                N,to,,I t opho(flood f~4,0)                         0,r,,It,,ooflofflood fiorool
0,0,00 00o,o,,t                                   ogto,,,,,pIoo,o,,,t
[o,, 0f 00102, h 02ohoIO ,d.,It                        floorrutol or,, poo,,oo,,                                    Io,o,.,, wfooO
100k 0f 00,1,,, floodo,  *                                0,01  jooflo,  0,0to
0,,,t~0 d,OOo,Ito          72,fî,0,îo,1 000t100t0     *0,,d 1,0,10,, ifitO     011t020000,,
1,0001,,, k I    *0,,, g,,,0,,t d,lt,0,100            00,1,4,                                             0001040         *000g,,,t,,t OfOtooltO
0,000017
0,0                                          70,11,100,0,0,0,,,                           0g,jOOIt202I,,O0
O  10 20 30 40 10 10 70                                0 10 20 30 40 10 10 70                              O 10 20 10 40 10 10 70
Sha,e othoos,ttold,l%l                                 Sh.,.of hooholdo lU)                               Sfl.,.of ho, Ofloldo (X)
Source: World Bank estimates using ECVAM (2007) and ECVAM (2008 and 2010)
Figure 2.3: Urban household perceptions of shocks and difficulties during Iast 12 months
r                  2007-Urban                                          2008-Urban                                          2010-Urban
Lackofmonbo                                       boa p,r,,,o,oo,,                                    1000 ooooooo
U,,omploo,o,nr  s                        Il, Ofo ,p1,,t,Ifl0oloott,,0,
SetjOOO,I000SOt 000401                                            Ooooî                                    Otfloff,,,I0,oo,,0o,
M,,kot ,oooso                                       I,,ooo,,,ty!ofloft                      IIoo,f o,oop,io,,,
F004 aoailah,I,to                               000110110,010 0020001                                           01000
[0,000001,0 hoo,,flold adn,                                     lîtogolor o, o,                                          On
100002010                                              200100,                                         OOo,o, 0~tog0
LaokOiaom4lO                                   10,01 tt00000t01000t00t                      f010 10,0 000101f000, fi,, 0000)
F002                                 Oîooolîo,aî opor 0~10o ,0,o,,o                         rO,gbo,00000,,oOO,oo,o
00021010
For,                                  FoIft,,,,poîo pOo0tOo~oOoo                        Og,oolto,Ooîo,
LaokofpaSoo,0 ffodde,   Oroatoso OfliCoto                   pO,0o,Ooî,gO                                       îo.~17fî,fî        0,0,1,01 Ofli0oW
flloîkod,Oad(to,oa,k,Oo,..  .200 gocatoît 0710,1v  u,î,,,îo,,î,opflîîîood f,,,,o,d)                           o,,,, p,io,i,o,,,,o  *000 4,00010770,11
1,0101,0k                                       10,1,1 0O00tl0~0000                            toolhaoopoo o,o,i,o,,,oo
O   10   20  10   40   10                            O 1020004010007000                                  O   10  20   30   40  10
Share If hous,hotds (10)                             Sh.,o,t h oooh,Id, (fi>                             SIO,0 If 10.11000 lU>
Source: World Bank estimates using ECVAM (2007) and ECVAM (2008 and 2010)
2.9          The prevalence and ranking of these self-reported shocks is consistent with
measured price and rainfaîl shocks using alternative data. Figures 2.4 plot mean rainfaîl
and variation derived from monthly rainfaîl measurements from 1989 to 2009 from 43
stations across Niger. The graphs indicate that average rainfaîl in 2007 and 2008 was
not bad, in the sense that the average rainfaîl level in most departments was above the
20 year mean rainfaîl. In fact> grain harvest in 2008 was considered the second highest
(surpassed only by 2010) in the ast 20 years (Reuters/FEWS NET> 2010). On the other
hand> 2009 could be considered a "bad" year, as confirmed by the well-publicized
drought where the government had to make a massive international appeal for food.
-30-


Figure 2.4.: Department-level annual rainfall means (2007, 2008, and 2009 vs. 20 year mean)
1000
-20 year mean (1989-2009)
-- 2007 department mean
800       2008 department mean
-U-2009 department mean
-700
E
M 600
-ru 500
400
300               _
200
100
0
<n 0       0        0                   0
o     0     0
Depatmet(Reion                   0
0                        <
H~~  0T              M H       M~ E <  >- -
0                                          0 = 0  0
HZ<
H                 Department (Region)
Source: World Bank staff estimates using rainfall data
2.10   Rainfall variability and timing are just as important. Figure 2.5 shows the
average rainfall (again over 20 years), and coefficient of variation (a measure of volatility
calculated as the ratio of the standard deviation to mean rainfall) for each department
in Niger. The mean rainfall during the rainy season months (June to September) varies a
lot from department to department. The coefficient of variation confirms that inter-
annual rainfall variability relative to the mean (the CV) is high, ranging from 22 to 35
percent above long run departmental means. Therefore, it is not unlikely that while
2007 was a relatively good year as Figure 2.4 suggests, many households may have had
irregular rains. In particular, even if the rainfall may have been adequate compared to
the 20 year mean, its arrival may have been irregular and disruptive enough that many
households may have ended up having poor harvests, underscoring their self-reported
characterization of the shock that year (see Figure 2.1). Figure 2.4 provides some
evidence for this possibility. Despite the above average rains in 2007 and 2008, only
one-third of the departments could be said to have received exceptional rains that year,
while another 15 percent received poor rains. The rest (almost 50 percent) received
average rainfall, suggesting that a large fraction of Nigerien families may still have
experienced poor harvests if the timing of rainfall was off. Figure 2.6 confirms that
-31-


almost 50 percent of households who reported having a bad harvest in 2007 relative to
the previous year, attributed it to early end to rains, while in 2009 the problem was one
of inadequate rainfall.
Figure 2.5: Mean and CV of rainfall, 1989-2009, by department
900                                                              0.40
800                                                              0.35
700                                                              0.30
0.30
600
0.25
 500
00.20
-@   400
C                                                 0.15
30% -
300                         Z year mean (18-ZUU9
200                                 coefficient of-variation (s.d/mean)  0.10
100                                                              0.05
0                                                               0.00
<                            00    0     0 < -4  - -QO
<  <                  < <
0< Z
O%)
ranalDeaesipartmento reginsan
Source: World Bank staff estimates from srainal data
F   : R e s  s a
Z) R~ 50%                           0Z z0z
< 0
U                                       0  <
7 ~  0%
0%
-
50%
40%
2-32


2.11  In the Nigerien context price and weather shocks are highly correlated shocks.
Figure 2.7 shows average monthly grain prices of a kilogram of millet from 2000-2010.
The graph also plots the incidents of reported droughts through that decade. First,
there is high volatility within and across years. For instance, there is obvious seasonality
to prices. Prices tend to be higher during the hungry season in the June-August quarter,
the period just before the harvest when all domestic grain stocks from the previous
agricultural season are likely to have been depleted, and lower in October-December
quarter following the harvests. Second, 2009 and 2010 prices were indeed higher than
2007 or 2008. However, in the last 5 years, the June-July prices in 2005 still remain the
highest due to the combination of drought and food supply shocks not only in Niger but
in the region. As shown in the graph, a drought is always followed by a sharp rise in
prices after about 4-5 months with a peak in the subsequent lean season. In particular, if
rains fail in the June to September months, October prices of millet and sorghum are
higher relative to the previous October price. When this is not the case, it has to do
with the production and supply conditions as well as the status of grain markets in
northern Nigeria and Benin, Niger's main trading partners. However, overall millet
prices in October following a drought would tend to rise by at least 20 percent relative
to October average prices. Finally, there is a large spatial variation to prices. Figure A1.1
(Annex 1) shows that even at the height of the food crisis in 2008, prices in Maradi were
about 60 percent lower than prices in Diffa and Niamey, reflecting perhaps Maradi's
proximity to Nigerian border grain markets.
Figure 2.7: Monthly millet prices and cumulative monthly rainfall during the rainy season
(F CFA), 2000-2010
dro. h                 drought                     droght
250                                                              250.1
200                                                              200
Source: World Bank staff estimates using data from SIMA.
2.12  Global food price shocks in 2008 were mitigated by a good domestic harvest.
Global food prices rose sharply in 2008, increasing by almost 50 percent between
January 2007 and June 2008 (see Figure A1.2 in Annex 1). More specifically, rice prices
more than tripled (217 percent) between 2006 and 2008, while maize and wheat prices
more than doubled (125 and 137 percent respectively) in the same period. These price
-33-


hikes were felt by many consumers around the world. However, in Niger, while prices
rose in 2008, they were not as severe for a number of reasons. Rural Nigeriens consume
millet as their principal grain. While millet is available in regional West African markets,
overall, it is far less tradable than rice and maize, so its price is likely to be affected less
by movements in other grain prices. Also, Niger cereal production in 2008 was one of its
best in two decades, surpassed by only the 2010 harvest. Since most of the rural
population rely on consumption of own production to meet their cereal needs, such a
harvest would have shielded them from the global price increase.
2.13  The intra-and inter-annual fluctuation of rainfall introduces significant supply
shocks in the grain markets, and in the most severe cases, it leads to seasonal food
insecurity and hunger for some households. The welfare costs of these shocks are
therefore expected to be large and are of interest for a number of reasons. There are
three important observations to note. First, they cause widespread short term poverty
in any given year in which they occur. Second, they influence long term poverty if
households' responses to protect themselves leads to asset depletion or participation in
less risky but also less profitable activities - in other words, it leads to poverty traps.
Finally, an estimate of welfare costs also provides a measure of the likely cost of insuring
such consumption shortfall either through safety nets or private insurance. In the next
section we look at the welfare costs of these shocks.
C.    Shocks Impose High Welfare Costs
2.14  Although Nigerien families face multiple shocks, the overwhelming shocks are
the irregular rainfall and related price movements. Therefore, our efforts to measure
the impact of shocks on welfare will concentrate on these two shocks. Households were
asked not just to rank the biggest shocks they experienced, but also to report what
impact it has had. Table 2.1 shows reported impacts.
2.15  The main impact of poor harvests is felt through loss of assets and reduced
consumption. Poor harvest could be due to drought, delayed on-set of rainfall, floods,
or pest attack. Almost 35 percent of households that ranked drought or flood as most
important felt the impact through reduced consumption and loss of productive and
durable assets. However, only 87 households in the sample could state clearly that
drought and floods were their greatest shocks. By far the single largest framing of the
weather shock is "poor harvest", where almost 1300 households ranked it as most
important in the past 12 months. In this case, about 45 percent of such households felt
the impact as reduced consumption, loss of productive and durable assets, or both (see
top row of Table 2.1).
-34-


Table 2.1: Reported household impacts of shocks ranked as most important (2007)
Self-Reported Impacts
(% of households that reported the shock as most important)
# of households                                            Asset loss plus
Income loss and
ranking shock as most Loss of household       Loss of durable or income loss or  No reported
important during last  income      reduced    productive assets  reduced    consequence
12 months                                                 consumption
Poor harvest                      1,303             0              4             11            28            56
Lack of money                      945              2              6             11            20            61
Food price increase                587              2              2             5             9             82
Lack of water                      188              0              2             3             6             89
Serious illness or disease         143              9              8             15            24            45
Unemployment                       100              2             18             3             16            61
Food availability                   71              4              3             7             25            61
Market access                       60              0              3             2             10            85
Drought                             57              2              9            23             4             63
Lack of manual labor                44              0              5             5             11            80
Loss of active household adult      38              5             11             8             18            58
Lackof animals                      32              0              0             3             13            84
Flood                               30              13            10             13            7             57
Insecurity                          24              0              8             8             0             83
Blocked road (to market or          14              0              0             0             7             93
administrative center)
Lack of pasture/ fodder             13              0              8             15            15            62
Livestock disease                   12              8              0             17            8             67
Fire                                7               0              0             0             86            14
Pests                               7               14             0             0             0             86
Lack of milk                        4               0              0             0             25            75
Source: World Bank staff estimates from 2007 ENBC
2.16      ... while food price shocks lead to food insecurity. Figure 2.8 displays perceived
impact of increases in food price on food security, among households that ranked food
price increases as one of the three greatest difficulties during the previous 12 months.
Almost 50 percent of households felt that the impact was very important in their lives.
In 2008, when the global food price increases reached their peak, substantial number of
households in both rural and urban areas felt that the price shocks had very important
impacts in their food security. This is despite the fact that 2008 was actually a good year
in the sense that production was one of the highest in the last 20 years. However, in
2010, following the severe drought of 2009, these perceptions were even stronger, to a
large extent because of the sharp rise in the share of the rural households who felt the
impact of food price shocks on food security to be very important.
-35-


Figure 2.8: Perceived impact of food price increases during the preceding
12 months on household food security
60%
50%
2 40%
0
30%                                            Very important
S30%
0
0  Moderate
o 20%-
I0 Weak
10%                     -                      No impact
0%
Pooled  Rural Urban Pooled  Rural Urban
2008                2010
Source: World Bank staff estimates from ECVAM.
2.17   ... and the nutrition of children. Figure 2.9 reports households' perception of the
impact of food price increases on nutrition of children below 5 years of age. Again, 2010
was the year where many more households felt these impacts more strongly and as in
the general case, rural households appear to have felt the impacts more strongly. In
2008, 25 percent of rural households felt that the impact of food price shocks on food
security were very important and that rose to 35 percent in 2010.
Figure 2.9: Perceived impact of food price increase on nutrition of children under 5 years
40%
35%
30%
Z 25%-
. 25No children
20%
o 2                                                   Very important
o 15%  -                                              Moderate
10% -                                               Weak
5%                                               E No impact
0%
Pooled  Rural  Urban  Pooled  Rural  Urban
2008                  2010
Source: World Bank staff estimates from ECVAM.
2.18   The self-reported perceptions of impacts of shocks suggest that shocks affect
welfare of households through multiple channels. As illustrated above, households
-36-


suffer income losses which lead to lower consumption, food insecurity, poor nutrition of
children and asset depletion.
2.19   We now look at quantifying these impacts more systematically. We begin with
the simplest exploration which runs a regression of total consumption, food
consumption, and staples (primarily millet) on an indicator on whether or not a
household reported experiencing a shock.         We control for region of residence,
household size, the fraction of members under 5 years of age, characteristics of head of
household (age, education, marital status, gender, and principal activity), assets and
share of household members who are active.
2.20    Health and irregular rainfall shocks are the most costly. According to the results
in Table 2.2, households hit by a death of a family member have, on average, per capita
consumption that is 11 percent less. A welfare loss of a similar magnitude is estimated
when per capita food consumption is used as the welfare measure. However, the
average millet consumption for the same households is lower but not statistically
significant. Households reporting difficulties related to a major illness of a family
member or exceptionally high health expenses show, on average, 7 percent lower per
capita millet consumption, and the impacts were reported to be significant in 2008 but
not in 2010. In 2008, millet consumption is estimated to have risen for those who
reported experiencing a price shock, but we suspect this reflects the fact that farmers
had a good harvest and therefore more consumption.
2.21    By comparison, irregular rainfall is reported to have significant negative impacts
in both years, regardless of what measure of welfare we use. According to the estimates
in Table 2.2, rainfall shocks led to 6-8 percent reduction, on average, in per capita total
and food consumption and up to 13 percent reduction in per capita millet consumption,
in 2008. Controlling for the households receipt of any transfer program, does not
change the results.
Table 2.2: Impact of shocks on per capita consumption, food expenditures, and millet consumption:
regression coefficients for various self-reported shocks
2008                        2010
Components of (log) pc expenditure  Components of (log) pc expenditure
Greatest self-reported shock / difficulty  total  food  millet  total  food     millet
Lower income                     -0.02    -0.01    -0.05  *  -0.03    -0.03    -0.04
Illness / exceptional health expenses  0.04  0.00  -0.07  **  0.06    0.03     0.04
Death of family member           -0.11  **  -0.11  **  -0.09  0.03    0.05     -0.02
Food price increase              0.02     0.02     0.08  *  -0.01     -0.01    -0.02
Irregular rain                  -0.08  **  -0.08  **  -0.13  **  -0.06  **  -0.05  *  -0.08  *
legend: p< 0.1 *; p< 0.05 ** p< 0.01 ***
Source: World Bank staff estimates using survey data
2.22    Rather than use a binary variable of self-reported incidence of shocks, we obtain
rainfall and price shocks from rainfall and price data collected in departments where
these households live in order to estimate their impact on welfare. We measure rainfall
shocks experienced by a household as the standardized deviation of mean rainfall from
-37-


the long run mean in the household's department of residence. We also try alternative
measures, such as absolute deviation of rainfall from the long run mean. We estimated
the long run mean as the average rainfall for 20 years (1990-2009). Figure 2.10 shows
the distribution of rainfall shocks in the last 3 years.
Figure 2.10: Exposure of rural households to rainfall irregularities: absolute (left) and standard deviations
(right) from 20 year mean rainfall during the rainy season, June-September, at department level
90%                                              m2007           90%                                        I 2007
80%                                              02008           80%                                        m 2008
70%                                              .2009           70%
-3 60%                                                             60%
50%                                                             50%
40%                                                            i 40%
30%                                                           130%    -
20%                                                             20%
10% -10%
0%                                              0%
<-150 <-100 <-50   <0     >0   >50   >100  > 150                < -15 < -10 < -05  <0  >0   >0.5  >1.0  > 1.5
deviation (mm) from 20 year mean rainfall                      deviation (s.d.) from 20 year mean rainfall
Source: World Bank staff estimates.
2.23      Having defined the rainfall shocks, we estimate its impact by pooling the three
ECVAM surveys and do a regression of the measures of welfare - (log) per capita total
consumption, (log) per capita food consumption, and (log) per capita millet
consumption - on these shocks. We control for department and year fixed effects and a
long   list of individual characteristics.             Effectively    we   run    a  difference-in-difference
estimation. Table 2.3 shows the results. The first three columns show the regression
results for each year, while the last two columns show the difference-in-difference
estimates, which pools all three years of data, which are also our preferred
specifications.
Table 2.3: Impact of rainfall shocks on per capita consumption (total, food, millet): regression results
2007             2008             2010           Pooled           Pooled
standard deviations standard deviations standard deviations  at least 1 s.d.  at least 100mm
Rainfall variable/shock:  of rainfall    of rainfall      of rainfall  below 20year mean below 20year mean
from20yearmean from 20yearmean from 20yearmean (indicatorvariable) (indicatorvariable)
Regression coefficient for rainfall variable
Dependent variable (log)
pce (total)                         0.010            0.030 *          0.095           -0.008          -0.075 **
pce (food)                          0.010            0.026            0.102            0.002           -0.064 *
pce (millet)                        0.014            0.094            0.103 ***       -0.102 *         -0.132 ***
Control variables
household characteristics and assets  Yes             Yes              Yes              Yes              Yes
agro-ecological zones                Yes              Yes              Yes              Yes              Yes
departments                          No               No               No               Yes              Yes
year indicator variable               No              No               No               Yes              Yes
- 2008for pce (total)                                  -                -              0.004            0.011
- 2010for pce (total)                                  -                -             -0.236           -0.225 **
- 2008 for pce (food)                                  -                -              0.005            0.013
- 2010for pce (food)                                   -                -             -0.241           -0.230 **
-2008for pce (millet)                 -                -                -              0.163 ***        0.167 **
- 2010for pce (millet)                -                -                -              0.084 ***        0.086
legend: p<0.1*; p<0.05 *;p<0.01*
-38-


F
2.24    Drought leads to large welfare losses. The most remarkable result in Table 2.3 is
just how bad 2010, which captures the drought of 2009, happens to be. The year
dummies show that there was no difference in average per capita consumption
between 2007 and 2008. However, average per capita consumption in 2010 was about a
quarter less (24 percent lower) than the average in 2007. Our estimates of rainfall
shock indicate that households who received rainfall that is at least 100mm less than
the 20 year mean had a per capita consumption of about 7 to 13 percent less than the
reference household not exposed to the shock. Overall, these estimates, confirm the
estimates we obtained using household self-reported responses on the impact of
irregular shocks.
2.25    We turn now to the welfare cost of price shocks. Prices in Niger are determined
by what happens in Niger and also the neighboring grain markets, particularly in Benin
and Northern Nigeria. It is not uncommon for prices in Niger to remain stable even
when production falls sharply in Niger simply because of the availability of imports from
Nigeria and/or Benin. To estimate the impact of "pure" price shocks is therefore
difficult. To solve this difficulty we use the information about the correlation between
rainfall and price shocks in Niger that can be recovered from           Figure 2.7.    More
specifically, we use rainfall shocks as an instrument for price shocks. First we estimate
millet prices in October of each year as a function of rainfall recorded in June to
September of the same year at the departmental level. We then predict the October
prices and use the predicted millet prices to create price shocks that we can use to
estimate welfare shocks. Table 2.4 shows the results.
Table 2.4: Price/rainfall elasticity of per capita household consumption: regression results
2007            2008           2010           Pooled
Regression coefficient for (log) millet price in October, predicted from  rainfall t
Dependent variable (log)
pce (total)                       0.041          -0.229 *        0.160 *        -0.323
pce (food)                        0.128          -0.155 *        0.155          -0.252
pce (millet)                     -0.138           0.340 ***     -0.385 *        -0.942
Control variables
household characteristics and assets  Yes          Yes            Yes             Yes
agro-ecological zones              Yes             Yes            Yes             Yes
departments                        No              No              No             Yes
year indicator variable            No              No              No             Yes
- 2008 for pce (total)                              -              -             0.021
- 2010for pce (total)                               -              -             -0.197
- 2008for pce (food)                                -              -             0.020
- 2010for pce (food)                                -              -             -0.209
- 2008for pce (millet)                              -              -             0.205
- 2010forpce (millet)                               -              -             0.181
legend: p<0.1*; p<0.05**; p<0.01***
t October millet prices were predicted with rainfall (during the rainy season of same year), department, and year variables.
2.26   ...So do price shocks. The last column which shows the pooled results shows the
sensitivity of total, food and millet consumption to changes in prices. The price variable
is in logs, therefore the estimated coefficient suggests that a 10 percent increase in
prices reduces total or food consumption by 3 percent, but the values are not
statistically significant for total and food consumption. But millet consumption is much
-39-


more sensitive, as we would expect. In this case, a 10 percent increase in the price of
millet leads to almost 10 percent (9.4) reduction in millet consumption per capita. Since
millet is such an important staple and accounts for a large share of household
consumption, these estimates imply that price shocks impose a lot of hardship to
Nigerien households.
2.27   The shortfalls in welfare, especially in 2010, obtained via regression methods are
large and supported by the household consumption data. Figure 2.11 plots the
breakdown of household consumption from own production, purchases and exchanges.
It also provides a glimpse of what happened between 2008 and 2010. In 2010, the
average share of food consumption from own production dropped substantially,
particularly in rural areas. Between 2008 and 2010, the share from own production fell
from 62% to 22% in rural areas and from 21% to 6% in urban areas. In above average
rainfall years (i.e. 2007 and 2008), rural households depend on own production for over
half of food consumption needs. Given lower production output with the 2009 drought,
households were forced to purchase more of their food needs. However, average per
capita food expenditures (including own production) in 2010 were lower, indicating that
households were not able to compensate fully for the production loss. How did
households, especially poor households, with limited income maintain consumption,
especially food consumption that would normally be met via own production?
Figure 2.11: Sources of household food consumption/expenditures.
100%                                   100%
90%   Rural                            90%   Urban                 Q1O%
80%                        77%         80%             .
70%                                    70%
60%                7                   60%
50%    o-                   -own production  50%                 -  own production
40%                -purchases          40%                       -purchases
30%                          exchange  30%                           change
20%                       No           20%
10%                                    10%                        6%
0%                                     0%
2007     2008     2010                 2007     2008     2010
Source: World Bank staff estimates from ECVAM
Figure 2.12: Households reporting fewer meals than usual
consumed by households and children.
35%
Fewer Meals Consumed by Households
30%
25%- Rural                p  76
E 25%
-W-Urban
g 20%
50 18%
15%
5%       6%
0%
2007       2008      2010
Source: World Bank staff estimates from ECVAM
-40-


