163
Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso A.C. van der Kraan Master Thesis January 2008

Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

  • Upload
    others

  • View
    2

  • Download
    0

Embed Size (px)

Citation preview

Page 1: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso

A.C. van der Kraan Master Thesis January 2008

Page 2: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in
Page 3: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso

Master thesis submitted in partial fulfilment

of the requirements of the degree of Master of Science

Faculty of Civil Engineering and Geosciences

Department of Water Resources Management

Delft University of Technology

in cooperation with

Small Reservoirs Project

Anneke van der Kraan

Final Report

January, 2008

Page 4: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in
Page 5: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

– Thinking differently about water is essential for achieving our triple

goal of ensuring food security, reducing poverty, and conserving ecosystems –

Water for food, water for life

Page 6: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in
Page 7: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

i

Preface

This is the final report for my master thesis project. The research is carried out at the section of

Integrated Water Resources Management of the Faculty of Civil Engineering and Geosciences of Delft

University of Technology (DUT). The project is conducted under supervision of Prof. Dr. Ir. Nick van de

Giesen.

The research is part of the Small Reservoirs Project (SRP)1, which is embedded in the CGIAR

Challenge Program Water for Food. The objectives of the SRP are (1) at basin/watershed level: to

promote and support the planning, development and management of small reservoir ensembles, (2) at

local/community level: to support use of small multi-purpose reservoirs that are properly located, well

designed, operated and maintained in sustainable fashion, and economically viable while assuring

they improve the livelihoods of the local residents (SRP, undated).

This research focuses on the second objective – more specific – the improvement of livelihoods due to

use of small multi-purpose reservoirs. The aim is to understand the interdependences between the

presence – or absence – of these reservoirs and the well-being of rural households living in their

vicinity.

I would like to thank my committee members for their advice and critics.

Prof. Dr. Ir. Nick van de Giesen – Chair of the department of Water Resources Management at

the faculty of Civil Engineering and Geosciences at Delft University of Technology, The

Netherlands.

Dr. Ir. Olivier Hoes – Lecturer at the department of Water Resources Management at the

faculty of Civil Engineering and Geosciences at Delft University of Technology, The

Netherlands and consultant for Neelen & Schuurmans Consultants in Utrecht, The

Netherlands.

1 The full title is: Planning and evaluating ensembles of small, multi-purpose reservoirs for the improvement of

smallholder livelihoods and food security: tools and procedures. The brief title is: Small Multi-purpose Reservoir

Ensemble Planning.

Page 8: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

ii

Dr. Eric Molin – Associate professor at the department of Transport Policy and Logistics

Organisation at the faculty of Technology, Policy and Management at Delft University of

Technology, The Netherlands.

Ir. IJsbrand de Jong – Senior water resources specialist at the department of Rural

Development Operations Eastern and Southern Africa at the World Bank in Washington DC,

United States of America.

Dr. Ir. Rhodante Ahlers – Senior lecturer in Water Management at the department of

Management and Institutions at the UNESCO-IHE Institute for water education in Delft, The

Netherlands.

Delft, January 20th, 2008

Page 9: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

iii

Summary

People living in Burkina Faso face highly variable rainfall, experience droughts and floods and

resulting have insecure livelihoods. Not only the year-to-year rainfall has a high variance, more

important is the variable distribution of rain over the growing season. To overcome dry periods, people

have built small reservoirs. According to several sources there are between 1400 and 2000 reservoirs

of all sizes in Burkina Faso. They are an important socio-economic infrastructure, as their use

supports many different purposes. This research focuses on the question how the use of small multi-

purpose reservoirs affects the well-being of poor rural livelihoods. The aim is to gain insight into the

relation between the presence – or absence – of small multi-purpose reservoirs and the state of well-

being of rural households living in the vicinity of those dams. So far, the extent and direction of the

relations between various dimensions of poverty and socio-economic values of small reservoirs are

based on assumptions. Knowledge of – positive and negative – impacts will be of significant value for

planning and management of these reservoirs. Therefore, the main research question is stated as:

“What is the relationship between the presence of small multi-purpose surface water reservoirs and

the state of well-being of rural households?”