D.    Coping Strategies Suggest Weak Resilience
2.28  Coping     strategies  are    Figure 2.13: Households reporting fewer meals than
inadequate and potentially harmful.    usual consumed by households and children.
The list of strategies that households  35%
reported to cope with shocks, shown     3  Fewer Meals Consumed by Households
in Annex 1 (Figures A1.5-A1.17), is  r 2        Rural
long. It includes migration of some  o-c
family members, borrowing from       o                               18%
family and friends, default on debt,                     6
begging and so on.  But when we
compare    the   proportions  of                         6%
households   using  the   various             2007      2008      2010
strategies, the largest fractions either
reduce  consumption  or deplete            Fewer Meals Consumed by Children  3
productive assets. Figure 2.13 shows
aa25%
the share of households who reduce            -U-Urban
5~20%
consumption, including feeding of    25%
children. Almost a quarter of rural
-610%
households   reduce    household                          7%
consumption and one third reduced
the consumption of children during     0%     2007      2008      2010
the early part of 2010, which is also
the worst year in the period. But     SoreWolBaktffsimesrmECA
even during the normal year (2007) and a good year (2008) between 8 to 6 percent of
rural households, respectively, would use this strategy to cope with shocks. Figure 2.14
shows that the fraction of rural households who would reduce their daily rations almost
on a daily basis (that is every day of the week) rose from around 3 percent in 2008 to 8
percent in 2010. Furthermore, the proportion that would reduce their ration at least
one day a week increased four-fold (from around 10 percent to 40 percent) in the same
period. Figure 2.15 shows that sometimes households cope by eating less nutritious
foods, or "shortage food". In 2010, nearly 20 percent of rural households had chosen to
substitute this type of food for at least a day.
-41-


Figure 2.14: Number of days (out of last 7 days)  Figure 2.15: Number of days (out of last 7 days) in which
in which households decreased daily rations.  households resorted to consuming food "de p6nurie"
45                                               25
40
35                                               20
M 7 days                                       M 7 days
:30                               M 6 days                                       M6 days
25                               M 5 days                                       M 5 days
020                       m4days                                        M *4 days
020  -1
1M 2 days                                      a 2 days
10 -                               I day          5 -                           E 1 day
Rural  Urban  Rural  Urban                      Rural  Urban  Rural  Urban
2008          2010                              2008         2010
Source: World Bank staff estimates from ECVAM
2.29    Another form of coping is depletion of productive assets. Figure 2.16 indicates
that the share of households who were compelled to sell their breeding livestock in
order to cope with shocks rose from around 8 percent in a normal year to about 25
percent in 2010. There is general agreement that the drop in livestock prices relative to
grains combined with threat of mass starvation because of unavailability of forage
compelled a lot of families to liquidate their herds. Consumption of seeds needed for
replanting is usually also not something that households do, but Figure 2.17 shows that
the share of rural households who had to deplete their seed stock rose from 2 percent
in 2008 to 10 percent in 2010.
Figure 2.16: Household reporting the sale of breeding animals to meet
food needs in last 30 days.
30
-2- Rural
25-- Urban
20
o 15-
10                                    9
1 9
5
0
2007                 2008                 2010
Source: World Bank staff estimates from ECVAM
-42-


Figure 2.17: Number of days (out of last 7 days) in which households consumed seeds
due to food insecurity.
10
M 7 days
= 6 days
M 5 days
M 4 days
m 3 days
=2 days
2                                          W I day
0
Rural    Urban    Rural    Urban
2008              2010
Source: World Bank staff estimates from ECVAM
2.30  Some of these coping strategies are, of course, taken only as an act of
desperation. For instance, "decapitalization"- depletion of breeding livestock and
planting seeds and reducing consumption - in the short run may be necessary to cope
with the immediate consequences of an extreme shock, but it carries potentially
harmful long term consequences. In the long run, these strategies deprive household
good nutrition, compromise their health and deprive them of productive capital. As a
consequence, the avenues for escaping poverty become narrowed, and in the worst
scenario become closed off.
2.31  These bad outcomes can be avoided if insurance markets and/or social
protection systems are widely available. For a household to fall into poverty after a
shock, other sources of protection must be missing, or inadequate. These include (a)
absent or insufficient mutual insurance in the presence of large and systemic shocks, (b)
incomplete, thin or missing markets for insurance, and (c) poor or lack of social
protection. Formal insurance markets in Niger are certainly incomplete or missing
altogether and do not provide a mechanism for households to cope with systemic or
idiosyncratic shocks. And while mutual insurance through kin, family, and community
may be widespread, they are known to be insufficient when households are
experiencing covariate shocks. When insurance markets are inadequate, social
protection can often fill the need, especially for the most vulnerable populations.
Unfortunately, Niger's social protection programs do not appear to be well-designed.
E.    Existing Safety Net programs are not well-targeted
2.32  Most social protection programs are poorly targeted. Figure 2.18 shows the
fraction of households in each consumption quintile who received any of several social
-43-


protection programs. As can be seen from the graph, of the three years examined, far
fewer households received benefits in 2007, not surprisingly because it was a normal
year. However, and surprisingly, similar fractions of households received benefits from
at least one transfer program in 2008 (a very good year) as in 2010 (a very bad year). It
should be noted that transfer programs were scaled up during the 2010 lean season,
that is, after the 2010 ECVAM survey, so such transfers may not be reflected in the
figure. An additional observation from the figure is the fact that the fraction of
households in each consumption group who received any transfer is about the same, or
even slightly higher for richer groups, meaning that the programs are not well-targeted.
Figure 2.18: Fraction of population receiving any transfer programs, 2007-2010.
60              Any Transfer Program
50 -
40 -
S2007
u 30 -
E                                        N 2008
C20  -                                            -
S0                                                    02010
1-
10
poorest    2nd ConsumpAro Quintile 4th  richest
Source: World Bank staff estimates from ECVAM survey data.
2.33  The poor targeting results observed above do not arise from    failure to
distinguish universal programs from programs specifically for the poor because a look at
individual programs leads to the same conclusion. Figure 2.19 shows the distribution of
beneficiaries for the subsidized cereal program. The fraction of beneficiaries in each
quintile are only slightly higher in 2010 compared to 2008. Moreover, in both years, the
share of beneficiaries in the top quintile was higher than in the lower quintiles. In fact,
in 2010, the fraction of beneficiaries in the lower quintile was less than in every other
quintile, making this particular program regressive. Other programs such as cash and
food for work, cash transfers and Zakat (albeit a private transfer) are just as badly
targeted, and are only better in the sense that they are not as regressive as subsidized
cereal programs (see Annex 1, Figures A1.18-A1.21).
2.34  Ideally, a good safety net will provide income support to the very poor during
normal and bad times, but also scale up to cover, only temporarily, those affected by
large and covariate shock. Therefore, while we should be concerned about the
targeting issues in a normal year such as 2007 or 2008, these issues become less of a
concern in a bad year such as 2010. Looking at the Niger programs, two programs that
are supposed to be self-targeting, food and cash for work (Annex 1, Figures A1.18-
A1.19), appear to have functioned this way. As can be seen from the figures, the
programs' coverage targeted higher fractions of households in the poorer quintiles in
-44-


2008, but targeting worsened in 2010, as we would expect. It is not clear if this was the
original intention of the Niger program design, but it is still a good thing that it
happened. That said, it should be clear that this does not mean that there are no
targeting issues to resolve for these programs. In fact, even if the programs cover
higher fractions of households in poorer quintiles, the coverage among the lower
quintiles is too low and the ones in higher quintiles potentially too high during the
normal years.
Figure 2.19: Benefit incidence of subsidized cereal program, 2008 and 2010.
20
m02008 Subsidized Cereals
N 2010
e15 -
C10 10
poorest    2nd        3rd       4th      richest
Consumption Quintile
Source: World Bank staff estimates from ECVAM
F.    Therefore chronic poverty is high
2.35  The frequency of shocks, the large welfare losses they impose, and the
inadequate coping mechanisms suggest high vulnerability. As there are several
definitions of vulnerability, for the purposes of this report we define vulnerability as the
probability of having a consumption (or income) level below the poverty line conditional
on initial consumption (or income). Put differently, vulnerability is the propensity of
households to fall into poverty in the future because of changes to income, especially
changes induced by shocks. Ideally, one would need panel data to follow the transitions
of households in and out of poverty and to estimate the likelihood of future poverty.
However, since such data are not available for Niger, we create pseudo-panels, from
three years of ECVAM cross-sectional data, to examine income dynamics and
vulnerability of households from 2007 to 2010.
2.36  We use a method proposed by Lanjouw, Luoto and McKenzie (2009) to obtain a
measure of vulnerability. The basic structure of this method is as follows. A model of
consumption (or income) is estimated using the second and third rounds of cross-
section data using only variables that do not readily change over time. The parameter
estimates from the second and third round models are used together with the time-
invariant covariates of households in the first round cross-section to predict
consumption in the second and third periods for a household surveyed in the first cross-
section. Analysis of transitions can then be obtained from actual consumption in the
second period and the estimated first round consumption. The model allows for a lower
-45-


and an upper bound estimate of mobility depending on how the unobserved
components of consumption (the error term), is treated.
2.37   Table 2.5 shows the transitions in and out of poverty for the period 2007-2010
for Niger. The details of how these results were arrived at are contained in Box 2.1.
Since consumption from ECVAM and ENBC data sets are not comparable, we calibrate
the poverty line in ECVAM in order to produce the same poverty headcount incidence
rate (60.8 percent) as was obtained in the ENBC 2007. This allows us to set aside the
complex issues of arriving at comparability of the poverty rates from the two different
surveys and focus instead on the dynamics of poverty.
Box: 2.1: Estimating Poverty Dynamics without a Panel
To estimate mobility using Lanjouw, Luoto and McKenzie (2009), we drop observations where
the age of head of household was below 25 years old or greater than 60 years so that the
underlying populations are not changing due to births and deaths. In step 1, using round 1 data,
we estimate consumption on time-invariant covariates in round 1, and obtain the parameters
and the error term. In step 2, we take a random draw (with replacement) of the error terms in
round 1 data, together with the parameters and the same time-invariant variables in round 2
data to obtain predicted round 1 consumption for round 2 individuals. In step 3, we calculate
transitions in and out of poverty, using predicted consumption in round 1 for round 2
individuals. Finally we repeat these three steps 500 times and calculate the transitions as the
average of the 500 repetitions. We also obtain the characteristics of the chronic poor in the
same three steps as the average characteristics over the 500 repetitions. Two additional issues
are worth noting. First, in this exercise the poverty line is calibrated so as to produce the same
initial poverty rate as the 2007 ENBC data. Second, the method described above is likely to
overstate mobility. So to provide a lower bound on mobility, an alternative assumption where
the individuals are given the same error term is imposed. In other words, in step 1 above we
regress round 1 consumption on round 1 covariates and round 2 consumption on round 2
covariates and obtain parameters and errors terms in period 1 and 2 respectively. In step 2 we
use parameters in round 1, round 2 covariates and round 2 error terms to estimate
consumption in round 1 for round 2 individuals.
2.38   Chronic poverty is widespread. Table 2.5 shows upper and lower bound
estimates of mobility. The lower and upper bound estimates provide radically different
and sometimes conflicting views of Nigerien context. The lower bound estimates
suggest that there is hardly any mobility in Niger. Instead chronic poverty is widespread.
We define chronic poverty as the population that was in poverty in all three years. The
estimates suggest that more than 90 percent of the poor in 2007 were still in poverty in
each of the following two survey years. Specifically, 54 percent of the population is
estimated to have been in poverty in all three years, implying that in those three years
only 10 percent of the poor population exited chronic poverty. A comparison of 2007
and 2008 suggest a small fraction of transient poor (9 percent of the population), which
almost doubles to 17 percent between 2007 and 2010. However, in chapter 1 we
noticed that the poverty gap also decreased - at least between 2005 and 2007. What
this implies is that the poor may becoming less poor, but not fast enough.
-46-


Table 2.5: Movement In and Out of Poverty (% of population)
Lower bound estimates of mobility  Upper bound estimates of mobility
Poverty Status       Poor in 2007  Nonpoor in 2007  Poor in 2007  Nonpoor in 2007
Poorin2008              54             3             35            17
Nonpoor in 2008         6             36             26            23
Poorin2010              58            14             42            22
Nonpoor in 2010         3             25            19             18
Poor in all 3 years            54                           26
Never poor in 3 years          25                           12
Source: World Bank staff estimates using ECVAM data (per capita expenditure deflated by CPI).
2.39  As the model predicts, the upper bound estimates would tend to overstate
mobility. And true to form, the transitions in and out of poverty using upper bound
estimates are large. They suggest that at least 40 percent of the population can be
classified as transient poor. Between 2007 and 2008 this was estimated at 43 percent
while a comparison of 2007 and 2010 leads to an estimate of 41 percent transient poor.
Because of these higher levels of transient poor, the estimated levels of chronic poverty
is much lower than would be predicted from the lower bound estimates. In fact only 26
percent of the population was estimated to have been in poverty in all the three years.
While that is still a high level of chronic poverty it is significantly smaller than the
implied levels of chronic poor using the lower bound estimates.
2.40  The majority of chronic poor are crop farmers. The distribution of the chronic
poor appears in Table 2.6, which provides the fraction of the chronic poor by primary
activity, education, land holding, and primary income source. Almost 8 in 10 of the
chronic poor live in households where the primary activity of the head of the household
is a crop farmer. The next highest group among the ranks of the chronic poor is heads
of households whose primary activity is commerce. According to these data sets, the
representation of the heads of households whose primary activity is livestock among the
chronic poor is small: only 2 percent of the chronic poor are heads of households whose
principal activity is livestock, and only 6 percent of the chronic poor live in the primarily
pastoral zones. The proportion is higher if we look only at the primary source of income
for the households. In that case, the share of the chronic poor whose primary source of
income is sales of livestock is about 12 percent (compared to 30 percent for crop sales)
and 15 percent if source is small commerce. In any case, it is important to keep in mind
that the nomadic herders are probably underrepresented in the surveys.
2.41  .. and also the least educated. A look at the education profile of the chronic
poor shows the highest concentration among heads of households with a Koranic
education. This is even higher than those without any education. Over 80 percent of the
chronic poor are households with a head without education or with only a Koranic
education.
2.42  Across space, the chronic poor are mostly in Maradi, followed by Tillabdri. A
little over one-quarter of the chronic poor live in Maradi while another 21 percent live in
-47-


Tillab6ri. By comparison, only small fractions of the chronic poor live in Niamey, Agadez
and Diffa. Partly this is a reflection that the chronic poor are less likely to be urban, since
the population of these three regions live mostly in the urban areas by the same names.
2.43    Proportionately more female heads of households are chronically poor. Only 7
percent of the households are headed by a female. However, about 10 percent of the
chronic   poor live   in  households which      are  headed    by  a  female.     Similarly,
proportionately more people in monogamous households are counted in the ranks of
the chronic poor compared to those living in polygamous households, a finding which is
altogether not surprising since the latter is usually a sign of relatively higher wealth.
Table 2.6: Characteristics of the Chronic Poor
Percent of Chronically Poor*
lower bound         upper bound
estimate of mobility estimate of mobility
Location
urban                                               12                  8
zone: agricultural                                  59                 59
zone: agri-pastoral                                 34                 35
zone: pastoral                                      6                   6
Agadez                                              1                   0
Diffa                                               3                   2
osso                                               17                 16
M    jaradi                                        27                  31
Tahoua                                              17                 14
Tillaberi                                           21                 25
Zinder                                              14                 12
Niamey                                             0                   0
Household Structure
average household size (# of members)               11                 12
aerage age of head (aerage age) -                   44                 44
female head                                        1o                  10
Education of Head
none                                                35                 38
Koranic                                            48                  49
literacy                                            4                   3
primary                                             10                  8
secondary                                           3                   1
tertiary                                            0                   0
IVarital Status of Head
married (monogamous)                                60                 53
married (polygamous)                                35                 42
divorced                                            1                   1
widowed                                             4                   4
single                                              O                   0
Principal Activity of Head
agriculture                                         78                 81
livestock                                           2                   2
commerce                                            8                   8
artisan                                             3                   2
administration                                      11
other                                               8                   6
House hold Agricultural Activity
agriculture in past year                            94                 96
land area, if agriculture in past year (hectares) -  5                  5
House hold Primary Income Source
sales of agricultural products                      30                 32
sales of livestock                                  12                 13
small commerce                                      15                 14
day labor                                           7                   6
sales of wood/straw                                 4                   4
sale of artisanal products                          5                   4
salary                                              I                   1
commerce / business                                 3                   4
transfers                                           9                   9
borrowing                                           1                   1
other                                               13                 13
.except for those categories denoted otherwise
Source: World Bank staff estimates using ECVAM data (per capita expenditure deflated by CPI).
-48-


G.    Conclusion:
2.44  In this chapter we documented major shocks that affect the lives of Nigeriens.
We also estimated the welfare consequences of these shocks. We find that price and
weather shocks, which are correlated, dominate the lives of Nigerien households. These
shocks impose huge welfare costs and high levels of vulnerability to poverty. In such an
environment, a durable and robust system of protection anchored by markets for
insurance and supplemented with public social safety net programs could help
households cope better with these shocks. Unfortunately, Niger's insurance markets are
weak or missing, while the existing safety net programs are not well-targeted and a few
are regressive. Therefore, there is very low resilience to shocks. So what could be done?
2.45  Niger needs to build resilience. That will require some long term investments
and short term actions that are not very costly. In the short run, the most urgent thing is
to build information system on market prices, on weather shocks, and on knowing the
poor. There is already a system for collecting regional prices of major staples, but the
system is not built for speed or transparency. To strengthen this system, it will be
necessary to build a connected system that uses modern information technologies to
collect information on prices from farmers, wholesalers and large traders from
neighboring trading partners. To know the poor, Niger needs to build regular,
consistent and comprehensive surveys that again use the capabilities of modern
technology. In the long run, investment in improving agricultural productivity and
developing insurance markets will be essential along with a diversified portfolio of
household incomes.
-49-


Chapter 3:            Children's Opportunities In Niger
A.    Introduction
3.1   As we saw in the first two chapters, material deprivation is widespread in Niger.
Headcount poverty of 60 percent masks large differences across regions, urban and
rural areas, and across socio-economic groups. We showed how a large part of the
observed deprivation can be explained by high exposure to shocks - from natural events
such as weather shocks and external shocks such as trade - which affect income profiles
and material well-being of Nigeriens. However, such large differences in material
outcomes can also be due to huge differences in past opportunities. In turn, current
inequality of opportunities can have large intergenerational differences in future
outcomes, including labor earnings, consumption and income, occupation, health and,
human capacity more generally. Therefore, understanding the nature of inequality of
opportunities is important in understanding the patterns of both current and future
poverty.
3.2   We define opportunities as a set of core basic services which enable individuals
to participate - with dignity - in basic functions of the society they live in. While what
such services include have varied across time, cultures and contexts, in recent years,
they have come to be associated with access to a minimum level of education, nutrition,
water and sanitation, electricity, and health. For example, the latest global consensus -
the MDGs - has mobilized global effort to universalize access to basic education, health
and nutrition. We use opportunity in this chapter to mean not just widespread
availability of services, but combined access and effective utilization of such services.
Therefore, it is not enough that a school is available in a slum. Rather, society must do
more and ensure that the children who live in the slum attend the school and learn as
their peers who live outside the slum. When they don't, we would say that there are
unequal opportunities in education.
3.3   There are at least three reasons why inequality of opportunity is problematic.
The first is that it violates a common social norm of fairness. Most outcomes such as
labor market earnings, incomes, wealth, occupation, etc., are determined by several
factors such as preferences, effort and ability, circumstances, and luck. In a just (or
ideal) world, differences in observed outcomes would be primarily determined by
differences in levels of effort and ability (or talent). However, as the experiences in
many contexts have shown, most observed differences are rarely due to differences in
effort or ability. Instead, they are due to factors that are inherited (e.g. gender,
ethnicity, parental education or income) or those that individuals find themselves in
(e.g. region, rural residence) - all of which are completely outside of the control of the
individuals. So long as such factors that are outside the control of the individuals play an
outsized role in determining life chances, opportunities will be judged unequal. Second,
when outcomes in society are not determined by effort and ability, but circumstances of
birth, inefficiency of economic activity will result. Such misallocation will undermine
-50-


overall productivity of the society, which in turn diminish prospects for inclusive wealth
creation and potentially lead to social conflict. Finally, inequality of opportunities and
resulting inefficiencies is also associated with unequal distribution of power, which in
many contexts lead to additional distortions and possibly conflict.
3.4    This chapter explores the inequality of opportunity in Niger. It attempts to
answer a few questions. Are opportunities in Niger shaped by circumstances? If so,
which circumstances are the most important? Has the country made progress in
improving opportunities for its children?  To answer these questions, we begin by
documenting patterns of inequality for two basic opportunities for children - education
and health. We then examine the progress made by Niger in improving opportunities for
its children and compare its performance to that of its neighbors. Along the process, we
show some of the key circumstances that appear to be the most important in
determining the trajectory of life chances in Niger.
3.5    Our goals are modest in this chapter. We hope to highlight certain human
development outcomes, such as school enrollment and children's nutrition,
purposefully. While we realize that there is much more to education than just showing
up for class, for millions of poor children in Niger, just getting into school will be a giant
leap forward. Similarly, children in Niger suffer from some of the highest incidence
rates of malnutrition and mortality in the world. Therefore, while there is a plethora of
health issues which merit serious attention, we have chosen in this chapter to focus on
highlighting factors which prevent a child from living to celebrate her/his 5th birthday.
Also, our goals are modest in part because we are limited by lack of datasets which
collect information on both human development outcomes as well as on household
welfare. In that vein, we draw upon those datasets which can be used to empirically
examine the relationship between poverty and human development outcomes.
Nonetheless, we hope to contribute to an almost non-existent empirical literature on
circumstances that alter children's life chances in Niger.
3.6    As the following sections demonstrate, human development outcomes, such as
educational attainment, incidence of child malnutrition and mortality, are poor in Niger.
Poor outcomes are a result of complex interaction of supply and demand side factors,
such as: insufficient spending on public services (budget constraint), not spending on
services used by the poor (regressive benefit-incidence), money and inputs not reaching
the poor (lack of planning, corruption), providers not showing up for work
(absenteeism), providers not providing adequate services while at work (quality of
provision), lack of market linkages (low demand for skilled labor), intra-household
preferences, social norms, and geographic isolation. Very little empirical work has been
done on any one of these factors in Niger , far less on the complex interplay.
A recent study by the World Bank (2009b) did a Public Expenditure Review on the Education and Health
sector examining issues of resource flows and accountability of personnel. Interestingly this came in the
aftermath of a certain episode in Niger where not a single text book reached a single pupil in rural Niger
-51-


3.7    One exception is a recent study of the Public Expenditure Tracking in education
and health. The study noted that weak public expenditure management in Niger may
exist for several reasons. First, the expenditure management chain is opaque and
difficult to monitor flows of resources from one entity to the next. The survey found a
general lack of systematic record keeping, particularly between central and regional
levels. Information on unit costs or transaction dates were missing systematically.
Without reporting such information it is difficult to know what, how much, and when
input materials should arrive at local providers. Second, citizens- or end users of
services- possess limited information about the inputs and services they should receive
in their schools and health facilities because the budget process and its specificity is not
revealed particularly to the public. Without such knowledge it is difficult to hold service
providers accountable for delivery of timely inputs and services. Relieving bottlenecks
and inefficiencies in the public expenditure management system for public services are
critical for improving outcomes in human development. Recommendations from the
study called for improving transparency in resource management through media and
public access to information. Below we hope to fill some gaps with regard to children's
education and health outcomes, beginning with education.
B.    Poverty and Education
3.8   The overwhelming majority of Niger's adult population cannot read or write.
Unfortunately this is a pervasive problem among the youth as well - Niger is one of 5
countries where less than 40 percent of the population between 15 and 24 years are
literate: Mali (24.2percent), Burkina Faso (33 percent), Afghanistan (34.3 percent), Niger
(36.5 percent), and Chad (37.6 percent). Therefore, both the Government of Niger and
development partners are in strong agreement that investment in the human capital
base has to be an essential component of Niger's pro-poor growth strategy. With the
framework of the Fast Track Initiative (FTI) and its PRSP, Niger launched a 10 year
development for the education sector in 2002. So have children's opportunities
improved as a result of the investments?
3.9   There is a sharp increase in enrollment rates. Niger's primary cycle lasts for six
years, while the secondary cycle lasts for seven years. Children are officially supposed
to enter primary schooling at the age of 6 (the age group 6 to 12 being for primary
schooling), and school enrollment is compulsory for children aged 7 - 15. The results
presented in this chapter will focus on the schooling outcomes in the primary cycle, but
because there is substantial late entry, it will sometimes present results for children
between 7-15 years, those for whom schooling is supposed to be mandatory. Table 3.2
shows that in 1998, about 47 percent of children between 7-15 years were enrolled.
Most recently, that is, in 2006 about 68 percent of this age group was reported enrolled
in school. By some estimate, close to 67percent of children in the reference age group (7
- 13) are currently enrolled in school. This is a 20 percentage point increase in
for an entire academic year. This underscores the lack of accountability permeating through the entire
system of service delivery.
-52-


enrollment in less than 10 years. Despite this dramatic (50 percent) increase in
enrollment, overall enrollment rates are still low and it is unlikely that Niger will meet its
MDG Education targets by 2015. The education system remains the least efficient even
compared to poorly performing neighboring countries: Niger has the highest spending
per pupil as a percentage of GDP per capita but the lowest primary school enrollment
and completion rate (World Bank 2009).
3.10   Education outcomes are, however, not uniformly dismal. They are relatively
lower for rural citizens, for girls, and for the poor. In this chapter we will present some
statistics of basic educational outcomes and explore factors which shape these
outcomes within the context of poverty. For that purpose we will draw upon the ENBC
2007/2008 dataset given that it is the only data source in Niger which has information
both on education outcomes as well as poverty (consumption based measure of
poverty).
3.11  The increase in coverage rate masks significant differences by income group.
Only 54 percent of children living in households from the poorest quartile attend school,
compared to 83 percent of children from the richest households (Figure 3.1). As Figure
3.1 illustrates, not only is there a huge gap between the poorest and the richest, but a
wide gap exists between the richest and the second richest as well.
Figure 3.1: Enrollment Rate for Children aged 7-13 by Consumption Quintile
(Poorest to Richest)
School Enrollment (%) and Poverty
90
80
70
60
50
40                                       . Consumption Quintile
30
20
10
0
Q1       Q2      Q3       Q4
World Bank staff estimates using ENBC 2007/2008
3.12   Location is a much larger disadvantage than income. A comparison of
enrollment rates by rural and urban residence shows that location turns out to be a
bigger factor in shaping schooling outcomes than poverty (Figure 3.2). The gap in
enrollment between poor children living in rural areas and poor children living in urban
areas is almost as large as the gap in enrollment between poor and rich children
(irrespective of residence) highlighted above. So while poor children are less likely to go
to school, this problem is most acute for poor rural children. Note that even for children
-53-


from households in the richest consumption quintile, 38 percent of rural children are
still not enrolled in school while most of their urban counterparts are in school.
Figure 3.2: Enrollment by location and poverty status
a Rural U urban
100
90
80
- 70
- 60
E
'~40
30
20
10
0
Q1           Q2           Q3          Q4
Quartile of expenditure
Source: World Bank staff estimates from ENBC 2007/2008
3.13   Gender inequity in access is still a problem. The overall enrollment rate for boys
is about 72 percent% while that of girls is 62 percent, so%, which means that there are 3
girls in school for every 4 boys. This gender gap (10 percentage point) in enrollment is,
however, modulated by poverty and location (Figure 3.3). While there is only a 2
percentage point enrollment gap between poor boys and poor girls living in urban areas,
there is a 15 percentage point gap in enrollment between poor boys and poor girls living
in rural areas. These findings highlight the fact that the problem of low enrollment
among poor girls living in rural areas is most acute.
Figure 3.3: Enrollment rate by gender, location and poverty status
MQ1 MQ2    Q3 MQ4
100
90
80
70
C
(V  50
E
S40
C   30
20
10
0
Girls rural  Girlsurban   Boys rural   Boys urban
Quartile of expenditure
Source: World Bank staff estimates using ENBC 2007/2008
-54-


C.    Correlates of Enrollment
3.14  The preceding graphs show the average enrollments in each location, by gender
and wealth. Now, we explore the correlates of school enrollment using a detailed probit
regression specification. The binary dependent variable takes on the value 1 if the child
is enrolled in school, 0 otherwise. We run a pooled regression, as well as separate
regressions for the rural and urban sample to explore whether certain covariates are
more influential depending upon the location of the household (as suggested by the raw
data highlighted above). The covariates we use include (a) child specific information
such as gender, age of the child (13 year olds are the left out category), whether the
child has a handicap/disability, and (b) household characteristics: female headed
household, age of household head, number of girls and boys between the ages of 0-4,
number of women between the ages of 15-64, number of men between the ages of 15-
64, number of women and men aged 65 and over, whether there is a school in the area,
whether there is a school less than 1 km away, consumption quintile (richest quintile is
left out category). Additionally, we also control whether the household lives in a rural
area in the pooled regression, regional effects (Niamey is the left out category), and
month of the survey (January is the left out category).
3.15  Regression results (Table A2.1, Annex 2) highlight the effect of several important
factors which shape school enrollment. Corroborating what was previously highlighted
in the raw data, children from the poorest quintile are 21% less likely to be enrolled in
school, children living in rural areas are 26% less likely to be enrolled in school, while
girls are 13% less likely to be enrolled in school.
3.16  Physical isolation is a major obstacle to improving equity of access, especially
for rural girls. We find that distance to school has a significant impact on school
enrollment, and this factor is particularly important in rural areas. Having a school
located in the village catchment area means that a rural child is 29% more likely to be
enrolled in school. Even if a school is within 1 km, the rural child is 25% more likely to
be enrolled in school compared to a child who lives further. While it also matters
whether a school is located nearby for an urban household, the marginal effect is much
smaller. This reflects the overall lack of infrastructure and limited transportation in rural
areas which makes it difficult for rural children to travel even long distances to attend
schools. There are also other notable differences between rural and urban areas. While
urban girls are 5% less likely (compared to boys) to be enrolled in school, rural girls are
17% less likely to be enrolled in school, even after controlling for a host of household
characteristics. This could reflect differences in social norms between rural and urban
areas, as well as opportunity cost and labor market returns of females. The presence of
other female children and older females in the household both positively affect the
probability of enrollment. Female members are likely to pick up the household chores
and make available time for children to go to school. This supports the idea that rural
children, especially girls, may have greater time constraints to participate fully in school.
The time constraints may increase with age. The age group variable left out in the
-55-


regression is 13. The closer children are in age to 13, the less likely they are to be in
school.
3.17  Poor children fare relatively worse in urban areas. While children from the
poorest quintile in rural areas are 14% less likely to be enrolled in school, children from
the poorest quintile in urban areas are 26% less likely to be enrolled in school. This
implies that special emphasis has to be devoted to reaching out to poor children in
urban areas.
3.18  There is a need for addressing problems on low enrollment among specific
category of children and by region. We find that children with a disability are almost
28% less likely to be enrolled in school. While overall enrollment rates are low, and the
first priority of a resource-constrained country should be to focus on the general
population, it is important to flag that special attention needs to be devoted to reaching
out to children with disabilities. Overall enrollment rates are low in Dosso and Maradi,
and enrollment rates are low particularly for rural children in these two regions.
D.    Poverty and Nutrition
3.19  While there is considerable variation in adult height and weight, a broad
consensus in biomedical research has established that children below the age of five
should share similar growth patterns in height and weight regardless of ethnicity and
gender under 'ideal' circumstances (e.g., adequate nutrition). Malnutrition by itself is
one of major killers of children, associated with almost half of all child deaths
worldwide. Malnourished children have lowered resistance to infection, and are more
likely to die from common childhood ailments like diarrheal diseases and respiratory
infections. Malnourished children who survive continue to be more susceptible to
morbidity spells with frequent illness further diminishing their nutritional status, locking
them into a vicious cycle of recurring sickness, diminished learning ability - with diverse
implications on productivity and earnings. Niger has high incidence of child malnutrition,
reflected in the fact that almost half of the under five children are stunted (low height
for age) and 40% are underweight. Given the fact that poor households often lack the
resources to purchase/produce sufficient quantity/quality of food, children in those
households often suffer from malnutrition.
3.20  In this section we will explore some of the factors which influence child height
and weight in Niger. Again, we draw upon the ENBC 2007/2008 dataset given that is the
only data source in Niger which has information both on nutritional outcomes as well as
poverty (consumption based measure of poverty). We will supplement this analysis
with DHS 2006 data which has information on nutritional outcomes, but does not have
any consumption/expenditure data. There are various other large sample surveys which
collect detailed information on child nutrition, such as UNICEF sentinel surveys.
However, like the DHS surveys, they do not collect information on household welfare.
For our sample, we draw upon children aged less than or equal to 59 months in the
ENBC 2007/2008 dataset. For the dependent variables, height (measured in
centimeters) and weight (measured in kilograms), we take log transformations to
-56-


reduce the influence of outliers and normalize the distribution. We control for child,
household, infrastructure, and regional factors in the regression. Table A2.2 (Annex 2)
presents the correlates of child height and weight.
3.21  It is not surprising that biological progression in age is the most important factor
which shapes growth in height and weight for children under five - about 70% of the
variation in the data is controlled for by just controlling for age variable (measured in
months).
3.22   Children in female headed household are more malnourished. While the
magnitude of the gender effect is not large for height - girls on average weigh 5% less
than boys - children in female headed households tend to be shorter (-1.4%), and weigh
less (-1.7%) as well. It is not clear why this is the case. Perhaps female headed
households are poorer and the result is simply the income effect. The results indeed
show that poverty is strongly correlated with child nutritional status. For instance, a one
percent increase in per-capita consumption is associated with a 3% increase in child
weight, and a 1% increase in height. This suggests a strong relationship between
poverty and child nutritional status. However, as we saw in chapter 1, female headed
households are not on average poorer and in fact proportionately more of them live in
urban areas. The alternative is mother's knowledge. A sizeable literature posits
mother's nutritional knowledge is strongly correlated with better nutritional outcomes
for children, a possibility nonetheless we could not ascertain in the case of Niger.
3.23   Children's nutritional status varies considerably by region. Compared to
children in Niamey, children in all other regions weigh less and are shorter (they are
shorter in Tillab6ri as well, however, the magnitude is small). It is interesting to note
that even though Agadez and Diffa are relatively better off compared to Niamey, (or at
least not worse off in terms of poverty), nutritional status of children are still lower
(Table 3.1). However, comparing Niamey to the other four (relatively) poor regions, this
finding of lower nutritional status reflects the fact that the average household in Niamey
is relatively wealthier, has greater access to nutrients, as well as access to services. This
is also consistent with the finding that children living in rural areas, weigh less (-3.3%)
and are shorter (-1%).
Table 3.1: Differences in nutritional status of children across regions
Height (logs)        1A/eight (logs)
Agadez             -3                    -8
Diffa            -1.5                  -5.3
Dosso            -0.7                  -2.7
Iaradi           -2.1                  -4
TahoLua           -1.3                  -2.5
Tillaberi          -0.5                  -2
Zinder            -2                   -6.9
Note: All estimates are percent differences from Niamey region's average
height and weight.
Source: World Bank staff estimates from survey data.
-57-


E.    Poverty and Mortality
3.24  After staying essentially unchanged for decades, the child mortality rate (number
of deaths per thousand children who die before reaching their 5th birthday) started a
secular decline in Niger starting from the 1990s. For example, the under-five child
mortality rate (deaths per thousand live births) was 309 in 1970, 310 in 1980, 305 in
1990, while falling to 227 by 2000 (last reliable estimate is 160 in 2006). The
Demographic and Health Survey (DHS), a standardized survey conducted in various
years in 75 countries, is the only survey which has a sample size large enough for
estimation of under-five child mortality rate (U5MR). The last DHS was conducted in
Niger in 2006.
3.25  Like all DHS surveys, no information is collected on income or consumption.
Some information is, however, collected on assets and dwelling structure. We use that
information to first create an "asset index" to proxy for household welfare. We use
principle component analysis to come up with a composite correlation index (using the
first eigenvalues) using the following variables: type of water source used by the
household (e.g., piped water into the house, public tap outside the house), type of
toilet, floor material of the dwelling, connection to electricity, and whether or not the
household owns the following - radio, television, refrigerator, bicycle, motorbike, car,
mobile phone.
3.26  Child mortality information is collected in the DHS by detailed retrospective
questions to adult women on their reproductive histories. To mitigate against censoring
bias, we only look at the sample of children who were born at least 5 years before the
date of the survey. We examine the correlates of child mortality in a Probit regression
framework. The dependent binary variable takes on the value of 1 if the child died
before reaching her/his 5th birthday, 0 otherwise. We control for gender of the child,
household asset index, education of the mother, age of mother when she had her first
child, whether or not the head of the household is a female, regional effects, and
whether the household resides in a rural area.
3.27  Probit regression results are presented in Table A2.3, Annex 2 (coefficients are
reported as marginal probability). While the interpretation of the coefficient of the
asset index is difficult, we see that it is significant and negatively related to the child
mortality, suggesting that children from wealthier households are less likely to die
before reaching the age of five. Consistent with studies across the world, we find that
mothers' education has a strong impact on child health - children of uneducated
mothers are 2.5% more likely to die before reaching their 5th birthday. Also consistent
with other studies, we find that delaying age of first birth (which in Niger is strongly
related to delaying age of marriage) lowers the likelihood of child mortality, although,
the relationship is not significant.
3.28  Similar to our findings of factors which influence education and nutrition
outcomes, we also find strong regional difference. However, like education and
nutritional outcomes which on average are lower in other regions compared to Niamey,
-58-


the picture is more nuanced when it comes to child mortality. Compared to children in
Niamey, children in Agadez and Diffa are 8% and 6%, respectively, less likely to die
before reaching their 5th birthday. Given that Agadez and Diffa are relatively wealthier
than Niamey, this is an expected finding, in contrast to our findings reported in the
previous section on child nutrition. On the other hand, children in Zinder are 6.5% more
likely to die - which is consistent with the fact that Zinder is poorer than Niamey.
However, we do not find a statistically significant result for other regions in contrast to
our findings on child nutrition. The one consistent finding, however, is that human
capital outcomes are lower in rural areas. Compared to children born in urban areas,
children born in rural areas are 3.2% more likely to die before reaching their 5th
birthday.
F.    Evolution of opportunities
3.29  Thus far we have shown that, on average, children's education and health
outcomes have improved over time. However, we also showed that coverage or access
to services is influenced by gender of children, residence in urban or rural areas, region
of birth and family wealth. This is an indication that opportunities for Niger's children
may not be equal. Recall that opportunity measures not only access to services but how
fairly the coverage is distributed. Therefore, in this section we look at the evolution of
opportunities in Niger. To track progress on opportunities we use human opportunity
index (HOI) which is coverage rate that takes into consideration the equity of access to
the service. The HOI is defined as the coverage rate plus a "penalty" for the share of
access to services considered universal that is not equitably distributed. As an
illustration, suppose that there are 100 children in a country, split equally between rich
and poor (i.e. 50 poor and 50 rich). Suppose that 40 rich children are enrolled, while
only 30 poor children are. The coverage rate would be estimated as 70 percent. But we
suspect that the distribution of the coverage violates the notion of fairness, where we
would expect the 70 slots to be equally distributed between rich and poor. Therefore,
the HOI would be 70*(1-5/70))=65. The ratio of the shortfall in equal coverage relative
to average coverage (5/70) is called Dissimilarity index (D-index) and it is a "penalty"
imposed on lack of equity in access. Table 3.2 shows the evolution of HOI in Niger
between 1998 and 2006 for some basic opportunities in education and health. The
calculations are all based on the DHS surveys.
3.30  Opportunities in Niger have increased sharply between 1998 and 2006. Table
3.2 shows the progress made in coverage and more importantly in equalizing
opportunities. Two of three education opportunities - school attendance and starting
primary on time - have improved. Similarly sharp improvements have happened in two
opportunities in health. The average coverage rate for school attendance for 6 to 15
year olds increased from 47 percent in 1998 to 68 percent in 2006. This is an impressive
45 percent improvement. However the HOI, which we have defined as inequality-
sensitive coverage rate rose more sharply, from 31 percent to 53 percent, which turns
out to be a 71 percent improvement. The opportunities for beginning primary on time
also improved dramatically, but no improvements are seen in opportunities for
-59-


completing primary on time. On health immunization against measles and underweight
- a measure of nutrition adequacy in childhood - also saw large changes, in the desired
direction.
Table 3.2: Human Opportunity index in education and health
Coverage      D-Index         HOI       Significant
change in
1998   2006   1998   2006   1998   2006  HO? (95%)
Education
School Attendance (6 to 11 years)  22.1%  33.1%  35.1%  22.7%  14.3%  25.6%  Yes
School Attendance (12 to 15 years) 25.0%  35.0%  32.2%  20.7%  17.0%  27.8%  Yes
Begin primary on time      5.4%   11.4%  46.7%  28.4%  2.9%   8.2%      Yes
Finished primary on time   6.1%   5.9%   40.7%  47.3%  3.6%   3.1%      No
Health
Immunization against measles  44.0%  52.6%  15.4%  12.0%  37.2%  46.3%  Yes
No underweight (Weight-for-age)  54.4%  62.1%  6.5%  5.2%  50.9%  58.8%  Yes
Source: World Bank staff estimates
3.31   The evidence suggests that both boys and girls benefited equally from the
expansion in opportunities. Tables A2.4 and A2.5 show the evolution of HOI in education
and health for girls and boys separately. The initial gaps in opportunities were very high
in education, but about the same in health, for girls and boys. In education, where the
opportunities are relatively pronounced, boys had around 7 percentage point higher
chance of attending school than girls, and these initial disparities did not decline
because of equal expansion of opportunities. However, despite the initial advantage
boys had in finishing primary on time, by 2006, the differences have vanished. By
contrast, girls' opportunities in health appear to be either as good or better than those
of boys. There are no differences on immunization against measles, while underweight
measures appear to favor girls - that is, a slightly higher fraction of girls are found to be
of normal weight.
3.32   Expanding access has been the key reason for the improvement in
opportunities. Table 3.3 shows a decomposition of the changes in HOI. Composition
means that the mix of children changed, that is the ratios between old and young. Scale,
as the name suggests, captures the role played by expansion of coverage to all groups.
Finally equalization captures the fraction of improvement that is attributable to
reduction in the differences between rich and poor, urban and rural and so on - in other
words, improving equity of access and thereby making coverage fairer. The findings in
Table 3.3 show that attempts to reach as many children as possible (scale effects)
explained most of the improvements observed in equalizing opportunities, but doing so
in an equitable way (equalization effect) also played an important role. This can be seen
by the fact that the dissimilarity index, or the penalty for not providing access according
to the equality of opportunity principle has declined a lot. Tables A2.6 and A2.7 confirm
that the same pattern holds when the decomposition is done separately for boys and
girls. That said, household composition seems to have relatively little weight in
explaining the changes in opportunities for girls. This suggests that large and small
-60-


families seem to treat girls the same. Similarly, location seems to carry more weight in
explaining changes in opportunities for girls than for boys (see Tables A2.8 and A2.9).
Table 3.3: Decomposing total change in Human Opportunity Index
Total Change in HOI               Decomposition
1998 - 2006                           Coverage
(Percentage Points)    Composition   Scale    Equalization
Education
School Attendance (6 to 11 years)    11.2                -0.8        8.3         3.8
School Attendance (12 to 15 years)   10.8                -0.2        7.1         3.9
Begin primary on time                 5.3                -0.2        3.6         2.0
Finished primary on time             -0.5                -0.1        0.1        -0.5
Health
Immunization against measles          9.1                 3.2        4.0         1.9
No underweight (Weight-for-age)       7.9                 0.5        6.9         0.5
Source: World Bank staff estimates
3.33    Despite this remarkable progress, Niger lags behind all its neighbors in
providing opportunities for all. Figures 3.4 and 3.5 compare Niger's basic coverage and
HOI for education and health, respectively, with 5 neighbors. In all the identified
opportunities, Niger lags behind its neighbors. The gulf separating Niger and the other
comparator countries appear smaller in health than it is in education, with the sole
exception of completing primary on time. On this last opportunity, there is not much
difference between Niger and its Francophone comparators, but the gap remains large
with Anglophone comparators.
Figure 3.4: Opportunities in Education, Niger compared to neighbors,
circa. 2006
School attendance (6 to 11 years)
100%
90%
80%
70%
600/
50%/
40%/
30%/
20%/
10%
Ghana    Nigeria Sierra Leone  Senegal  Mali   Niger
M Coverage     M HOI
-61-