First concern is identification of the socio-economic values of small reservoirs in rural areas in Burkina

Faso. Therefore, a literature review is conducted on the values of water ecosystems in general and

small surface water reservoirs in particular. Not so much the (monetary) economic value of water is

regarded, more the economic characteristics; water as a natural asset that is used by agriculture and

households, and so provides a means for human well-being. Identified socio-economic values – goods

and services – are water supply of domestic, agricultural and animal use, raw material, food and

nutrient supply, and other uses like recreation and education. Water from small reservoirs is used in

and around the house, e.g. for cleaning, bathing, washing, cooking. Generally, it is not a source for

drinking water; that is extracted from the groundwater; however, in areas where rainfall is very low

people may have no other choice than to use reservoir water. Additionally, general agriculture and

other agricultural purposes – such as fruit trees and vegetable gardens – are served. The diversion of

water to home gardens may contribute substantially to a varied diet or increase the household income.

Livestock may depend directly on water from small reservoirs, in addition to profiting from the higher

availability of fodder from crop stubble. Easier access to water can also contribute the development of

local economic activities, be it small scale and informal such as brick making, beer brewing, and mat

weaving.

Second concern is the need to define poverty as to be relevant to the aim and context of this research,

and to be compatible with data availability. Recurrently, a literature review is conducted on the various

Page 10: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

iv

concepts of poverty, and related dimensions. In order to define poverty the following steps are taken:

(1) Definition of the dimensions of poverty: related to the concepts of poverty adapted, a range of

dimensions is identified; (2) Selection of indicators of poverty: for every (sub)dimension one or more

indicators or proxies can be identified. These indicators are mainly determined by the data source: a

multi-topic household survey over 8,500 households covering the whole country of Burkina Faso. The

reported socio-economic values correspond with poverty dimensions that fall under the poverty lines

and basic needs concept, and therefore these are selected to represent poverty. As income, health

and nutrition being the main dimensions that explain poverty, their sub-dimensions are supporting

them. These sub-dimensions are measures of access (proximity + availability) to resources and

services, measures on health, nutrition and income levels and expenditures on resources and assets.

Consequently, the definition of poverty is formulated as: “Poverty is the lack of sufficient access to

financial and material assets, and public and natural resources, as to ensure being nutritioned and

healthy.”

Based on the qualitative analysis a (theoretical) cause-effect diagram showing the relations within and

between ‘storage’ and ‘poverty’ is drawn up. Together these relations form the conceptual model,

which is verified and quantified in the second part of the research. A valid cause-effect diagram is

essential for the outcomes of this research, as it functions as the conceptual basis for testing. Applied

methods for quantitative analysis are bivariate correlation analysis and multiple regression analysis,

both performed by using SPSS software package. This statistical approach allows quantify the system

of relations step-by-step as to finally estimate series of multiple regression equations between and

within (sub)systems of dependence and independence relations. The value of this approach is its

ability to estimate the impact of interventions in parallel with the interaction between multiple factors

within the poverty reduction process.

The analysis reveals that the main (positive) direct effect of small reservoirs is income generation and

education. Mainly employment rates and education levels benefit from storage. This leads to the new

hypothesis that when small reservoirs are more proximate this leads to considerable time savings.

Access to small reservoirs does not seem to contribute directly to improved nutritional status. The

state of food security is mostly determined by purchase of nutritional products and sufficient access to

stocks, while autoconsumption (self-sufficiency) of nutritional products plays a minor role. In turn,

expenditures on food are mainly determined by income levels and market access. There is no

evidence found that small reservoirs have a positive direct impact on food supply from (irrigated)

agriculture, dairy farming or aquaculture – that are considered as the main socio-economic values of

small reservoirs.