School attendance (12 to 15 years)
80%
70%
80%--
50%o
40%/
30%/
20%/
10%
0%
Ghana      Nigeria  Sierra Leone  Senegal     Mali       Niger
M Coverage        M HOI
Begin primary on time
100%
90%
80%
70%
60%
50%
40%
100%
200/.
90/o
0%
Nigeri    na Sierra Leone MGh       Seneg.l    Mali       Niger
M Coverage        M HOI
Finished primary on time
100%
900/
80%/
70%/
60%/
500/
40%/
30%/
20%/
10%OL-
0%M                                      M.
Nigen.a     Ghana   Sierr Le,ne    Mal       Senegal     Niger
0 Coverage        M HOI
-62-


Immunization against measles
100%
90%
80%
70%
60%
50%
40%
30%-
20%-
10%
0%
Ghana   Senegal  Mali Sierra Leone  Niger  Nigena
M Coverage   m HOI
Figure 3.5: Opportunities in Health, Niger compared to neighbors, circa. 2006
No underweight
100%
90%
80%-
70% --
60%
50% -
40%/,
30%/-
20%
10%-
Senegal  Ghana Sierra Leone Nigeria  Mali  Niger
M Coverage   m HOI
3.34   The main reason for this huge gap with neighbors is because circumstances
continue to play a huge impact in access to services in Niger. Of the circumstances
examined, three explain most of the inequalities in opportunities: the income rank of
the household, the characteristics of the household head (that is gender, education, and
age) and residence in rural areas (Figure 3.6). This is true for both education and health
opportunities. Since rural residents tend to be poorer, and poverty is highly correlated
with education, age and gender of household heads, it is tempting to conclude that the
three circumstances identified here may in fact capture the role income differences play
in entrenching inequality of opportunities. But the analysis in this report shows that
these circumstances matter in their own respect. For instance, children of rich rural
residents are still only 66 percent more likely to enroll in schools as the children of rich
urban residence (see Figure 3.2), reflecting perhaps issues of simple availability (supply)
of schools and differences in urban and rural social norms (demand) that affect
household perceptions about the benefits of education and in particular the benefits of
education for girls.
-63-


Figure 3.6 : Dissimilarity Index using one by one each circumstance (education)
Geuder and Age
Presence ofparnt and orphan3               Presence ofchildren and elderly
Wealth quintiles                     Location (Urban/rural)
Household head chacteristics
(education, gender ard age)
-School Attendance (6 to 11 years) -School Attendance (12 to 15 years)
Begin primary on time  - Finished primaryon time
Figure 3.7: Dissimilarity Index using one by one each circumstance (health)
Gender and birth order
10%
Wealthquintiles                         Presence ofchildren and elderly
Mother characteristics (education,   Location(Urban/rural
employment, marital status and age)
- Immunization against measles  -No underweight (Weight-for-age)
G.      Conclusion
3.35    While   problems of low       school enrollment, inadequate nutrition, and           child
mortality, plague the nation as a whole - it is the poor, rural children of Niger who
disproportionately suffer. This inequality in opportunities is particularly pronounced in
regards to educational outcomes. Consistent with micro-econometric evidence from
Africa, we also find a strong bias against investment in girls' schooling . The confluence
of poverty (children from the poorest quintile are 21% less likely to be enrolled in
school), remoteness (children living in rural areas are 26% less likely to be enrolled in
8In contrast to micro-econometric evidence from South Asia for example, where there is systematically
stronger evidence of intra-household gender bias in both education and health investments.
-64-


school), and social norms (girls are 13% less likely to be enrolled in school), make it a
daunting challenge for policymakers to improve outcomes. However, that does not
mean that public investments cannot make a difference. For example, our analysis
shows that making public schools accessible to rural households has a strong impact on
increasing enrollment rates. Our analysis also highlights important regional variations in
educational outcomes - for example, not only are enrollment rates lower in Dosso and
Maradi, enrollment rates are particularly low for rural children in these two regions.
This implies that attention needs to be devoted towards a geographically differentiated
pubic strategy to reaching out to poor children.
3.36  Our analysis also finds a strong relationship between poverty and child
nutritional status. There is also some evidence that there might be a gender bias in
nutritional inputs - girls under five on average weigh 5% less than boys. Nutritional
status is also significantly lower in rural areas, as well as considerable variation in child
nutritional status by region. On average, compared to children in Niamey, children in all
other regions are shorter and weigh less. While for most regions this finding is
consistent with being poorer and having lower levels of public investments, this holds
even in relatively richer regions of Agadez and Diffa. Again, this calls for a detailed
spatially differentiated exploration of determinants of child malnutrition, as well as
strategies to geographically prioritize and target public interventions. Even though
there is no dataset in Niger which has systematic information on both child mortality
and income/consumption, we construct a proxy for household wealth (primarily using
asset information) to illustrate the relationship between child mortality and welfare.
While our findings suggest that there is a significant relationship between poverty and
child mortality, the most consistent finding which again emerges is that incidence of
child mortality is higher in rural areas. Our analysis of the correlates of child mortality
also highlights another important issue - the positive spill-over effects of human capital
investments. We find that mothers' education is significantly associated with lower
incidence of child mortality.
-65-


Chapter 4: Agriculture, Income, and Rural
Poverty
A.    Overview of the Chapter
4.1   We noted in the first chapter that rural areas experienced a faster reduction in
poverty between 2005 and 2007 and some of the change may be driven by a reduction
in inequality. However, rural areas still account for over 80 percent of the population
and the largest share of the poor. Part of the explanation, of course is that the primary
reliance on low-input, low-technology rain-fed agriculture is not sufficient for moving
rural households out of poverty. Thus, it is important to explore how Niger's agricultural
sector, especially the productivity of small farmers, contributes to rural welfare. This
chapter opens with an overview of crop production and characterizes the main
participants in each of the existing sub-sectors. The first section looks at aggregate data
and the evolution of agricultural productivity in Niger over time and in comparison to its
neighbors (section B). We then present some stylized facts about the agricultural system
using detailed, farm level information from the Niger General Agricultural Census
(NGAC) in section C. This provides the backdrop for looking at the links between
participation in farming and poverty issues which are tackled in the section that follows
(section D). Section E looks at correlates of farm level efficiency, while section F explores
the potential links between government spending and agricultural growth and therefore
incomes in rural areas.
4.2   As we suspect, most of the cultivated land (85 percent) is devoted to rain-fed
agriculture. Producers tend to deploy an intercropping system although pure stand
systems have demonstrated higher yields in most departments. More than 93 percent
of households engaged in agriculture are headed by males. Compared with male
farmers, females use systematically lower levels of inputs and produce lower yields.
Based on a stochastic production frontier model, we find that education, mechanization,
distance to nearest market (that is, isolation) and rainfall are among key correlates of
agricultural efficiency.
4.3   We also find that the share of agricultural income is much higher among poor
households than among the non-poor across departments, suggesting that agricultural
incomes are too low to lift many households out of poverty. Poor households rely upon
ni6be (cowpea), millet, and groundnut production as their main sources of income.
Growth accounting analysis suggests that fertilizer and agricultural labor have been
instrumental in maintaining a positive agricultural growth trend. In the most recent
period (1991-2006), Total Factor Productivity (TFP) has turned positive, growing at an
annual rate of 7.5, compared to -0.6 percent from 1962 to 1990. However, shocks
continue to take their toll, and the chapter provides evidence of the negative effects of
droughts, which have contributed to a lower than expected growth rate and long-term
low per capita income.
-66-


4.4    Finally, we use a computable general equilibrium (CGE) model to simulate the
impact of agricultural investment on agricultural growth and poverty reduction. Policy
simulations on possible impacts of promoting irrigation are also conducted. The results
suggest that expansion of irrigated land will boost agricultural growth as well as overall
economic growth. However, competition over limited inputs will temper growth in non-
irrigated and other agricultural sub-sectors. We offer recommendations for improving
agricultural production and increasing returns to poor producers.
B.     Overview of Agricultural Production
4.5    Producers cultivate during two seasons in Niger: the rainy season, considered
the main cropping period when grains and cowpeas (niebe) are produced, and the
irrigated season which includes the production of horticultural commodities and onions.
Rice is grown during both periods but the irrigated season predominates. Most of the
grain harvest is consumed locally although some of it may be traded regionally. Trade
activity depends upon regional market conditions and the degree to which foreign
markets face deficits or surpluses. These markets are extremely sensitive to weather
conditions and price shocks.
4.6    Depending upon the geographic zone, irrigated crops in Niger may be grown in
succession to grains on the same plot of land or they may be produced on a separate
parcel which features greater access to water. The bulk of irrigated production - almost
90 percent- comprises export commodities, chiefly onions and peppers. The crops
grown during these two seasons embody distinct technologies that command different
levels of inputs, coordination and infrastructure, and therefore yield different returns.
Below, Figure 4.1 represents the broad distribution of rainy and irrigated cropping
seasons in 2008.
Figure 4.1: Shares of Rain-Fed and Irrigated Production in the Agricultural sector (%)
46.7
27.6
22.6
3.1
Rain fed food crops Rain fed cash crops  Irrigated food  Irrigated cash crops
crops
Source : Niger 2008 Social Accounting Matrix (SAM), World Bank staff calculations.
4.7    The import and export data reveals that the country is a net exporter of
agricultural products to its sub-regional partners (data from BCEAO and World Bank
Niger Diagnostic Trade Integration Study (DTIS)). Table 4.1 shows that agro-pastoral
-67-


production contributes a significant portion of exports to Niger's economy and it has
increased over time. The adjusted trade data suggest that a large but uncertain portion
of trade in this sector is informal and largely elusive to capture via the customs
reporting.
Table 4.1: Main exports by Value, 2001-2005
BCEAO Adjusted Data (CFAF billion)
2002          2003          2004           2005          2006
Uranium         62.5         65.5           70.1           78.5          79.6
Live animals    36.7         33.3           26.8          31.8           48.6
Onions          13.8         15.5           35.7           38.4          42.3
Cowpeas         7.6          10.8           14.0           13.5          19.1
Gold            N/D          N/D            10.8           34.2          25.0
Misc.           74.2         79.40          73.3           55.5          47.1
Total           194.8         204.5         230.7          251.9         263.2
Source: BCEAO *Includes CFAF 478 million for leathers and skins.
4.8    Despite the important contribution of the agriculture sector to overall GDP
growth, agricultural performance is weak in comparison to its Sahelien neighbors. Table
4.2 below compares yields of key food crops across the Sahel. With the exception of
rice, yields of other Nigerien commodities presented lag behind those of comparator
countries. Yields per hectare of main agricultural products are low for all the three main
crops (millet, sorghum, and groundnuts) and are 43.2 percent lower on average
compared to other SSA countries. Rice yield per hectare is in line with the region's
average, and other commodities are driving exports.
Table 4.2: Comparison of Niger's yields of cereal production with Sahelien countries
Average yields, in Kg/Hectare (2003-2007)
Millet          Sorghum           Rice           Groundnuts
Burkina Faso         885.4           1,034.3          1898.5            733.7
Chad                 571.6            702.9           1,259.8            877.0
Mali                 722.9            812.2           2,297.9           887.3
Mauritania           184.1            397.9           4,218.7           800.5
Niger                444.6            335.0           3435.6             534.4
Senegal              617.7            811.6           2,523.5            775.4
Western Africa       875.5            978.2           1,574.0           1,177.3
SSA                  793.5            900.6           2294.5             990.2
Source: FAOSTAT.
4.9    Agricultural growth in Niger has fluctuated considerably over the last 50 years
resulting from both changes in intensification and productivity. Table 4.3 presents some
of these dynamics. We decompose change by decade to examine how various factors
contributed to overall growth. From the early 1960s to the late 1980s, growth was
mainly driven by the change in inputs. Overall, fertilizer and agricultural labor have been
critical in maintaining a positive growth trend with the contribution of labor remaining
relatively constant and fertilizer being highly variable. Capital intensification has been
flat over the entire period.
-68-


4.10   The largest gains occurred during the 1991-2000 period when TFP9 increased by
11.6 percent. Although the magnitude of TFP growth may reflect drastic changes in
fertilizer consumption during this period - for instance fertilizer consumption declined
by 76.9 percent in 1991, only to then increase by 164.2 percent in 1992, and 307.7
percent in 1994 - it may just as well be indicative of several other factors. First, initial
conditions of efficiency may have been low and the dramatic annual increases suggest
that Niger was catching up to its technological frontier. The literature on growth
accounting shows that the initial human capital endowment plays a crucial role in
determining the future level of total factor productivity for a given country. It also
emphasizes that the more favorable the initial conditions, the higher the TFP growth
performance. The decade experienced a large increase in fertilizer usage which would
have contributed to labor productivity. In Niger, agricultural technological change
surged dramatically from -43.2 percent in 1993, to an annual average of 7.5 percent
from 1994 to 1996.
4.11   Second, these efficiency gains may have been enabled through changes in
agricultural policy. During the second half of the 1990s, Niger is among countries that
enabled increases in real producer prices for exports through a combination of actions
to lower export taxes, raise administered producer prices, reduce marketing costs and
depreciate the exchange rate of the domestic currency.
4.12   Third, the ever present negative impacts of weather are likely to have reduced
the effects of technical change on agricultural growth. For instance, although this
decade accounted for the largest increases in fertilizer usage and total factor
productivity (TFP) over the entire period analyzed but the end result was only medium-
level growth. This may be because during the decade Niger encountered two severe
droughts, which combined with the ever pervasive weather risk, potentially muted the
impact of the TFP on agricultural growth.
4.13   The 2001-2006 period witnessed increases in both labor and land usage which
may have been driven by population growth. Improving returns to labor -and reducing
rural poverty--means that yields (output/hectare) must increase faster than the labor to
land ratio (worker/hectare). Changes in the relative use of inputs will affect labor and
land productivity.
There are two widely used methods to estimated TFP, parametric or non-parametric. The parametric
approach assumes a specific functional form for the relationship between the inputs and the outputs as
well as for the inefficiency term incorporated in the deviation of the observed values from the frontier.
For Niger, we are reporting only the non-parametric (Malmquist) estimates. It is also worth pointing out
that agricultural gross production (constant 1999-2001, US$1,000, smoothed using the Hodrick-Prescott
filter with A = 6.25) considered in this section includes both crop and livestock.
-69-


Table 4.3: Growth accounting of agricultural sector0 (%)
Agricultural  Fertilizer  Livestock  Machinery  Labor  Land  Total  TFP  Other
growth                                                  inputs
1962-1970   2.6          0.8       0.3        1.6        0.8     0.3   3.8     -2.5  1.4
1971-1980   3.5          2.3       0.1        0.3        0.8     0.2   3.8     -0.1  -0.2
1981-1990   1.6          0.4       -0.5       0.2        0.9     0.9   2.0     0.9   -1.3
1991-2000   4.9          4.5       0.4       -0.1        0.9     0.3   6.0     11.6  -12.8
2001-2006    6.2         0.1       0.2        0.0         1.0    0.9   2.2     3.4   0.7
1962-2006   3.6          1.8       0.1        0.4        0.9     0.5   3.7     2.7   -2.8
Source: Authors' calculation, data from FAO (2009) and Fuglie (2008)
C.      Farm   Level Production: Stylized Facts
4.14   In this section we use farm level production information from the Niger General
Agricultural Census (NGAC) conducted between 2004 and 2008 to highlight some
stylized facts about the characteristics of rural agricultural production. These stylized
facts provide the context for understanding rural agricultural productivity and poverty
outcomes.
4.15    The NGAC covered seven regions, thirty-six departments and urban areas of
Niamey. The census included about 1.6 million farm households out of a total
agricultural population of 10.1 million. Approximately 1.6 million hectares (excluding
Agadez) are cultivated. The cropping system is dominated by intercropping practices
which accounts for 77.3 percent of cultivated land, followed by monocropping, or pure
stand, practices which account for 17.3 percent. As a risk management strategy,
intercropping seems relevant to the risky circumstances of Niger. A little over five
percent of land is fallowed. Maradi (25.8 percent) and Zinder (26.9 percent) farm the
largest proportions of intercropped land (see Figure 4.2 ). Tillab6ri, alone, accounts for
44.5 percent of total mono-cropped land followed by Zinder (13.1 percent) and Dosso
(12.4 percent).
0 d, = j wi(t + TFP + F, where cit represents agricultural growth rate, wi share of input i, Yt input i
growth rate, and F other factors such as weather and seeds. For input shares, we used estimated
elasticities from Evenson and Dias Avila (2007) and Pratt and Yu (2008).
-70-


Figure 4.2: Distribution of Cultivated Land by Region (percent)
50.0
45.0
40.0
4  Pure stand  Intercropping
35.0
30.0
25.0
20.0
15.0
10.0 -
5.0 -
0.0                j
Diffa   Dosso   Maradi Tahoua Tillabery  Zinder  Niamey
Source: World Bank staff estimates from NGAC
4.16   There is some regional concentration of main crops grown. Millet, sorghum,
groundnuts and nidb6 are the main pure stand crops, while most intercropping involves
the combination of millet-ni6b6, which take up about 2.1 million hectares, and millet-
sorghum-ni6b, cultivated on 1.8 million hectares. The regional distribution of
cultivated land in pure stand by crop as reported in Table 4.4 is as follows:
*   Tillabdri accounts for 50.2 percent of the total millet area. The remaining is shared more
or less equally by other regions except urban Niamey, where it is rather negligible;
*  Zinder has the highest share of land allocated to sorghum (24 percent) followed by
Tillabdri (20 percent), Tahoua (18 percent), and Maradi (13 percent);
*   Dosso leads groundnut cultivation, with 55 percent of total area, followed Tillab6ri (19
percent), and Maradi (11 percent);
*   Tillabdri accounts for the largest area allocated to ni6b6 (44 percent), followed by Zinder
(18 percent), and Dosso (17 percent).
Table 4.4: Cultivated Land in Pure Stand by Regions and Crops (%)
Millet  Sorghum      Ni6b6  Groundnuts
Dosso                  9         21       16.6        54.6
Diffa                 8.1       4.3         14         0.6
Maradi                9.9      12.8        5.9        11.2
Tahoua               11.2       17.6       4.0         5.8
Niamey                 0          0        1.5          0
Zinder               11.7      23.7       17.6         8.9
Tillab6ri            50.2      20.4       43.6        18.8
Total                100        100        100         100
Source: World Bank staff estimates from NGAC
-71-


4.17  Cultivated land size per family tends to be small. On average, cultivated land is
around 5 hectares per household. However, as shown in Figure 4.3, land size varies
significantly across departments; from 0.6 ha in Niamey to 5.8 ha in Goure (Zinder
region). Across regions, the size of land cultivated by males is systematically higher than
by females (Figure 4.6) and the highest differences are observed in Dosso and Tillab6ri.
Much of the land is also of poor quality, so when land quality problem is factored in,
household's productive land tends to be even smaller than the reported sizes.
Figure 4.3: Average cultivated land by regions and departments (ha)
7
m6
c 5
r4
.23
2
20
0~        0     >J  w 0       0
0           0f  M E      ~              M
0                      F-    = i0  O
0   F-            M  ~
ID            F
Diffa  Dosso  Maradi  liameA  Tahoua  Tillaberi  Zinder
Source: World Bank staff estimates from NGAC
4.18  Cultivated land per capita has changed relatively little since 1980. Figure 4.4,
shows that the size of per capita cultivated land has remained steady except in Tahoua
where it has declined and Dosso where it has increased. Since population growth has
been relatively high during the same period, this suggests that agricultural expansion
has been achieved through expansion into previously uncultivated land. About 90
percent of cultivated land is allocated to millet, 5.4 percent to sorghum and 1.1 percent
to ni6b6. The rest, less than 4 percent, is shared by all other crops. On average, parcel
size for millet is slightly above 2 hectares, 1.6 hectares for sorghum, and 1.4 hectares for
ni6b6.
Figure 4.4: Per capita Cultivated Land (hectares)
0 1980  2004-2008
0.97
0.77      0.740.74           0.810.84
0.560.55                     0.55
1   1         10.44
II
Diffa    Dosso    Maradi   Tahoua     Zinder
Source: World Bank staff estimates from NGAC
-72-


4.19   But there is considerable variation of land cultivated per household within and
across region. Figure 4.5 reports the distribution of households by land quintile across
regions. On average, agricultural households cultivate approximately a little over 5
hectares of land but there is wide dispersion across regions. About twenty percent of
households cultivate less than 1.4 hectares of land compared with households located in
the top twenty percent of the distribution who farm over 4.3 hectares of land.
Households in Diffa, Tahoua, and Niamey regions farm comparatively less land than the
other regions. Approximately 60 percent of households located in these three regions
cultivate less than 2.5 hectares of land, that is, they have access to the two smallest land
sizes (bottom two quintiles of land holdings). By contrast, 60 percent of households in
Dosso and Tillab6ri work on land holdings in the top two quintiles (that is at least 4.3
hectares of land). Maradi producers fall in between the distribution with 60 percent
farming on less than 4.3 hectares of land.
Figure 4.5 : Land Size Farmed by Region
Land Size Farmed by Region
100%
50%
0%
Diffa  Dosso  Maradi Tahoua Tilaberi Zinder Niamey
a Quintile 1 m Quintile 2  Quintile 3 m Quintile 4 NQuintile 5
Source World Bank staff estimates from NGAC
Figure 4.6: Average Cultivated Land by Region and Gender (hectares)
3.5
3.0
m Male 0Female
2.5
£2.0
2~ 1.5
M~ 1.0
2
1j0.5
0.0
Tahoua   Diffa  Maradi Niamey  Zinder  Dosso  Tillaberi
Source: World Bank staff estimates from NGAC
-73-


4.20  There are important gender differences in size of land cultivated, activity
specialization, and input use. Figure 4.7 reports the distribution of activities by gender of
household head. Almost eighty percent of male-headed households combined
agricultural and livestock activities while the other 20 percent were equally split
between exclusive crop and livestock production. A smaller majority of female-headed
households also engage jointly in crop and livestock production. Compared to male
households, they are relatively more specialized in the management of livestock and are
less likely to produce only crops. Rural households are likely to engage in both
agricultural and livestock activities because joint engagement offers the possibility to
diversify income sources and reduce risk from sole reliance on rain fed agriculture. In
addition, livestock ownership will ensure some minimum access to organic fertilizer
inputs for field crops. Although female members within households tend most
frequently to livestock, management does not imply ownership: females may be more
likely to manage livestock but they may be less likely to have authority to sell livestock
or to retain income from these transactions.
4.21  The majority of male-headed households (80.7 percent) were engaged in millet
production compared with less than half of female-headed households (41 percent)
engaged in millet. These households also cultivated sorghum, groundnuts, and
voandzou.
Figure 4.7: Main Activity by Gender of Household Head (percent)
U Male U Female
78.1
65.6
25.0
0.6 1.8      11.0 76       10.
Non-agriculture  Crop production  Livestock only  Crop production
only                    and livestock
Source: World Bank staff estimates from NGAC
4.22  The amount of land managed by males and females and the type of
management also differs across the country (see Table 4.5 and Figure 4.8). The ratio of
total size of parcels managed by males and females varies between 9.2 in Zinder and
20.9 in Tillab6ri. Overall, 76 percent of parcels managed by males are under collective
regime while 60 percent of parcels under female responsibility are individual parcels.
There are also some geographical differences. For example, in Diffa 74.2 percent of
parcels under female control are individually managed compared to 60.9 percent of
parcels under male control.
-74-


Table 4.5: Land Management by Gender (%)
Male                         Female
Total    Collective  Individual  Total  Collective  Individual
Diffa           2         1          5       2         1          2
Dosso          18        15         29      12         6         15
Maradi         23        25         15      26        26         27
Tahoua         12        12         14      11        18          7
Tillab6ri      21         22        17      14         8         17
Zinder         23        25         16      34        40         30
Niamey          2         0          5       1         1          1
Total         100        100       100     100        100     268632
Source: World Bank staff estimates from NGAC.
4.23  Differences in land use according to gender may be observed for several reasons.
Household-heads are responsible for providing grains for family consumption. In an
autarkic setting in which rural households rely mostly upon own production for
consumption, household heads cultivate traditional food grains on the family-managed
plots. As noted above, these plots are typically larger than individual plots managed by
either males or females and larger than collective plots managed by female heads.
Additionally, other circumstances may hold for production of rice, a high value crop.
The crop is irrigated and, thus, it must be planted where land has been improved with
irrigation canals. Investments in irrigation increase the returns to land and its
ownership or use is more likely to be contested among family members. Research in
the region has demonstrated that when investments to land improve its return, female
members may lose their use rights to that land or to farming a crop in which
technological advancements are made (Jones 1986 and Von Braun and Webb 1989).
4.24  Another reason for women managing smaller sized plots is cultural norms. Land
tenure and natural resource management is governed by the Rural Code, a body of laws
that lays the foundation for rural land policy and is executed locally by citizen-
empowered land commissions. In practice, the interpretation of land inheritance and
succession are based on the intersection of civil, customary, and Islamic norms. With
regard to female inheritance claims, communities are likely to revert to customary and
Islamic practices which proscribe inheritance with some element of differentiation.
Under Islam, land is partitioned with two shares allocated to males and one share
allocated to female members. Thus, female-headed households are more likely to farm
smaller-sized fields compared with male-headed households.
4.25  Furthermore, customary land use rights may differ according to ethnic groups
and their outcomes may have strong geographic differentiation since by tradition
groups settled together. In particular, these groups established the notion of
collectively- and individually- managed fields within the household. The household head
is charged with the management of collective fields which are considered essential for
meeting household nutritional needs, and this objective obligates household members
to contribute their labor to collective fields over the course of the agricultural season.
-75-


Furthermore, adult members may be each assigned an individual field in which
management and returns from production of that field are under their respective
control. Although, household heads may manage both collective and individual fields,
other members would usually manage their own fields.  Thus, fields managed by
women are likely to be smaller compared to those managed by men because women
tend to manage mostly individual fields and such fields are smaller relative to collective
fields. With the possibility of questionable tenure arrangements, women may under-
invest in improvements to land parcels or utilize inputs less intensively.
Figure 4.8: Cultivated land" size by crops and gender (ha)
E Male  Female
34.81
28.28
2.19 1.49    0.18 0.08    0.87 1.06
Millet      Sorghum        Niebe      Groundnuts
Source: World Bank staff estimates from NGAC
4.26  Yields from pure stand crops are higher than yields obtained from intercropped
plots. As Table 4.2 shows, compared to the region, crop yields in Niger lag behind yields
of its neighbors. Yields across departments for select crops are shown in Figures A3.1-
A3.4 (Annex 3). Overall, except for a handful of departments, yields from pure stand
cropping are higher than those obtained by intercropping. The gap in average yield
between the two systems is largest for groundnuts. There is a potential for significant
productivity gains by a universal adoption of pure stand agriculture, even under current
production technology - such as quantity and quality of inputs applied.
4.27  However, input usage continues to vary widely by region and remains a
constraint on agricultural productivity. Input use and its impact on agricultural
production depend on the underlying crop production technology and whether inputs
are being used to maximize existing technologies or to adopt new technological
improvements. Figure 4.9 shows that a large fraction of households report using seeds,
and to some extent fertilizer. Unfortunately, it was not clear from the data whether
these are modern seeds or not. However, only half the farmers use pesticides and even
fewer use fungicides.
1 The NGAC 2004/2008 has a lot of missing values on land use; therefore we used zone level land use
whenever the reported crop production was positive.
-76-


Figure 4.9 : Input Usage and Access through Market (%)
0 Input use  Access through market
85.8
64.0               67.6       69.9
50.8510       53
~Eim i27.8
11.6                                 10.7       2.
Fungicides  Pesticides  Fertilizer  Manure    Seeds
Source: World Bank staff estimates from NGAC
4.28  The relatively high up take of seeds in Niger, is a reflection of growing
competence of institutions for seed distribution. A highly functional seed system
requires (a) being able to produce and disseminate high quality seed of multiple
varieties to large numbers of farmers; (b) developing quality standards and quality
control practices that work for both sellers and buyers; (c) monitoring impact of seed
usage and improving adoption rates; and (d) ramping up commercialization. The
capacity for seed production has increased through seed groups or cooperatives in
several locations across the country. A partnership of development partners and local
institutions such as Seed Certification and Legislation Agency (SICCLA, French acronym)
and seed producer farmers have increased their competency in producing high quality
seeds. Moreover, community-managed input stores have been established to supply
farmers with inputs on a cost-recovery basis.
4.29   However, it is important to note that these stores often do not stock large
volumes of modern seed varieties. Modern seeds are sold in local grain markets by
traders whose knowledge of seed production varies considerably. Traders in certain
markets may be familiar with specific varieties and yet in others traders do not
differentiate seeds from grain and they will mix different varieties together. Thus,
variation in both knowledge of seed attributes and quality control of production and
distribution and usage of modern seed varieties varies according to the locally available
seed production infrastructure across regions. This is one of the weaknesses of an
otherwise reasonably organized input distribution system.
4.30   Farmers' usage of agricultural inputs may be constrained by market access. To
get a glimpse of the functioning of the agricultural input markets in Niger, we provide, in
Table 4.6, a breakdown of where farmers obtain their inputs. Most frequently farmers
purchased fertilizer, fungicides, and pesticides from a local seller in comparison with
cooperatives, NGOs, own production or some other source. In contrast, farmers relied
upon a variety of sources for seeds and manure. For inputs in which a high frequency is
purchased through private sellers-or the private market-usage of that input is lower
in comparison to inputs that were acquired more frequently from outside of the market.
-77-


4.31   Several factors could limit the functionality of the input market for rural
households. First, farmers could have liquidity constraints. Often, producers don't have
resources available to finance inputs during the planting season-especially when
resources may be required to smooth household consumption until the next grain
harvest is available. A recently completed World Bank study on rural finance indicated
that the supply of rural credit in Niger was notably lower than comparator countries.
Thus limited access to financial resources may pose a barrier to purchasing input
commodities. Second, farmers may physically lack access to markets because of physical
isolation. Physical access would be impaired by poor road conditions or limited
transport. Third, if the quality of an input is difficult to ascertain or verification requires
sophisticated testing then buyers may again prefer to obtain inputs from sources in
which a personal relationship exists with the seller. Lack of trust between buyers and
the market may undercut its effectiveness.
Table 4.6 : Main sources of agricultural inputs (%)
Fertilizer Seeds Manure Fungicides Pesticides
Cooperative and association  12.5  9.5   0.2      13.8       8.9
Local seller              53.9   27.8    10.7     64.0      51.0
NGO                        1.1    4.3    0.2       0.0       1.7
Own-production             1.1   16.9   60.5       0.8       0.5
Other                     31.5   41.5   28.4      21.5      37.9
Source: World Bank staff estimates from NGAC
4.32   Both the yield differences between pure stand cropping and intercropping on
one hand and the varying use of modern inputs across regions on the other suggest that
Niger is foregoing large gains in production. Naturally, a system that moves from the
existing production technology to a new production technology characterized by
widespread use of modern inputs and specialization would improve productivity and
ultimately welfare. However, this is also unlikely to be widely adopted unless
households can be assured of reliable insurance against pervasive weather risk,
insurance which intercropping provides however inadequately. But, risk is just one of
many factors that affect yields and therefore productivity, and in section E we look at
some of the determinants of agricultural production and efficiency.
4.33   A promising new area for agricultural diversification is high value crops, but its
expansion is constrained by poor price stability, land for expansion and water
availability. Table 4.7 lists the main problems reported by farmers as being responsible
for changes in production from the previous harvest. The shares in parentheses show
the fraction of farmers who identified that problem as the main cause of change in
production. Among major causes driving change in production, water availability, price
and threats from parasites are the most cited. For example, about 33 percent of farmers
point to low price as the main raison for unchanged production. The importance of
these production drivers varies across crops. For example, land expansion was more
important in increasing production for tomato and onion compared to pepper, lettuce
and cabbage. Similarly, the role of price was higher for lettuce, cabbage and onion than
for tomato and pepper. It is worth mentioning that increase in production is still
dependent on land expansion, not necessarily because land is abundant when its
-78-


marginal productivity is considered, but more because there are too many constraints to
intensification of production.
Table 4.7: Main causes of production change
Water availability (13.1%)
Decreased        Low price (18.9%)
Parasite pressure (25.7%)
Water availability (14.9%)
Unchanged        Low price (32.7%)
Parasite pressure (14.6%)
Water availability (16.5%)
Increased        Good price (12.9%)
Land expansion (32.9%)
Source: World Bank staff estimates from NGAC
4.34   Undoubtedly, to improve diversification through high value crops and to
intensify production of such crops, access to financing is crucial. Below, we present
information on access to agricultural funding from farmers engaged in horticultural
production. Only very few farmers received financing, and among the recipients, most
were male farmers. Almost 19 percent of male respondents received funding compared
with about 6 percent of female respondents and the sums that males received were on
average 42 percent larger than sums received by females. Without knowing whether
the funding targeted the provision of specific inputs, there are also notable differences
in what the funding was used for. The top three uses of external funding were seeds,
storage and fertilizer. Around 25 percent of farmers used the additional financial
resources to buy seeds, another 18 percent used it on storage and about 17 percent on
fertilizer. Almost 14 percent used it on agricultural equipment. Table 4.8 presents the
reported use of funding by gender. The biggest difference between males and females
regarding what to finance with the external funding appears to be seeds; 18.6 percent
for males but 45.9 percent for females. Two other equally stark differences are in
agricultural equipment and crop management.
Table 4.8: Use of External Funding by Farmers (%)
Male      Female
Seeds                        18.6       45.9
Fertilizer                  23.8        29.9
Ag equipment                 16.4       9.6
Crop management              12.2       1.9
Salary                       2.1        0.6
Well                         0.4        1.9
Storage                      17.7       16.0
Other                        8.9       10.2
Source: World Bank staff estimates from NGAC
D.     Participation in agricultural sector and household poverty
4.35   The stylized facts of farm level production demonstrate that too many families
have small pieces of land, of poor quality and inputs markets that do not work.
-79-


Therefore we would expect there to be a strong association between participation in
the agricultural sector and poverty.
4.36  The data confirms that the poor are more dependent upon agricultural income
as their single source of income (Figure 4.10). More than 60 percent of the income of
poor households originates from crop production (26.7 percent) and livestock (39.4
percent). In contrast, the non-poor have diversified income portfolio and count upon
wage activities for a substantial portion of their income. The agricultural and livestock
income of the non-poor contribute substantially less - about 16.5 percent and 28.0
percent, respectively - to total income.
Figure 4.10: Sources of Household Income (%)
U Non poor  Poor
51.7
39.4
31.3
26.7      28.0
16.5
3.8  2.6
Agriculture    Livestock       Salary         Other
Source: World Bank staff estimates from ENBC 2007/08.
4.37  The overwhelming reliance of the poor on agricultural income is widespread
across the land. In Figure 4.11, we plot the share of agricultural income as a percentage
of overall income for both poor and non-poor households by departments. Except for
the departments of Diffa, Maine-soroa, Guidan Roumdji, Mayahi, Keita, and Madaoua,
agricultural income constitutes a larger proportion of the household rural income
portfolio among the poor than among the non-poor. Of the thirty six departments
presented, only households in three departments (Arlit, Tchirozerine and Niamey) earn
less than fifty percent of their total income from agriculture, compared to fourteen
departments for non-poor households.
-80-


Figure 4.11: Share of Agricultural Income by Departments and Poverty Status (%)
120.0
100.0             -    Non-poor     Poor
80.0
60.0A
40.0VkA                                    kAW,n
20.0
0.0
Source: World Bank staff estimates from ENBC 2007/08
4.38   A look at income profiles of households reaches the same conclusion. In Figure
4.12 we present the distribution of agricultural participation and poverty incidence by
region. In regions where poverty incidence is among the highest, there tends to be a
larger share of the population engaged in agriculture. Some of the high incidence of
poverty has to do with high dependency ratios on account of higher family sizes. On
average, the regions of Dosso, Maradi, and Tillab6ri have larger household sizes, and
they also tend to have some of the highest incidence of poverty.
Figure 4.12: Agricultural Participation and Poverty Incidence
Agricultural Participation and Poverty Incidence
100                                                       8
7
50                                                     -
5
4
0                             -                          3
Agadez Diffa  Dosso Maradi Niamey Tahoua Tilaberi Zinder
Agricultural Population  Poverty Incidence  HH Size
Source: ENBC, 2007/08
4.39   Livestock alone accounts for 63.5% of household agricultural income; evidence
of the dominance of livestock is observed in almost all departments except Dakoro,
Guidan Roumdji, Kollo, Gour6, Magaria, and Kantch6. Both non-poor and poor
households rely mostly on sales from sheep and goats (see Table 4.9). However, cattle
and poultry also constitute an important source of income for poor households.
-81-


Table 4.9: Average livestock income (millions, FCFA)
Non-poor         Poor
Cow                            60.3        114.2
Sheep                         473.5        198.5
Goat                          249.0        190.3
Camel                          10.3         36.8
Pork                            0.0          0.0
Donkey                         54.2          1.5
Horse                           0.5          2.9
Poultry                        14.0        162.4
Fish                            4.0          1.5
Other                           1.1          3.4
Source: World Bank staff calculation from ENBC 2007.
4.40   Among crops, cereals are the main sources of income for poor and non-poor
households alike. As shown in Table 4.10, the three top sources of income for non-poor
households are millet, sesame, and ni6be, while ni6be, millet and groundnuts are the
most important for poor households.
Table 4.10: Average crop income (millions, FCFA)
Non-poor      Poor
Millet                             321.4     137.1
Sorghum and maize                   60.0      14.4
Other cereals                        8.7       5.9
Ni6be                              105.1     144.3
Groundnuts                          33.8      79.5
Sesame                              125.2      1.5
Boodlel2
Tobacco                              35.9     12.1
Other                               81.5      61.9
Source: World Bank staff estimates from ENBC 2007/08.
4.41   As the preceding sections have shown, Niger's agriculture is characterized by
overreliance on small family plots tilled by family labor, utilizing little modern inputs.
Except for a few islands of high value crops, most of the production is produced for own
consumption. As a consequence, most families who rely on agriculture live in poverty.
And yet, as the stylized facts demonstrate, there is enough scope to improve yields, if
hurdles imposed by natural forces (weather risk), dysfunctional markets, and
government investment can be overcome. This will mean improving productivity and
efficiency through improved production technology and deploying public investments
better. In the next two sections, we look at determinants of agricultural production and
efficiency and government expenditure and their likely impact on poverty.
E.     Determinants of farm-level agricultural production and efficiency
4.42   We use a frontier production model to analyze the key determinants of
household level agricultural efficiency. A common approach used to explain efficiency
12 From French "Oseille".
-82-


consists of first estimating a stochastic frontier production function, from which the
agricultural efficiency index is computed and subsequently regressing the efficiency
index obtained on farmers' characteristics. But this procedure may be biased for two
reasons. First, a possible correlation exists between variables in the frontier production
function and the inefficiency term. Second, the inefficiency term from the first step is
measured with error and it is correlated with the exogenous factors (Liu and Myers,
2009). Consequently, we assume a stochastic production frontier of the form following
Battese and Coelli (1995) and Kumbhakar and Lovell (2000). We present the approach in
Annex II.
4.43   Returns to land are higher than returns to labor and seeds. Estimation results
from such an approach are reported in Table 4.11. We used the unrestricted translog
specification to allow for interactions between inputs. Before getting to the details of
the determinants of farm efficiency, we summarize the findings of the production
model, a discussion that is relevant in its own right. The top panel of Table 4.11 (panel
A) suggests that expansion of land and labor inputs increase output, although the
relationship is concave meaning that production rises with input use, peaks at a certain
land size or labor input and then declines. Both production and input levels are in logs,
so the elasticities'3 are reported in Figure 4.13. Overall, when accounting for cross-
inputs effects, land effect is the most dominant compared to labor and seeds; a 10
percent increase in land and labor at the existing production technology will increase
output by about 3.8 and 0.2 percent respectively. This means that land is still more
productive than labor in the Nigerien context. The returns to using more seeds only are
not positive. The findings suggest that increasing usage of seeds with current technology
will reduce output - a 10 percent increase will lead to around 7 percent reduction in
output if land and labor effects on seeds are not accounted for. But a joint increase of
land and seeds, or labor and seeds will increase output, although the increase is small.
However, it is worth noting that the results do not distinguish between regular and
modern seeds. Therefore, part of the lower returns attributed to seeds, could be
because of usage of regular seeds predominate.
13 The translog production function used in this paper is of the following form:
Lny = Lnao + Ei aiLnx + 1/2 E Ej fliLnxi * Lnxj + yZ, where y is the output, xi is the quantity of input i, and Z
a control variable (in this case, rainfall). To maintain the consistency with Young's theorem of integrable functions,
flj = flj. It follows that elasticity of output with respect to inputs quantities is given by El = aLny = a + EjflijLnxj
aLnx-
-83-


Table 4.11: Estimation results14
Panel A: Production - Frontier production estimation results
Coefficient      SE
Land (ha)                                               0.7189a          0.0657
Labor (person days)                                    -0.0478           0.0433
Seed (kg)                                              -0.6579a          0.0718
Rainfall (mm)                                           6.0960a          0.1227
Rainfall squared                                       -0.7285           0.0200
Land squared                                           -0.2239a          0.0079
Land*Labor                                              -0.0334a         0.0068
Land*Seed                                               0.1327a          0.0097
Labor squared                                           -0.0184a         0.0062
Labor*Seed                                              0.0688           0.0100
Seed squared                                            0.0394           0.0111
Panel B: Determinants of Inefficiency
Demographics
Gender (1 if female, 0 if male)                      -0.1653c         0.0857
Age (years)                                          -0.0503a         0.0138
Age squared                                          0.0004a          0.0002
Self-sufficiency (default=more than 9 months)
Intermediate (3 to 9 months)                            0.5794           0.1045
Poor (less than 3 months)                               1.1107a          0.1327
School attendance (1 if no formal education, 0 otherwise)  -0.6830a    0.1156
Agricultural services
Crop management (1 if didn't use, 0 otherwise)        0.8959a          0.1589
Mechanization (1 if didn't use, 0 otherwise)          2.3534a          0.1490
Productive assets
Number of farming parcels                             0.3863a          0.0774
Number of cattle                                      0.0163           0.0141
Number of sheep                                       0.1769a          0.0169
Remoteness
Distance to the nearest food market (km)              -0.6243a         0.1795
Distance to the nearest food market squared           0.1697           0.0580
Agro ecological zones (Default=zone 1)
Zone 2                                                0.6219b          0.2452
Zone 3                                                -6.1947a         0.7801
Zone 4                                                0.4544b          0.2063
Zone 5                                                -2.0146a         0.2185
Zone 6                                                1.0346           0.2136
Zone 7                                                1.8311          0.2560
Zone 8                                                1.3701a          0.2778
Intercept                                               -1.5126a         0.4172
# observations                                          3294
Log likelihood                                         -1933.6
Wald chi2(11)                                           704137.9
Note: a,b means significant at 1% and 10% respectively; SE=standard error.
14 Data from the General Agricultural Census Survey 2004/2008
-84-