The concern is that the presence of small reservoirs would cause higher prevalence of water-related

diseases; water resources development in general has often been blamed for negative impacts on

Page 11: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

v

human health such as increased spread of malaria and schistosomiasis. Fortunately, the analysis

shows that this concern is dispensable; there is no evidence found that reservoir density relates to

gauges of water-related diseases (fever and diarrhoea). However, neither better sanitation nor

availability of improved potable water sources contribute to the reduction of water-related diseases. In

fact, the prevalence of water-related diseases is not explained by the variables within the model. We

can conclude population density is an important motivation for improved accessibility of water storage

(small reservoirs), and other resources (food market, potable water) and services (schools, health

services) become better accessible. Moreover, access to (public) transport contributes significantly to

the accessibility of resources and services.

The applied statistical techniques puzzle out the existence and strength of relations, however, the

direction remains only given in by theory. In general, the strength of the relations and the explained

variance by the regression models is low to medium; hence, we should be careful drawing strong

conclusions. Relations are weak due to (1) measurement error and the use of proxies to represent

concepts, and (2) disaggregation of variables. First proposed solution is re-defining of the

questionnaire as to upgrade the data by re-formulating survey questionnaires as to obtain data of (at

least) interval level. Additionally, re-formulating should lead to more reliable answers (reduce

measurement error) and more accurate indicators (to omit imperfect representation of concepts).

Secondly, we propose geo-referencing of the survey to be able to estimate the real proximity of small

reservoirs.

Page 12: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in
Page 13: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

vii

Glossary

Access For this research access is defined in a fairly simple way by only including

proximity and availability in the definition.

Covariance The averaged sum of combined difference.

Direct use values Direct use values arise from direct interaction with water resources. They may

be consumptive, such as use of water for irrigation or the harvesting of fish, or

they may be non-consumptive such as recreational swimming, or the aesthetic

value of enjoying a view.

Dummy variable A dummy variable is a numerical variable used in regression analysis to

represent subgroups of the sample.

Excreta Faeces and urine.

Food security Food security exists when all people, at all times, have access to sufficient,

safe and nutritious food to meet their dietary needs and food preferences for

an active and healthy life. The causal factors of food insecurity are equally

physical, economic and socio-political.

Homoscedasticity Homoscedasticity is the statistical assumption that the variance of the

dependent variable is the same for all the data.

Human capital Human capital is the attributes of a person that are productive in some

economic context.

Indirect use values Indirect use values are associated with services provided by water resources

but that do not entail direct interaction. E.g. they are derived from flood

protection provided by wetlands or the removal of pollutants by aquifer

recharge.

Interval level The interval measurement level represents quantitative data with a constant

unit of measurement, that have an arbitrary zero point. Therefore, it is not

possible to state that any value on an interval scale is a multiplication of any

other value on the scale.

Linearity Linearity is the statistical assumption that the relationship between variables is

a straight line.

Livelihood Refers to the means of gaining a living, including livelihood capabilities,

tangible assets and intangible assets.

Malnutrition The condition caused by deficiencies or imbalances in energy, protein and/or

other nutrients. Signs include wasting (thinness), stunting (shortness), or

being underweight.

Measurement error The degree to which observed values are not representative for ‘true’ values.

Page 14: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

viii

Multi-collinearity Multicollinearity refers to a situation of collinearity of independent variables.

Collinearity of two variables means that strong correlation exists between

them, making it difficult or impossible to estimate their individual regression

coefficients reliably.

Non-use values Non-use values are derived from the knowledge that a resource is maintained.

By definition, they are not associated with use of the resource or tangible

benefits that can be derived from it.

Nominal level The nominal measurement level is considered the lowest. It assigns numerical

values as labels to identify categorical data.

Normality Refers to the statistical assumption that at all variables and all combinations of

the variables are normally distributed (bell-shaped curve).

Option values Option value is the satisfaction that an individual derives from the ensuring

that a resource is available for the future given that the future availability of the

resource is uncertain. It can be regarded as insurance for possible future

demand for the resource.

Ordinal level In the ordinal measurement level categorical data are ordered or ranked in

relation to the amount of the attribute possessed. However, the scale is really

non-quantitative, because it indicates only relative positions in an ordered

series.

Outlier Observations with a unique combination of characteristics identifiable as

distinctly different from the other observations.