Figure 4.13: Output elasticities with respect to inputs.
0.381
0.024              0.018
Land               Labor              Seed
4.44   Figure 4.14 also provides some possible insights into why the productivity of
inputs like seeds and labor are low as we noted above. Given the quality of land, a
complement of modern inputs would be necessary to raise productivity. However,
already a large fraction of the farmers use seeds, although this is not all necessarily the
modern variety of seeds. So additional use of this input is unlikely to increase output
substantially as the marginal returns will likely be low. Furthermore, farmers do not
appear to be using other important complementary inputs - such as pesticides and
fungicides - as much.
Figure 4.14: Access through Market
E input use   Accessthrough market
85.8
64.0               67.65       69
50.8 51.053
I 27.8
16                               10.7
Fungicides  Pesticides  Fertilizer  Manure      Seeds
4.45   The distribution of agricultural efficiency is bi-modal. Figure 4.15 plots the
density of farmers' technical efficiency. On average, farmers' agricultural efficiencys is
estimated at 0.58. Thus, given the level of inputs, farmers are able to realize 58 percent
of possible achievable production, on average. This alone suggests that there is room for
improvement even without increasing the current quantity of agricultural inputs.
However, as the graph demonstrates, efficiency of farmers varies widely. In fact the
distribution is bi-modal: there is a large fraction of farmers who seem pretty efficient
and another smaller but still significant fraction who are very inefficient. The more
1s The measure of technical efficiency is given by E{exp(-ui)ci},{exp(-ui)lc},E(exp(-ui) lei}, whereci =yj -
Xifl -ci = Yi -Xifl.
-85-


efficient farmers are able to realize over 60 percent of potential output, and some can
harvest up to 90 percent, given the context. By contrast, the large proportion of
inefficient farmers cannot obtain more than 20 percent of potential output. Table 4.12
lists some potential drivers of efficiency differential between farmers. Below we discuss
some of the key correlates of efficiency and production.
Figure 4.15: Distribution of farming efficiency
0 i     ...       ... LII
O            .2           4           .6            81
M       Density  -        Kernel density
Source: World Bank staff estimates from NGAC
Table 4.12: Potential drivers of efficiency gap
Efficiency<=0.6  Efficiency>.6
Land investment (%)
Did not invest                       83.3            52.5
Invested                             16.7            47.5
Land slope (%)
Plain                                91.2            81.8
Lowlands                              8.8            16.7
Other                                 0.0              1.5
Farm management (%)
Individual                           33.7            25.7
Collective                           66.3            74.3
Crop management (%)
Apply                                 4.7              8.5
Do not apply                         95.3            91.5
Mechanization (%)
Use                                   8.0            11.2
Do not use                           92.0            88.8
Use of chemicalfertilizer (%)
Use                                   5.7            14.3
Do not use                           94.3            85.7
Gender (percentage)
Male                                84.8             83.9
Female                               15.2            16.1
Inputs quantities
Land (ha)                             9.7             3.3
Seeds (kg)                           28.1            35.0
Labor (person days)                  13.5            15.2
Source: World Bank staff estimates from survey data
4.46    The summary presented in Table. 4.12 confirm            the importance of on-farm
improvements. For example about 48 percent of high efficiency (efficiency >.6) farmers
-86-


report using various types of improvement compared to only 16.7 percent of low
efficiency (efficiency<=.6) farmers. At least 80 percent of these improvements involve
tillage, manure and irrigation. Use of crop management also explains the efficiency gap
between the high and low efficiency farmers. Only 4.7 percent of low efficiency farmers
use crop management compared to 8.5 percent of high efficiency farmers. A good deal
of the differences in the efficiency may be related to quality of land accessible. For
instance more high efficiency farmers cultivate on lowland, and on average, high
efficiency farmers cultivate less land (3.3 ha) than the low efficient farmers (9.7 ha),
another indication that low quality land is an important contributor to low efficiency.
Additionally, high efficiency farmers use more seeds and manpower than their low
efficiency counterparts. Finally, the findings suggest that high-efficiency farmers use
collective system, with respect to farm management type practices, compared to low
efficiency farmers.
4.47  There is no significant gender bias although women farmers seem to be more
efficient than male farmers. Why this is the case is not at this point clear and it is
unlikely to be empirically resolved using the existing data. But as we discussed, both
men and women in a family work in common plots and also have private plots.
Utilization of farm resources indicate that women farm less intensively than men: they
are less likely to mechanize tasks on their plots or use plough animals and less likely to
use fertilizer. As previous studies have shown, women farmers use systematically lower
levels of inputs for the same crop farmed within the same household as male farmers
which results in lower yields on female-managed plots and an overall loss in household
efficiency. (Udry, 1996 and Udry et al. 1995). Therefore, while women in Niger may be
disadvantaged when it comes to the quality and perhaps the size of plots they get for
private use, they may be very efficient in how they use resources to exploit these
plots-even though they produce lower yields.
4.48   A farmer's age does seem to matter for efficiency. From the estimates in Table
4.11, inefficiency decreases with age and then increases thereafter, suggesting that
younger farmers are more efficient than older farmers. However, as the coefficients are
very small, this difference disappears very early in the age distribution. In fact, when
evaluated at the mean age, it hardly exists at all. The results also suggest that farmers
who could meet at least 9 months of their family's food needs - that is those who are
self-sufficient in food consumption, are more efficient than the less self-sufficient. The
least efficient farmers are those who report meeting only 3 months of their family's
food needs. This should not surprise us at all. Farmers who can meet their nutritional
needs throughout the year are more likely to be stronger, healthier and more
productive and efficient, while the opposite to be true.
4.49  Education, agricultural services, remoteness and rainfall are among the key
correlates of agricultural efficiency. The findings show that education is not associated
with increasing efficiency. Since the coefficient is negative, this implies that those who
did not attend school are more likely to be efficient than those who have some
education. This is a surprising result, and a further exploration shows that this is mostly
-87-


true among female farmers, but not male farmers, because the average efficiency of
males who attended school is higher than those who did not (Figure 4.16). One possible
explanation could be that the educated who take up farming have very low levels of
education, or inappropriate skills, or low levels of experience in farming than those who
farm but have no education.  In one study, Vietnamese rice producers who had
vocational training were found to be more productive than farmers with only formal
primary and secondary education (Ulimwengu and Badiane (2010)). Hence, significant
productivity gains can be achieved through the promotion of education schemes
tailored to the specific technical needs of smallholder or poor farmers such as follow-up
extension services, farmer field demonstration plots, and short term technical training.
Figure 4.16: Efficiency by gender and education
E Educated oNot educated
0.732
0.715
0.670
0.631
Male                      Female
Source: World Bank staff estimates from NGAC
4.50  As we would expect agricultural mechanization is associated with higher levels of
efficiency. Notice that the coefficient on mechanization in Table 4.11 (panel B) is
positive. But this is a dummy variable for those who did not mechanize, suggesting that
they are more inefficient than those who did mechanize. Figure 4.17 plots the average
levels of efficiency for male and female farmers who mechanize compared to those who
did not. Both groups have higher average levels of efficiency than those who do not,
supporting the regression results.
Figure 4.17: Efficiency by gender and use of mechanization
1.00       m Practice mechanization
0.80  -     Dosatpa    ic
0.60  -
0.40  -
0.20  -
0.00
Male                     Female
Source: World Bank staff estimates from NGAC
-88-


4.51  Efficiency declines with remoteness. Our measure of remoteness is the distance
to the nearest food market. The results suggest that the level of efficiency increases the
further the farmer is from the nearest food market. But this is only up to about 2
kilometers (Figure 4.18). After 2 kilometers, efficiency declines with distance, thus
making the relationship an inverted U-shape (see Table 4.11). This makes intuitive
sense. We do not have knowledge of what was identified as the nearest food market,
but if it is the nearest town or nearest village market, then it is easy to imagine that
farmers that are very close to the town would be more efficient than farmers very far
away because they have lower costs of transportation (that is, less isolated), have better
access to inputs and information. However, up to the point at which efficiency declines,
it is possible that farmers a little further (say, 1.5 kilometers) are more efficient because
they can cultivate more land as they are not too close to the urban areas. But at the
same time they have access to just as much transportation, inputs and information as
those that are only 0.5 kilometers and therefore could produce more for the same levels
of inputs. And then of course the further the farmer is from the 2 kilometer threshold,
isolation sets in and technical efficiency declines.
Figure 4.18: Efficiency and market distance
0                 1                2                 3
Market distance
4.52  Farmers with more farming parcels are less efficient. In the extensive agricultural
production technologies of the Sahel, land is a valuable input. More parcels imply more
land and, more land means potentially more output. However, the finding that more
parcels lead to more inefficiency could arise for several reasons. One possibility is that
those with more parcels are in effect people with less good quality land, and are
offsetting good quality land with many poor quality pieces. More parcels could imply
that more individuals are involved in plot-level decision making and that input allocation
is being considered at the plot-level without concern for maximizing overall household
efficiency. The alternative is the possibility that more parcels mean many fragmented
parcels separated from each other by long distances. This increases transaction costs,
effort to monitor the work done on each parcel and reduces time spent on planting.
While we are not sure which of these situations prevails, the results suggest farmers
with multiple parcels are not utilizing them efficiently.
-89-


4.53   As we would expect, the findings confirm that rainfall is a key correlate of
efficiency (Figure 4.19).  In Niger, rainfall is as important an input to agricultural
production as any other, perhaps more so, because without it, no production can take
place. This is of course because agriculture in Niger is too rainfall-dependent. Rainfall
determines efficiency in the sense that it affects how farmers plan and optimize the rest
of their inputs - for instance, what should they plant if the weather report says that
expected rainfall is less than average? How much labor should be mobilized? Should
some land be left fallow? And so on. Therefore, yields are strongly correlated with
rainfall levels (see Figure 4.20 and 4.21), and it is not surprising that variability of rainfall
is found to be highly correlated with efficiency as well. Overall, efficiency rises with
more rainfall variability and then tapers or even falls as variance increases much more.
Figure 4.19: Efficiency and rainfall
0-
200            300           400            500            600
Rainfall (mm)
Figure 4.20: Sorghum yield and rainfall
Say
Kollo
Dogon-Doutchi
Madaoua                    Gar
Tillabery
Tchin-Ta bP( u  a          Niamey
ouza   Loga         Boboye
-a u        a      Bkonn,  Madarounfa
TchighW-WAi11ni             Magaria
Tessaous GroAVk"
Tanout     Filingu9II61a
Goure Matam-ye
MVW6-Sor,  Mayahi
0-
200         300         400         500         600
Average rainfall (mm)
-90-


Figure 4.21: Millet yield and rainfall
Madarounfa
Macaoua
Kollo          Say
Grourndji
Diffa   M     Tbr u4Aagar guiE  Niamey
Maine-SorofiDakBk
IVN'Gul           TIMiy9fht  MatAw0u-Doutchi
>:                HFilinguE                Dosso
Tanout     Oualam       L
TiMPIWfnabarade
Thighozerine
0
200         300        400         500        600
Average rainfall (rnm)
Source: Author's data from http://harvestchoice.org:8080/geonetwork/srv/en/main.home
4.54   The expected impact of farm       investments on efficiency is not homogenous
across investments, nor is it linear. One key investment is of course managerial effort.
Table 4.11 shows that farmers who do not practice sound crop management are
significantly more inefficient than those who do. Other investments are use of modern
inputs that would increase productivity and when properly applied, technical efficiency.
Figures 4.22a to 4.22d capture basic features of the relationship between farming
efficiency and on-farm investments. There is a critical level of seed investment where it
starts yielding a positive effect on farming efficiency. For example, investments in
fertilizer depict an inverse-U curves suggesting declining efficiency beyond a certain
level of investments. Only crop management and manure investment display a
monotonically increasing trend with farming efficiency. Though simple, these features
suggest that strategies to boost farming efficiency should account for heterogeneity of
the effects of individual investments.
Figure 4.22: Relationship between farming efficiency and on-farm investments
Figure 4.22a: Efficiency and seeds           Figure 4.22b: Efficiency and fertilizer
0     10000  20000   30000  40000  50000
Sees expn,ire (FGFA)          0           00         0001500001
Ferblizer expenditure (FCFA)
-91-


Figure 4.22c: Efficiency and crop management  Figure 4.22d: Efficiency and manure
6000    7000    8000    9000    10000    6000    7000    8000   9000    10000
Crop management expenditure (FCFA)       Crop manage nnt expenditure (FCFA
F.     Government spending and long term growth and poverty targets
4.55   While the debate over the direction and magnitude of the impact of public
spending on growth remains heated, growing evidence at both macroeconomic and
microeconomic levels asserts that carefully targeted public investments achieve
development results. In particular, public investment focused on areas characterized by
market failures or public goods externalities can yield high rates of return and
substantive benefits. In particular, public investments in basic infrastructure, human
capital formation and research and development (R&D) are necessary conditions for
other investments as they (public investments) promote technology adoption, stimulate
complementary on-farm investment and input use and are needed for marketing the
agricultural goods produced.
4.56   Several analytical and empirical studies in the growth literature have focused on
understanding how different types of public spending can affect growth through both
traditional and new channels. A direct effect relates to an increase in the economy's
capital stock (physical or human) reflecting higher flows of public funds, especially when
they are complementary to those privately financed. Public investment can also
contribute to growth indirectly by increasing the marginal productivity of both publicly
and privately supplied factors of production. For example, public expenditure on
agricultural research and development (R&D) can promote higher productivity by
improving the interaction between physical and human capital production inputs. Other
components of public spending, related for instance to the enforcement of land
property rights, can also exert a positive indirect effect on growth by contributing to
better use of existing assets. There is also growing evidence in developing countries that
externalities associated with infrastructure public spending may have non-trivial
linkages on human capital development.
4.57   Current research argues that agricultural investments are strongly pro-poor.
More specifically, the claim is that public agricultural investment is the most direct and
-92-


effective way for African countries to ramp up agricultural growth, which is a necessary
condition for achieving the first Millennium Development Goal of reducing absolute
poverty (Fan and Rosegrant (2008)). Aside from theory, some of the arguments are
based on recent studies that have documented a significant impact of agricultural
expenditures on agricultural growth and poverty reduction in India and China (Fan,
Hazell and Thorat (2000), Zhang and Fan (2004) and Fan, Zhang and Zhang (2002)).
Similar findings on the link between agricultural growth and agricultural development
expenditure are reported for several African countries, including Rwanda, Zambia, and
the entire continent of Africa (Diao et al., 2007; Thurlow et al., 2008; Fan and Rao,
2003). Strong linkages between agricultural growth and agricultural research are also
reported in Uganda (Fan et al., 2004).
4.58  In Niger, despite commitments to reduce rural poverty agricultural spending has
lagged behind spending in other sectors. From 1993-2007, agricultural sector (including
crop production, livestock, fishery, and forest) spending accounted for 12 percent6 of
the overall government budget compared with 23 percent and 16 percent, respectively,
for education and health. Both overall agricultural spending and agricultural investment
have been increasing at a rather slow pace. Total agricultural spending rose from 30.3
billion in 1994 (in 2005 FCFA) to 38.9 billion in 2007 (Figure 4.23). Similarly, agricultural
investment increased from 26.0 billion in 1994 (2005 FCFA) to 34.1 billion in 2007 (in
2005 FCFA).
Figure 4.23: Spending in agricultural sector
45.00
40.00
35.00
30.00
.225.00
a20.00
15.00
-   Ag investment
10.00
-Total ag spending
5.00
0.00
1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007
Source: World Bank staff estimates from INS data
4.59  In nominal terms, budget allocations to the sector declined significantly in 2009
(Table 4.13). The downward trend is mainly due to a sharp decrease in investments for
'6 This amount includes donors' budget support.
-93-


crop production from   23.7 billion FCFA in 2008 to 2.6 billion in 2009, a year
remembered for a severe drought that wiped out most of the agricultural output.
Table 4.13: Government budget allocation (%)
2003     2004    2005    2006     2007    2008      2009
Crop production    6.1      8.8     7.8      8.4     5.4      6.9      1.6
Livestock          0.2      0.6     0.4      0.6     1.4      2.8      1.1
Health             9.1      6.4     6.8      8.2     9.2     11.3      8.5
Education         17.9     17.6    17.3     18.6    17.8     20.6     24.1
Social protection  0.5      0.5     0.4      0.5     0.6      0.7      0.3
Other             66.2     66.2    67.2     63.7    65.7     57.6     64.3
Total             100.0   100.0   100.0    100.0    100.0   100.0    100.0
Source: INS data.
4.60   External development assistance is provided largely through projects and
captured in the investment budget. Assistance targets agriculture, education, and heath
- sectors considered instrumental for reducing poverty. Over the 2001-2008 period,
health, agriculture and education accounted for 47.3 percent of the total spending on
various projects across the country (Figure 4.24). Intra-sectoral spending levels vary
dramatically. Within agriculture, almost 91 percent of resources were targeted to crop
production, followed by about 9 percent to livestock and almost nothing - less than one
percent to the fishery sub-sector.
Figure 4.24: Disbursement of projects funds by sectors (%)
Other                                           40.8
HHealth                          24.9
Hydraulic          9.1
Education        6.2
Infrastructures L2.7
Food security  0P2
Source: Ministry of Finance
4.61   Table 4.14 summarizes key agricultural investments realized through projects
from 2001 to 2009. These investments include cereal banks, dams, feeder roads,
irrigation, reclaimed land and wells. As expected, the construction of dams has the
highest unit cost (76.7 million FCFA) followed by irrigated land (48.3 million FCFA) and
feeder road (28.6 million FCFA).
-94-


4.62   Investment efficiency is low in comparison to similar projects. The average cost
estimated for wells ($USD 19,484.9) falls within the range of average costs cited by
Africon (2008), for building an electric well ($USD 14,112-54,701) but out of range for a
well without pump or an electric well with hand pump. In particular, the average cost of
feeder road construction ($USD 55,175.3) is almost three times as expensive as the cited
range for re-gravelling projects ($USD 12,835-19,490).
Table 4.14: Key agricultural investments (2001-2008)
Amount (million                 Unit cost
Investments                    FCFA)       Quantity   (million FCFA)
Cereal banks                  8223.0          377.0           21.8
Dam                            230.0            3.0           76.7
Feeder road (Km)              1070.0           37.4           28.6
Irrigation network (liter)     10.0           130.1            0.1
Irrigated land (Ha)           8600.0          178.0           48.3
Reclaimed land (Ha)           2558.0         1279.0            2.0
Wells                         7475.0          741.0           10.1
Source: Ministry of Finance
Agricultural Research
4.63   The case for investing public money in agricultural research and development
(R&D) is compelling since market failures and public externalities abound. These
activities are crucial to ramping up productivity and promoting efficiency of resource
use. Yet, R&D spending has been erratic during the 1980s and 1990s and it plunged
dramatically in 1998 (see Figure 4.25). By 2008, Niger's investments in agricultural R&D
had dropped by 80 percent, totaling no more than approximately 1.4 billion 2005 FCFA
(Stads, Issoufou, and Massou, 2010).
Figure 4.25: Public agricultural R&D spending (Billion 2005 FCFA)
8.0
7.0
6.0
no
.E 5.0
4.0
ca 3.0
2.0
1.0
0.0
Source: Stads, Issoulou, and Massou (2010).
-95-


Spending, Long term Growth and Poverty Targets7
4.64   In an effort to increase investment as a way to accelerate rural sector growth
and reduce rural poverty, the government of Niger joined the Comprehensive African
Agricultural Development Program       (CAADP)'8 and prepared a National Agricultural
Investment Program (NAIP) for review in June 2010. The proposed investment plan is
based on the country's rural development strategy (SDR). The NAIP is dominated by
water and sanitation, land reclamation and reforestation, and rural infrastructures
which account for 56 percent of the total (Table 4.15). Still, agricultural research
represents less than one percent of the overall proposed budget. The NAIP total budget,
as reported in Figure 4.26, ranges between the long term funding levels required to
achieve poverty reduction targets outlined in the MDG 1 by 2020 and at a more
accelerated pace by 2015.
Figure 4.26: NAIP, CAADP and MDGI Budgets (billions of FCFA)
1878.7
1130.2
488.2
349.6
208.7
SDR        CAADP      MDG1-2020       NAIP     MDG1-2015
Source : Sunday et al. (2010)
17 From Sunday, 0., J. Ulimwengu and 0. Badiane. 2010. Niger compact review: Does Niger national agricultural
investment plan (NAIP) meet the Long term growth and poverty outcome benchmarks established during the
roundtable and underlying the compact? Manuscript.
19 Overall, CAADPs goal is to eliminate hunger and reduce poverty through agriculture. To do this, African
governments have agreed to increase public investment in agriculture by a minimum of lo per cent of their national
budgets and to raise agricultural productivity by at least 6 per cent.
-96-


Table 4.15: Composition of Niger National Agricultural Investment Program
Programs and sub-programs                                   Billion,   Million,   Share
FCFA     $USD"        (%)
Local and community development                               135.0      265.7     11.9
Local governance of natural resources                          37.0       72.8      3.3
Professional organizations and structuring of sub-sectors      32.5       64.0      2.9
Rural infrastructures                                         144.4      284.2     12.8
Rural financial system                                         12.2       24.0      1.1
Research-Training-Outreach                                     10.3       20.3      0.9
Capacity strengthening of rural public institutions             5.7       11.2      0.5
Water and sanitation                                          272.5      536.2     24.1
Reduction of households vulnerability                          72.0      141.7      6.4
Environmental preservation                                     32.0       63.0      2.8
Food security through development of irrigation                25.0       49.2      2.2
Pastoral management and security of pastoral systems           23.0       45.3      2.0
Land reclamation and reforestation                            221.5      435.9     19.6
Kandadji - ecosystem restoration and enhancement of the Niger  107.1     210.8      9.5
valley
Total                                                        1130.2     2224.1    100.0
Source: S6cr6tariat Ex4cutif/SDR (2010)
4.65   At the signing of the CAADP compact, total GDP, agricultural GDP, and non-
agricultural GDP were all expected to grow by 4.4 percent, 6.2 percent, and 2.9 percent,
respectively, by 2015. The scenario was based on a target of reducing poverty from 59
percent to 52.9 percent by 2015, even though the targets were not sufficient for Niger
to achieve the poverty MDG of halving poverty by 2015. Reducing poverty at this pace
would have required an unrealistic agricultural growth rate of 11.5 percent per annum.
The post-compact National Agricultural Investment Plan (NAIP) injected new funding
levels, and total GDP, agricultural GDP, and non-agricultural GDP have since been
estimated to grow by 5.0 percent, 7.4 percent and 3.0 percent, respectively by 2015.
Figure 4.27, highlights the growth rates associated with each strategic spending plan.
Figure 4.27: Growth rates by scenarios
14.0
E Overall  0 AgGDP    * Non-ag GDP
12.0
710.0
8.0
- 6.0
(2  4.0
2.0
0.0
SDR     NIAP-2015   CAADP    MDG1-2015 MDG1-2020
Source: Sunday et al. (2010).
Exchange rate: 508.18 FCFA for 1 $uS.
-97-


Irrigation Investment
4.66   Earlier simulations conducted by the World Bank (2007) indicate that positive
productivity shocks on irrigated export crops and irrigated food crops are likely to
generate the greatest gains in household welfare. Niger is an arid country where
agricultural harvests respond highly to climatic outcomes. The reliable supply of
irrigation is among several key requirements for sustained growth in agricultural
productivity. Improved access to irrigation with complementary agricultural investments
would remove several bottlenecks to agricultural growth. Just by itself, improving access
to irrigation infrastructure is not sufficient to enhance agricultural productivity.
Improving rural infrastructure in general (e.g., irrigation, roads, telecommunications,
water and sanitation, etc.) is critical for sustained growth and poverty reduction.
Additionally, farmers would require increased access to extension services to more
effectively utilize new irrigation technologies. Irrigation interventions must account for
the physical construction and the human dimension of management since irrigation
schemes usually are based on collective action.
4.67   Through the SDR (specifically, sub-program 4.1 of the SDR), agricultural and
water infrastructure, which incorporates mainly the national strategy for irrigation, the
government's objective - or more appropriately, aspiration - is to improve the
contribution of irrigated agriculture to agricultural GDP from 14% in 2001 to 28% in
2015 (Executive Secretariat, 2010). To this end, the government is planning to invest
about 147 billion FCFA in irrigation over the 2006-2015 period. A separate irrigation
related program called "Kandadji" exists, for which the government is expected to
spend about 138.9 billion FCFA during the same period. Compared to the past trend,
these are relatively large public expenditures for irrigation, and their operationalization
will therefore require evidence-based strategies and well-functioning monitoring and
evaluation system to ensure the highest possible return.
4.68   Given the ambitious nature of the planned irrigation investment program, we
use a CGE model to simulate the potential impacts of irrigation expansion on agricultural
growth. We make no effort to evaluate the feasibility of such an investment being
realized. Because the investment program was arrived at through multiple high level
CAADP consultations, we take the proposed plans as given. Table 4.16 presents key
parameters used for this simulation exercise. We first used econometric results from
Kibonge (2010) showing that a one percent change in the share of irrigated land will
significantly increase agricultural efficiency by 0.019 percent across SSA.
-98-


Table 4.16: Key simulation20 parameters
Change in share of irrigated  Change in overall crop  Size of irrigated  Share of irrigated
land (%)                efficiency (%)           land' (ha)          landa
10                       0.186                  7097.9             0.189
30                       0.558                  8388.4             0.223
Note: a compared to 2006.
4.69   We implement both the efficiency change and the change in the size of irrigated
land under two scenarios. Scenario I assumes a 10 percent change in the share of
irrigated land and scenario II assumes an increases in the share of irrigated land to 30
percent unless otherwise specified, we run the simulation through 2015. The simulation
results suggest that expansion of irrigated land will trigger a positive change in TFP
across all agricultural sub-sectors (Table 4.17).
Table 4.17: Annual average TFP growth (%)
Scenario I Scenario II
Food crop
Irrigated                0.23      0.69
Non-irrigated            0.26      0.79
Export crop
Irrigated                0.23      0.69
Non-irrigated            0.26      0.79
Livestock                0.26      0.78
Forestry                 0.28      0.84
Source : World Bank staff estimates
4.70   Expansion of irrigated land is expected to boost agricultural growth and overall
economic growth (Table 4.18). As expected, the highest impact is observed in the
irrigated sub-sector; 17.6% for irrigated export crops and 11.5% for irrigated food crops.
However, competition over limited inputs will temper the growth trend in non-irrigated
and other agricultural sub-sectors. This calls for an efficiency arbitrage in the design and
implementation of a government-led irrigation strategy.
20 The team is grateful to Dr. Odjo of IFPRI for assistance with running the simulations.
-99-


Table 4.18: Impact of irrigation expansion on growth rates (%)
BASE     Scenario I  Scenario II
Overall GDP             3.5        4.0        5.8
Agricultural GDP        4.4        5.4        9.2
Food crops             4.4         4.7        5.8
Irrigated               4.5        6.3       11.1
Non-irrigated           4.3        4.2        3.2
Export crops            5.8       11.5       22.6
Irrigated               7.8       17.6       31.8
Non-irrigated           4.0        3.4        1.4
Other agriculture       4.0        3.3        2.3
Livestock               4.1        3.3        2.3
Forestry                3.2        3.1        2.4
Non-agriculture        2.9        2.8         2.7
Source : World Bank staff estimates
4.71   The simulated scale of poverty reduction is large. As shown in Figures 4.28a-
4.28c, expansion of irrigated land will foster poverty reduction across rural and urban
locations. As expected, the highest rates of poverty reduction are predicted from the
second scenario which assumes an expansion of land under irrigation to 30 percent. In
the case where land under irrigation increases by 10 percent, national headcount ratios
- national, rural and urban - will decline only marginally. Over a decade, poverty is
simulated to decline by only 5 or so percentage points. However, should irrigated land
increase by 30 percent, poverty is predicted to decline by almost 20 percentage points,
and it declines even faster in rural areas.
Figure 4.28a: Impact of irrigation on national poverty rates by scenarios (%)
70.00
60.00
50.00
40.00
-BASE
30.00
-Scenario I
20.00                    -   Scenano II
10.00
0.00
2004 2005 2006 2007 2008 2009 2010 20112012 2013 2014 2015
Source : World Bank staff estimates
-100-


Figure 4.28b: Impact of irrigation on rural poverty rates by scenarios (%)
80.00
70.00
60.00
50.00
40.00
30.00              -BASE
20.00
10.00
0.00                               I
2004 2005 2006 2007 2008 2009 2010 20112012 2013 2014 2015
Source : World Bank staff estimates
Figure 4.28c: Impact of irrigation on urban poverty rates by scenarios (%)
-BASE
Scenario I
-Scenario II
Source : World Bank staff estimates
4.72   The simulated results are consistent with other findings that show that African
and Asian investments in irrigation, agricultural research and extension have large
impacts on agricultural productivity and poverty reduction, and investments in rural
infrastructure can bring even greater benefits (Fan, 2010). In addition, investing in food
crops, which are mainly produced by smallholder farmers in Niger, will be particularly
effective in reducing poverty and in generating economic growth linkages. Reliable food
production would significantly improve food security and reduce vulnerability in rural
areas. Furthermore, expansion of crop production in Niger should facilitate broader
growth through new opportunities for increasing exports and processing agricultural
products.
G.     Policies for improving agricultural production and increasing
returns to the Poor
4.73   In this chapter we have shown that after a slow start, Niger's agricultural
productivity has turned for the better, as measured using TFP. In addition, areas of
promise in the agricultural sector are emerging. Two such activities are the emerging
-101-


community-based institutions for distribution of inputs, especially seeds and the
growing diversification into high-value crops, such as the cowpeas and other
horticultural crops. However there remains major problems, not least the enduring low-
productivity status of Nigerien female and male farmers.
4.74  Part of the explanation for the low productivity is the reality that many of the
farmers live in a high risk environment and dealing with an uninsurable risk - weather
risk - at least to date. Other problems that compound this constant risk are the non-
functioning markets for inputs, for savings, and protection of income. Adding to the pile
of problems is government investments that are unstable, often inadequate, and poorly
executed. The result is volatile growth of agriculture and widespread vulnerability and
poverty in Niger.
4.75  There is no doubt that pro-poor growth must begin with robust agricultural
growth. However, while this has proven elusive in the past, the aspirations remain and
the new vision offers some prospects for success. First, a rigorous peer evaluation and
review process (CAADP) for the government investment program in the sector provides
the hope that spending will be anchored on strong governance and accountability
framework and monitored in a timely and transparent fashion. Second, a search for
improving agricultural productivity would have to consider investments in new crops
(such as horticulture) and new ways to deliver information (say agricultural extension
through mobile phones). Strategies to improve access to markets, information, and new
technologies must reach out to both male and female producers. Without being
inclusive, these strategies are likely to fail since women use substantially lower amounts
of inputs and they obtain lower yields. Low spending on agricultural research and
investment and a dysfunctional agricultural extension system suggest that Niger stands
to fall behind in efficiency and lose gains recently made to catch up to its technological
frontier. The expansion of irrigated agriculture offers the promise of improving returns
to the sector. However, given that scaling up in the short term, especially accessing land
that is suitable for irrigation, is constrained by human and hardware capital, careful
choices must be made regarding who receives access. Kandadji Dam is expected to
generate an additional 1000 hectares of land for irrigation per annum over the next 30
years. The allocation strategy of who receives these parcels should be transparent.
Monitoring of agricultural performance hinges on having reliable and comprehensive
data of the agriculture sector and the performance of households operating within the
sector. Finally, finding new forms to deliver insurance - weather, crops, or income - will
go a long way to protecting households and reducing their current levels of risk
aversion.
-102-


References
Africon. 2008. Unit costs of infrastructure projects in Sub-Saharan Africa: Summary of
Background Paper No. 11 for the Africa Infrastructure Country Diagnostic (AICD)
project. Washington, D.C.: World Bank. <http://www.eu-africa-infrastructure-
tf.net/attachments/library/aicd-background-paper- 11 -unit-costs-summary-en.pdf>.
Accessed September 2010.
Aker, J. 2009. "Information from Markets Near and Far: Information Technology, Search Costs
and Grain Markets". Mimeo
Aker. J. 2010. "Dial "A" For Agriculture: Using Information and Communication Technologies
for Agricultural Extension in Developing Countries". Tufts University. mimeo
Ali, A. T. Leber, and A. Amani (2005). "Rainfall Estimation in the Sahel. Part I: Error Function,
Journal of Applied Meteorology, November 2005, pp. 1691-1706.
Backiny-Yetna, Rospere, Quentin Wodon, Rose Mungai, and Clarence Tsimpo (2008).
Poverty in Liberia: Level, Profile and Determinants. The World Bank.
Battese, G.E. & Coelli, T.J. (1995) A Model for Technical Inefficiency Effects in a
Stochastic Frontier Production Function for Panel Data. Empirical Economics,
20(2), pp. 325-332.
Cleaver, K. M., and W. G. Donovan. 1995. Agriculture, poverty, and policy reform in
Sub-Saharan Africa. World Bank Discussion Papers, Africa Technical
Department Series, No. 208, World Bank, Washington, D.C.
Cooper, B. (1997). Marriage in Maradi: Gender and Culture in a Hausa Society in
Niger, 1900 - 89, published by James Currey.
Diao, X., P. Hazell , D. Resnick and J. Thurlow. (2007). "The Role of Agriculture in
Development: Implications for Sub-Saharan Africa," Research Report 153. International
Food Policy Research Institute, Washington, D.C.
Dorosh, Paul (2008). "Production Shocks, Regional Trade and Cereal Prices in Niger". A Report
Prepared for the World Bank, Washington, D.C.
Evenson, R.E. and A. F. Dias Avila, 2007. FAO Data-Based TFP Measures In: Chapter
31 Vol. 3 Handbook of Agricultural Economics: Agricultural Development:
Farmers, Farm Production and Farm Markets, Eds: R.E. Evenson, P. Pingali and
T.P. Schultz, Elsevier.
Fan, S. (2010). Halving Hunger: Meeting the First Millennium Development Goal
through "Business as Unusual". Food Policy Report. International Food Policy
Research Institute Washington, D.C.
-103-


Fan, S. and M. W. Rosegrant. (2008). "Investing in Agriculture to Overcome the World
Food Crisis and Reduce Poverty and Hunger," IFPRI Policy Brief No. 3.
International Food Policy Research Institute: Washington, DC.
Fan, S. and N. Rao. (2003). "Public spending in developing countries: trend,
determination and impact," Environment and Production Technology Division
Discussion Paper 99. International Food Policy Research Institute: Washington,
DC.
Fan, S., P. Hazell and S. Thorat. (2000). "Government Spending, Growth and Poverty in
Rural India," American Journal ofAgricultural Economics 82.
Fan, S., X. Zhang, and N. Rao. (2004). "Public expenditure, growth, and poverty
reduction in rural Uganda," Development Strategy and Governance Division
Discussion Paper 4. International Food Policy Research Institute: Washington
DC.
Fan,S., L. Zhang, and X. Zhang. (2002). "Growth, Inequality, and Poverty in Rural
China: The Role of Public Investments," IFPRI Research Report 125.
International Policy Food Research Institute: Washington, DC.)
FAO (Food and Agriculture Organization of the United Nations). 2009. FAOSTAT
database. <http://www.fao.org/>. Accessed February 15, 2009.
Foster, J. E., J. Greer, E. Thorbecke, (1984). A Class of Decomposable Poverty Indices,
Econometrica, 52: 761-766.
Guengant, JP (2001). "Comment B6n6ficier du Dividende D6mographique?" La
d6mographie au centre des trajectoires de d6veloppement dans les pays de
l'UEMOA Analyse pays Niger: Published by Agence Francaise D6veloppement
February, 2011.
Fuglie, K.O. (2008). Is a slowdown in agricultural productivity growth contributing to the
rise in commodity prices? Agricultural Economics 39 (1): 431 -441.
Institut National de la Statistique (2008). Tendances, profil et d6terminants de la pauvret6
au Niger 2005-2007/08. Minist&re de l'6conomie et des finances, Niamey.
Institut National de la Statistique (2006). Questionnaire des indicateurs de base du bien-
6tre, Profil de pauvret6 - Rapport d'analyse. Minist&re de l'6conomie et des
finances, Niamey.
Institut National du Statistique (INS) and World Bank (WB). 2006. << Questionnaire des
Indicateurs de Base du Bien-6tre (QUIBB 2005): Profil de Pauvret6 >>. Niamey,
Niger.
INS. 2008. << Tendances, Profil et D6terminants de la Pauvret6 au Niger: 2005-2008 >>
Niamey,Niger.INS metadata: http://www.ins.ne/nada/index.php?page=home.html
-104-


Jones, C. (1986). "Intra-household Bargaining in Response to the Introduction of New
Crops: A Case Study from North Cameroon." in Understanding Africa's Rural
Households and Farming Systems, edited by Joyce Lewinger Moock. Boulder,
Colorado: Westview.
Kibonge, A. 2010. Agricultural Productivity and Climate Change in Sub-Saharan Africa.
Paper presented at the Agricultural & Applied Economics Association 2010
AAEA, CAES, & WAEA Joint Annual Meeting, Denver, Colorado, July 25-27,
2010.
Kumbhakar, S.C. & Lovell, C.A.K. (2000) Stochastic Frontier Analysis. Cambridge
University Press, Cambridge.
Lanjouw, Jean Olson, and Peter Lanjouw. 2001. "How to Compare Apples and Oranges:
Poverty Measurement Based on Different Definitions of Consumption." Review of
Income and Wealth: 47(1): 25-42.
Lanjouw, Peter, Jill Luoto, and David McKenzie. 2009. "Using Repeated Cross-Sections
to Explore Movements in and out of Poverty." Mimeo.
Liu, Y. & Myers, R. (2009) Model Selection in Stochastic Frontier Analysis with an
Application to Maize Production in Kenya. Journal of Productivity Analysis,
31(1), pp. 33-46.
Lusigi, A., and C. Thirtle. 1997. Total factor productivity and the effects of R&D in
African agriculture. Journal of International Development 9: 529-38.
Minist&res du D6veloppement Agricole et des Ressources Animales. (2008).
Recensement g6n6ral de l'agriculture et du cheptel (RGAC 2004/2008).
Gouvernement du Niger.
Mogues, T., G. Ayele and Z. Paulos. (2008). "The Bang for the Birr: Public Expenditures
and Rural Welfare in Ethiopia, " IFPRJ Research Report 160. International Food
Policy Research Institute: Washington, DC.
Nin, A. and B. Yu (2008). "An Updated Look at the Recovery of Agricultural
Productivity in Sub-Saharan Africa," IFPRI Discussion Paper 00787.
Washington, DC: International Food Policy Research Institute.
Ravallion, Martin (1998). Poverty Lines in Theory and Practice, LSMS Working Paper
no. 133. The World Bank, Washington, D.C.
Ravallion, M et G. Datt (1991), Growth and redistribution components of changes in
poverty measures : a Decomposition with Application to Brazil and India in the
1980s, LSMS Working Paper, no 83. The World Bank, Washington.
-105-


Reardon, T. C. Delgado, and P. Matlon (1992), "Determinants and Effects of Income
Diversification amongst Farm Households in Burkina Faso," Journal of
Development Studies, Vol.28 (January 1992) pp. 264 - 296.
Reardon, T. and J.E. Taylor, "Agroclimatic shock, Income Inequality, and Poverty:
Evidence from Burkina Faso," World Development, Vol24, No.5 pp. 901- 914.
Reardon, T, JE Taylor, Stamoulis, K, Lanjouw, P. , Balisacan (2000), A. "Effects of
Non-Farm Employment on rural Income Inequality in Developing Countries: An
Investment Perspective", Journal of Agricultural Economics Vol51, Number 2,
pp266 -288.
S6cr6tariat Ex6cutif/SDR. 2010. Plan d'Investissement PNIA/SDR Niger. Gouvemement
du Niger.
Stads, G. J., M. H. Kabaley, and M. Gandah. 2004. Niger. ASTI Country Brief No. 24.
Washington, D.C. and Niamey: IFPRI and INRAN.
Stads, G.-J., M. Issoufou and A. M. Massou. (2010). Niger: Recent developments in
agricultural  research.  ASTI     country    note.   Available   at:
http://www.asti.cgiar.org/pdf/Niger-Note.pdf.
Sunday, 0., J. Ulimwengu and 0. Badiane. 2010. Niger compact review: does Niger
national agricultural investment plan (NAIP) meet the Long term growth and
poverty outcome benchmarks established during the roundtable and underlying
the compact? Manuscript.
Tarozzi, A. (2007). "Calculating Comparable Statistics From Incomparable Surveys,
With an Application to Poverty in India. Journal of Business and Economic
Statistics, Volume 25, No. 3, pp. 314-36.
Thurlow, J., S. Benin, X. Diao, H. Kalinda and T. Kalinda. (2008). "Agricultural Growth
and Investment Options for Poverty Reduction in Zambia," IFPRI Discussion
Paper 00791. International Food Policy Research Institute: Washington, DC.
Udry, C. (1996). "Gender, Agricultural Production, and the Theory of the Household."
Journal of Political Economy, Volume 104, Number 5, pp. 1010-46.
Udry, C. et al (1995). "Gender Differentials in Farm Production: Implications for
Household Efficiency and Agricultural Policy. "Food Policy Vol 20 (October
1995).
Ulimwengu, J. and 0. Badiane. (2010). Vocational Training and Agricultural
Productivity: Evidence from Rice Production in Vietnam. Journal of Agricultural
Education and Extension, Vol. 16, No. 4, 399-411.
-106-


Von Braun, J. and P. Webb (1989). "The Impact of New Crop Technology on the
Agricultural Division of Labor in a West African Setting." Economic
Development and Cultural Change 37 (April): pp513- 34.
WFP .2010. Report Executive Summary. "Enqu6te sur la S6curit6 Alimentaire des
M6nages au Niger (Avril 2010): R6sume Ex6cutif (mai 2010).
World Bank. 2001. Engendering development through gender equality in rights,
resources, and voice. World Bank Policy Research Report. Washington, D.C.:
World Bank.
World Bank, (2011). Niger: Rural Financial Services Expanding Financial Access to the Rural
Poor
World Bank,(2009). Niger: Food Security and Safety Nets: Final Report.
World Bank, (2009b) Niger: Public Expenditure Tracking Survey
World Bank (2008). Niger: Modernizing Trade During a Mining Boom. Diagnostic Trade
Integrations Study for the Integrated Framework Program.
World Bank (2007). Niger Country Economic Memorandum: Accelerating Growth and
Achieving Goals: Diagnosis and the Policy Agenda
World Bank (2006). Carte de Pauvretd pour le Niger. Unpublished mimeo. Q. Wodon.
World Bank (2005). "Providing Nigeriens with Food, Education, and Heath Care: a Demographic
Perspective".
World Bank. 2009. "Niger: Food Security and Safety Nets" (Report No. 44072-NE) Washington
D.C.
World Food Programme (WFP), 2010. "Chocs et Vuln6rabilitd au Niger: Analyse des Donndes
Secondaires." (Available at www.wfp.org/food-security)
Zhang, X. and S. Fan. (2004). "How Productive is Infrastructure? A new approach and evidence
from rural India," American Journal ofAgricultural Economics 86: 494-50 1.
-107-


-108-


Annex 1: Figures and Tables: Vulnerability and Resilience.
Figure Al.1: Regional millet prices
Millet
250
200
m Oct-05
n Nov-06
o Dec-07
0 Nov-08
'3                                                n Dec-09
100
* Apr-10
50
Agadez  Diffa   Dosso  Maradi  Naney  Tahoua  1illaberi  Zinder
Figure A1.2: Global Agricultural Price Indices
Ag ricu4tuw* PriceIx km6ca!F
300Food
150
an-15  Jan-06      La-07    Jn-OB     Ia   O9   Jan-10     Jan-11
Source: World Bank - Global Economic Monitor
-109-


Figure Al.3: Opinion of current agricultural season
Opinion of current agricultural season
80%
70%
E 60%
0
0
 50%
80%
4 70%
C
.  20%
o
0
10% -
0%
2007                2008                2010
Source: World Bank staff estimates from ECVAM
Figure A1.4: Household views of harvest levels relative to previous year's
rain fed agricultural season
80% -
U 2007/08 M 2008/09 IN2009/10
70%
0
6~ 0%
0
-c 50%
40%
.4-30
0
C
aj 20%
CL 10%                    E
0%
Higher       Same level       Lower       Not applicable
Source: World Bank staff estimates from ECVAM
-110-


Figure Al.5: Number of days (out of last 7 days) in which households
lacked food or money for food
50
40
m 7 days
M. 6 days
530-
M 5 days
* 4 days
--20   -                                             m3 days
a 2 days
10                                                m 1 day
0
Rural     Urban     Rural     Urban
2008                2010
Source: World Bank staff estimates from ECVAM
Figure A1.6: Households reporting higher departures than usual of household
workers in last 30 days
16
Rural
14 -A
-1- Urban
4.
2
0
2007                 2008                2010
Source: World Bank staff estimates from ECVAM
-111-


Figure A1.7: Number of days (out of last 7 days) in which households consumed
less preferred food
60
60 -
S7 days
40 -6 days
I5 days
30   -                                            4 days
M 3 days
o20   -
m days
10m                                                 day
0
Rural    Urban     Rural     Urban
2008                2010
Source: World Bank staff estimates from ECVAM
Figure Al.8: Number of days (out of last 7 days) in which households borrowed
food from parents, neighbors, or friends
25
20-
m7 days
- 6 days
U515d
5 days
M 4 days
=10 - I3 days
A  2 days
5 - H Iday
0
Rural    Urban      Rural    Urban
2008                2010
Source: World Bank staff estimates from ECVAM
-112-


Figure Al.9: Number of days (out of last 7 days) in which households
bought food on credit
30
25
7 days
20                                               6 days
-  5 days
M 4 days
M 3 days
M 2 days
M 1day
5
0
Rural    Urban     Rural     Urban
2008                2010
Source: World Bank staff estimates from ECVAM
Figure A1.10: Number of days (out of last 7 days) in which households had
to depend on food aid
10
8
7 days
- 6 days
.5 days
I4 days
4                                             M 3 days
= 2 days
2                                             M 1 day
0
Rural    Urban      Rural    Urban
2008                2010
Source: World Bank staff estimates from ECVAM
-113-


Figure Al.11: Number of days (out of last 7 days) in which households
skipped payment of debts to purchase food
14  -
12-
10                                             m 7 days
=                                            E 6 days
M 5 days
a                                                  m4 days
4 M 3 days
2                                             = 2 days
M I day
Rural    Urban    Rural    Urban
2008               2010
Source: World Bank staff estimates from ECVAM
Figure Al.12: Number of days (out of last 7 days) in which households had to ask
other households for food for the children
16
12  -7 days
- 6 days
M 5 days
M 4 days
6 -M 3 days
M 2 days
M   day
0
Rural    Urban     Rural     Urban
2008                2010
Source: World Bank staff estimates from ECVAM
-114-


Figure Al.13: Number of days (out of last 7 days) in which households were forced
to beg due to food insecurity
5
4
m7 days
6 days
-                                               5 days
M 4 days
o
- 2                                             M 3 days
0
M2 days
1  -mi day
0
0  Rural     Urban     Rural     Urban
2008                2010
Source: World Bank staff estimates from ECVAM
Figure A1.14: Number of days (out of last 7 days) in which households
decreased daily rations of adults to benefit children
35
30
25 - a7 days
6 days
20M 5 dlays
M 4 days
M 3 days
a 2 days
M1 day
5
0     Rural     Urban     Rural     Urban
2008                2010
Source: World Bank staff estimates from ECVAM
-115-


Figure Al .15: Number of days (out of last 7 days) in which households
were forced to buy meals in order to save money
16
14-
12 -
7 days
10 -
- 6 days
Cm 5 days
o)   6 -                                        m4 days
4 -                                          3 days
2 -                                          2 days
1 1day
0
Rural   Urban    Rural    Urban
2008              2010
Source: World Bank staff estimates from ECVAM
Figure Al .16: Number of days (out of last 7 days) in which households decreased
the number of meals per day
30
25
M 7 days
-2                                               M 6 days
M 5 days
m 4 days
m 3 days
1 o m 2 days
M1 day
0L
Rural    Urban     Rural     Urban
2008                2010
Source: World Bank staff estimates from ECVAM
-116-


Figure Al .17: Number of days (out of last 7 days) in which households
did not eat the entire day
9
8
M 7 days
6 days
5                                              m 5 days
(v
M 4 days
M 3 days
0  2 days
2                                               m I day
0  Rura;l    Urban      Rural     Urban
2008                 2010
Source: World Bank staff estimates from ECVAM
Figure Al.18: Food for work
Food for Work
0 2008
6
0 2010
'R
04
'2-
1
0
poorest      2nd        3rd        4th        richest
Consumption Quintile
Source: World Bank staff estimates from ECVAM
-117-


Figure Al.19: Cash for work
16
Cash for Work
14 -                                                       200
112~ -                                       2010
- 12
u 10 -
8 -
2
0
poorest       2nd         3rd         4th        richest
Consumption Quintile
Source: World Bank staff estimates from ECVAM
Figure Al.20: Cash Transfers
6
Cash Transfer                    M21
5
4
.~3
1
0 i
poorest      2nd         3rd         4th        richest
Consumption Quintile
Source: World Bank staff estimates from ECVAM
-118-


Figure Al .21: Zakat/Donations
25
Zakat /Donations               0 2008
20 -m 2010
5 --
0
poorest      2nd         3rd        4th       richest
Consumption Quintile
Figure Al.22: Zakat/Donations
25-
Zakat /Donations               0 2008
20 -                                                  02010
S15-
~10
5
0
poorest      2nd         3rd        4th       richest
Consumption Quintile
Source: World Bank staff estimates from ECVAM
-119-


Annex 2      Children's Opportunities in Niger
Table A2. 1: Correlates of School Enrollment
(Coefficients are reported as Marginal Probabilities)
(1) All            (2) Rural           (3) Urban
VARIABLES             Attend school       Attend school       Attend school
Girl                  -0.128***           -0.174***          -0.0457***
(0.0146)           (0.0206)            (0.0155)
Age7                  0.0102              -0.0315             0.0379**
(0.0224)           (0.0348)            (0.0180)
Age8                  0.108***            0.0817**            0.0913***
(0.0206)           (0.0352)            (0.0142)
Age9                  0.165***            0.185***            0.101***
(0.0189)           (0.0349)            (0.0134)
AgelO                 0.0983***           0.0653*             0.0868***
(0.0210)           (0.0355)            (0.0147)
Agell                 0.149***            0.152***           0.0995***
(0.0200)           (0.0367)            (0.0137)
Agel2                 0.0640***           0.0759**           0.0393**
(0.0228)           (0.0376)            (0.0187)
Handicap              -0.276***           -0.252***          -0.215**
(0.0609)           (0.0643)            (0.0981)
Female_hh             -0.0178             -0.0142            -0.00683
(0.0247)           (0.0420)            (0.0207)
Age_hh                -0.00119*           -0.00208**          0.000595
(0.000640)          (0.000890)         (0.000727)
Girl04                0.0116              0.0143             0.00830
(0.00713)          (0.00998)           (0.00791)
Boys04                0.00316             -0.000119           0.0117**
(0.00485)          (0.00667)           (0.00575)
Girls514              0.0150***           0.0211***          0.00966*
(0.00545)          (0.00793)           (0.00563)
Female1564            0.0198***           0.0235***          0.0125**
(0.00571)          (0.00842)           (0.00583)
Men1564                -0.00981*          -0.0300***          0.00323
(0.00532)          (0.00891)           (0.00476)
Female65              0.0209              0.0223             0.00716
(0.0194)           (0.0275)            (0.0209)
Male65                -0.00793            0.00256             -0.0336
(0.0231)           (0.0324)            (0.0257)
School_in_area        0.244***            0.293***           0.0843**
(0.0199)           (0.0226)            (0.0357)
School_withinlkm      0.192***            0.247***           0.0719***
(0.0168)           (0.0270)            (0.0237)
Q1                    -0.213***           -0.137***          -0.264***
(0.0229)           (0.0328)            (0.0327)
Q2                    -0.140***           -0.0958***         -0.128***
(0.0222)           (0.0324)            (0.0264)
-120-


(0.0219)           (0.0338)           (0.0217)
Rural                 -0.265***
(0.0147)
Agadez                -0.0429                               -0.0475*
(0.0398)                              (0.0277)
Diffa                 0.0130                                -0.00313
(0.0377)                              (0.0355)
Dosso                 -0.0933***         -0.140***          -0.0818**
(0.0342)           (0.0428)           (0.0383)
Maradi                -0.0958***         -0.163***          -0.0542**
(0.0340)           (0.0450)           (0.0265)
Tahoua                0.0245              -0.0409            0.0338
(0.0315)           (0.0469)           (0.0214)
Tillab6ri             -0.0627*           -0.135***           0.0149
(0.0342)           (0.0440)           (0.0273)
Zinder                -0.0695**          -0.0987**          -0.0682**
(0.0323)           (0.0434)           (0.0266)
February              0.0830***          0.0799*             0.0554**
(0.0284)           (0.0457)           (0.0250)
March                 0.0527              0.107**            0.0266
(0.0322)           (0.0488)           (0.0313)
April                 0.0700**            0.0789*            0.0373
(0.0292)           (0.0452)           (0.0286)
May                   -0.0441             -0.00575          -0.0675
(0.0339)           (0.0464)           (0.0429)
June                  0.0326              0.0405             0.00665
(0.0390)           (0.0626)           (0.0380)
July                  -0.0437            -0.0368            -0.0283
(0.0331)           (0.0461)           (0.0363)
August                0.00245             -0.00922           0.0245
(0.0325)           (0.0469)           (0.0303)
September             -0.0190            -0.0371             0.00395
(0.0359)           (0.0505)           (0.0359)
October               0.0195             0.0158              0.00541
(0.0310)           (0.0447)           (0.0336)
November              0.0388              0.0408             0.0363
(0.0322)           (0.0484)           (0.0297)
December              0.0662**           0.0904**            0.0259
(0.0290)           (0.0438)           (0.0306)
Observations          6,031               3,279              2,752
Note: Standard errors in parentheses;
*** p<0.01, ** p<0.05, * p<0.1
-171-


Table A2.2: Covariates of Child (less than 5 years): Height and Weight
(1)                 (2)
VARIABLES                             Ln (Height)         Ln (Weight)
In(age months)                        0.170***            0.321***
(0.00144)           (0.00246)
Girl                                  -0.00926***         -0.0527***
(0.00331)          (0.00566)
ln(pcexp)                             0.0134***           0.0310*
(0.00265)           (0.00453)
#Female Infants                       0.00368**           0.0112***
(0.00154)           (0.00264)
#Boys                                 0.00581***          0.00703***
(0.00151)          (0.00258)
#Female Adults                        -0.004l)8***        -0.00612***
(0.00126)           (0.00216)
#Male Adults                          0.000693            -0.000181
(0.00119)           (0.00204)
Female_headed_hh                      0.0142**            0.0170*
(0.00571)          (0.00977)
Agehh_head                            0.000269**          0.000839***
(0.000136)          (0.000233)
No Edu                                -0.00572            -0.0119**
(0.00350)           (0.00599)
Madrasa Edu                           0.000543            -0.00272
(0.00364)           (0.00623)
Agadez                                -0.0273***          -0.0790***
(0.00702)           (0.0120)
Diffa                                 -0.0147*            -0.0526***
(0.00790)           (0.0135)
Dosso                                 -0.00659            -0.0265**
(0.00615)           (0.0105)
Maradi                                -0.0212***          -0.0401***
(0.00607)           (0.0104)
Tahoua                                -0.0131**           -0.0246**
(0.00617)           (0.0106)
Tillab6ri                             -0.00407            -0.0195*
(0.00630)           (0.0108)
Zinder                                -0.0200***          -0.0693***
(0.00589)           (0.0101)
Rural                                 -0.0109***          -0.0327***
(0.00331)          (0.00566)
Distance transport                    0.000-1 10          0.000477**
(0.000134)          (0.000234)
Distance water                        -6.08e-05           -0.000203*
(6.75e-05)          (0.000116)
Distance market                       -6.60e-05           -0.000202
(0.000151)         (0.000261)
Constant                              3.729***            1.002***
(0.0344)            (0.0590)
Observations                          4,889               4,890
R-squared                             0.748               0.786
Note: Standard errors in parentheses; *** p<0.01, *p<0.05, * p<0.1
Note 2: Coefficients reported as marginal probabilities.
-122-


Table A2.3: Under 5 Child Mortality - Probit Regression
(1)
VARIABLES                       DiedU5
Girl                            -0.00529
(0.00830)
Asset Index                     -0.0249**
(0.00985)
Mother_noedu                    0.0246*
(0.0134)
Motherage_firstbirth            -0.00258*
(0.00134)
Female_headed_hh                0.00600
(0.0134)
Agadez                          -0.0782***
(0.0188)
Diffa                           -0.0615***
(0.0202)
Dosso                           0.0289
(0.0230)
Maradi                          0.0275
(0.0226)
Tahoua                          0.00602
(0.0219)
Tillab6ri                       -0.00499
(0.0219)
Zinder                          0.0645***
(0.0249)
Rural                           0.0322**
(0.0146)
Observations                    9,511
Note 1: Standard errors in parentheses; *** p<0.01, ** p<0.05, * p<O.1
Note 2: Coefficients are reported as Marginal Probability
-123-


Table A2.4: Human opportunity Index for Girls, Niger, 1998-2006
Girls
Coverage       D-Index          HOI        Significant
change in
1998   2006    1998   2006    1998   2006   HO? (95%)
Education
School Attendance (6 to 11 years)  18.1%   27.1%  39.5%   24.6%  10.9%   20.5%      Yes
School Attendance (12 to 15 years)  21.1%  27.6%  40.7%   22.0%  12.5%   21.5%      Yes
Begin primary on time               4.8%   9.6%   56.3%   31.3%   2.1%   6.6%       Yes
Finished primary on time            4.9%   5.3%   51.5%   47.3%   2.4%   2.8%       No
Health
Immunization against measles       35.0%   48.3%  19.5%   13.1%  28.2%   42.0%      Yes
No underweight (Weight-for-age)    55.5%   65.6%   5.9%   5.8%   52.2%   61.8%      Yes
Source: World Bank staff estimates from survey data
Table A2.5: Human opportunity Index for Girls, Niger, 1998-2006
Boys
Coverage       D-Index          HOI        Significant
change in
1998   2006    1998   2006    1998   2006   HO? (95%)
Education
School Attendance (6 to 11 years)  25.9%   38.8%  26.4%   16.0%  19.0%   32.6%      Yes
School Attendance (12 to 15 years)  28.9%  42.5%  26.1%   15.7%  21.3%   35.8%      Yes
Begin primary on time               5.9%   13.3%  41.1%   26.6%   3.5%   9.7%       Yes
Finished primary on time            7.3%   6.5%   34.6%   48.4%   4.8%   3.4%       No
Health
Immunization against measles       35.9%   47.1%  22.4%   13.4%  27.9%   40.8%      Yes
No underweight (Weight-for-age)    53.4%   58.8%   7.6%   4.4%   49.4%   56.2%      Yes
Source: World Bank staff estimates from survey data
-124-


Table A2.6: Decomposition of change in HOI for Girls, Niger, 1998-2006
Total Change in                  Decomposition
HOI
1998 - 2006                              Coverage
(Percentage Points)      Composition   Scale  Equalization
Education
School Attendance (6 to 11 years)     9.5                   -0.9       7.3        3.1
School Attendance (12 to 15 years)    9.0                   -0.5       4.7        4.9
Begin primary on time                 4.5                   -0.4       2.6        2.3
Finished primary on time              0.4                   -0.2       0.4        0.2
Health
Immunization against measles          13.7                  1.0        9.6        3.1
No underweight (Weight-for-age)       9.6                   0.8        8.8       -0.1
Source: World Bank staff estimates from survey data
Table A2.7: Decomposition of change in HOI for Boys, Niger, 1998-2006
Total Change in                  Decomposition
HOI
1998 - 2006                              Coverage
(Percentage Points)      Composition   Scale  Equalization
Education
School Attendance (6 to 11 years)     13.6                  -0.3      10.5        3.4
School Attendance (12 to 15 years)    14.5                  0.2        9.8        4.4
Begin primary on time                 6.2                   -0.1       4.4        2.0
Finished primary on time              -1.4                  0.1       -0.4       -1.1
Health
Immunization against measles          13.1                  0.2        8.5        4.3
No underweight (Weight-for-age)       6.6                   -0.1       5.2        1.5
Source: World Bank staff estimates from survey data
-125-


Table A2.8: The most important correlates of change in HOI, Niger, 1998-2006
Girls
Percentage explained by:
D-        Household    Wealt    Household    Child'
Variable                       Index       Compositio     h        head's       s     Locatio
n        Index  characteristic  gende    n
s          r
Education
School Attendance (6 to 11 years)  24.6%       3.3       39.6       22.5        -      34.6
School Attendance (12 to 15
years)                         22.0%           0.0       43.0       18.9        -      38.2
Begin primary on time          31.3%           8.1       38.6       20.4        -      33.0
Finished primary on time       47.3%           0.6       39.2       23.6        -      36.6
Health
Immunization against measles    12.8%          4.9       46.0       22.1        -      27.0
No underweight (Weight-for-age)  5.6%         17.4       32.6       18.1        -      32.0
Source: World Bank staff estimates from survey data
Table A2.9: The most important correlates of change in HOI for Boys, Niger, 1998-2006
Boys
Percentage explained by:
D-        Household    Wealt    Household    Child'
Variable                       Index       Compositio     h        head's       s     Locatio
n        Index  characteristic  gende    n
s          r
Education
School Attendance (6 to 11 years)  16.0%       2.5       42.7       25.0        -      29.9
School Attendance (12 to 15
years)                          15.7%          6.5       41.8       24.1        -      27.6
Begin primary on time          26.6%           8.1       42.3       19.6        -      30.0
Finished primary on time       48.4%           5.9       38.6       19.8        -      35.7
Health
Immunization against measles    14.1%          8.0       33.6       38.0        -      20.5
No underweight (Weight-for-age)  5.1%          9.6       25.7       43.5        -      21.2
-126-


Annex 3: Figure s and Tables - Agriculture, Income and Rural
Poverty
Figure A3.1: Millet yield by cropping systems and departments (kg/ha)
700
600
500 -P re stand                   Intercropping
400 -                                    I
300                  -  -   -
200
100
0                                    ..... . ...
Source: NGAC, Author's calculations
Figure A3.2: Sorghum yield by cropping systems and departments(kg/ha)
500
450
400  -MF               re stand    Intercr  pping
350 -   -            -
300 -   -          -    -                -
250  -   --             --
200 -                  -i-
150  -                   -               -            -       -
100  -                -  ----
50 -
-fd -11A                                Ilia
0-   0M          0E            -0U
E   ~0
0         0         M.
Source: NGAC, Author's calculations
-127-


Figure A3.3: Niébé yield by cropping systems and departments (kg/ha)
600
500
9 Pure stand    Intercropping
400
300                                                                    -
200                                                                    -
100                         -k-
E                                 -c  E    '  E  E
-0     0                                   E 8 E t  O ~ C  O C
CO 0     =     (9                                f
O C3 -0im                 m    w
b.O                      m>~
Source: NGAC, Author's calculations
Figure A3.4: Groundnuts yield by cropping systems and departments (kg/ha)
900
800                  * Pure stand  e Intercropping
700                                                     _
600 -
500 -                                                                  -
400 -
300  -
200                                        e                           -
100
0
01 0  0                                              vJc
Source: NGAC, Author's calculations
-128-


Table A3.1: Descriptive statistics
Sample    Mean    SD     Min    Max
6403.  6435.         80024.
Weighted21 crop production (kg)               13945     6      0       0.0   2
Land (ha)22                                   13945     3.9     17.3   0.0    290.5
Labor (person days)                           7837      14.9    17.9   0.0   213.0
Seed (kg)                                    7790       33.8    29.8   2.0   210.0
190.
Precipitation (mm)                            13945     445.5   96.6   3     822.6
Gender (1 if male, 2 if female)               7830                     1.0    2.0
Age                                           7760      42.0    13.0   15.0   75.0
Self-sufficiency (1 if sufficient, 2 if intermediate,
3 if poor)                                    7278                     1.0    3.0
School attendance (1 if no formal school, 0
otherwise)                                    7278                     0.0    1.0
ICrop management (1 if didn't use, 0 otherwise)  7830                  1.0   2.0
Mechanization (1 if didn't use, 0 otherwise)  7830                     1.0    2.0
Number of farming parcels                     7278      1.7     0.7    1.0    3.0
Number of cattle                              7278      1.0     1.8    0.0    30.0
Number of sheep                               7278      1.3     2.3    0.0    20.0
Distance to the nearest food market (km)      5815      1.2     1.2    0.0   3.0
Source: NGAC (2008)
21Shares of land allocated to each crop are used as weights.
22 For estimation, whenever production was positive, missing land values were changed to average
cultivated land for that particular crop at zones level.
-129-


Annex 4: Stochastic Production Frontier Model
In this report we estimate the stochastic production frontier of the following form
qh = f (xh,fl)EChexp(vh),qh = f(xh,fl)Ehexp(vA),
(1)
where h = 1, ..., H, indexes farmers, qh qh is a (n x 1)(n x 1) vector of output for farmer h,
Xh Xh is a (1 x k)(1 x k) vector of associated inputs, ft is a (1 x k)(1 x k) vector of unknown
parameters to be estimated, and 88hrepresents farmer h's level of efficiency. In addition, the
farmer's production activity is subject to a stochastic shock vh-N (0, o1).v -~N (0, a,2).
In log form, equation (1) can be written as
lnqh = fto + Zt  flj lnxh j + InEh + vh lnqh = fto +Zt  flj lnxhj1 + InEh + Vh
(2)
Let uh = -lnEh,Uh = -lnEh, it follows that
Inqh = flo + Yt1 fnxhJ - Uh + vhtlnqh = flo +    i/ 1lnxhj - Uh + Vh
(3)
where Uh~N +(0, au), and A = oU/Uv -Uh-N+(0, au), and A = au/v,
Since variables influencing agricultural efficiency (Eh)(Eh) may also directly impact agricultural
production (qh),(qh), we adopt the approach proposed by Wang and Schmidt (2002), and Liu and
Myers (2009) where equation (3) is re-written as follows:
Inqh = flo + Zfj1lnxhj + Vh - Uh (z0,8), Uh(Zh,8)   0.            (4)
lnqh = fto +  1 =ljnxh + vh - Uh (zh, 8), uh (zh, 8)  0.            (4)
where ZhZh include farmers' education and gender, extension services, market access, and
rainfall. The frontier function and the inefficiency segment are then jointly estimated using a one-
step maximum likelihood estimation (MLE) procedure.
The marginal effect of ZhZh on production (qh)(qh) and efficiency (Uh)(Uh) is given by
0[E(qh|xh,zh)]/'zh1 = )[E(-uh|Xh,zh)]/8zf)Z.                        (5)
a[E(qh|xh,zh)]/8z)h = )[E(-uh1X|,Zh)1]/)Zh1.                         (5)
Equation (5) represents the semi-elasticity of output (efficiency) with respect to exogenous
factors, i.e., the percentage change in expected output (efficiency) when ZhlZhl increases by one
unit.
-130-


-131-