Poverty For this research poverty is formulated as follows: poverty is the lack of

sufficient access to financial and material assets, and public and natural

resources, as to ensure being nutritioned and health.

Poverty line The threshold below which a given household or individual will be classified as

poor.

Poverty mapping Refers to the use of maps in policy making and targeting assistance,

particularly in the areas of food security and environmental management.

Poverty maps are spatial representations of poverty assessments.

Quasi-option values Quasi-option value is derived from the potential benefits of waiting for

improved information prior to giving up the option to preserve a resource for

the future. This is based on a desire to take advantage of the prospect of

improved information in the future and act on subsequent revision of

preferences.

Ratio level The ratio measurement level represents the highest form of measurement

precision because they possess the advantages of all lower scales plus an

absolute zero point.

Page 15: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

ix

Small reservoirs In this research small reservoirs are defined as surface water reservoirs with a

surface until 100 hectare.

Stunted Low height for age (shortness).

Underweight Low weight for age.

Wasted Low weight for height (thinness).

Well-being Generally, well-being is the experience of good quality of life. Within this

research well-being is used as substitute (inversely proportional) for poverty.

Page 16: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in
Page 17: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

xi

Table of Contents

Preface ....................................................................................................................................... i Summary ...................................................................................................................................iii Glossary ...................................................................................................................................vii Table of Contents....................................................................................................................... xi Figures and Tables ...................................................................................................................xiv

1. Introduction ........................................................................................................................ 3 1.1 Context of the research .............................................................................................................. 3 1.2 Problem definition....................................................................................................................... 4

2. Approach............................................................................................................................ 7 2.1 Theory on socio-economic values.............................................................................................. 7 2.1.1 Framework for valuation......................................................................................................... 8 2.2 Conceptualisation of poverty .................................................................................................... 10 2.2.1 Dimensions of poverty.......................................................................................................... 10 2.2.2 Indicators of poverty............................................................................................................. 12 2.3 Methods for quantification of linkages...................................................................................... 13 2.3.1 Cause-effect relations .......................................................................................................... 13 2.3.2 Selection of statistical techniques ........................................................................................ 14 2.3.3 Correlation analysis.............................................................................................................. 16 2.3.4 Regression analysis ............................................................................................................. 17 2.4 Outline of the report.................................................................................................................. 20

3. Socio-Economic Values of Small Reservoirs........................................................................ 23 3.1 Water as a natural asset .......................................................................................................... 23 3.2 Values and classifications found in literature ........................................................................... 23 3.3 Socio-economic values of small reservoirs .............................................................................. 25 3.4 Characteristics of small reservoirs ........................................................................................... 27 3.5 Indicators of storage................................................................................................................. 29

4. Definition of Poverty .......................................................................................................... 31 4.1 Poverty in literature................................................................................................................... 31 4.2 Classification of dimensions ..................................................................................................... 31 4.3 Relevant dimensions of poverty ............................................................................................... 33 4.4 Indicators of poverty ................................................................................................................. 33

Page 18: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

xii

5. Linking Storage to Poverty ................................................................................................. 39 5.1 Storage as explanatory factor for poverty ................................................................................ 39 5.2 Linking storage and poverty ..................................................................................................... 40 5.2.1 Links between storage and poverty ..................................................................................... 41 5.2.2 External explanatory factors................................................................................................. 44

6. Verification of Relationships ............................................................................................... 49 6.1 Model specification................................................................................................................... 49 6.2 Bivariate correlation analysis.................................................................................................... 50 6.2.1 Direct links between storage and poverty ............................................................................ 50 6.2.2 Indirect links (and interdependencies) between storage and poverty ................................. 53 6.2.3 Links with external factors.................................................................................................... 58 6.3 Discussion ................................................................................................................................ 63

7. Quantification of Relationships ........................................................................................... 65 7.1 Model specification................................................................................................................... 65 7.2 Multiple regression analysis ..................................................................................................... 66 7.2.1 External factors .................................................................................................................... 66 7.2.2 Income.................................................................................................................................. 67 7.2.3 Nutrition................................................................................................................................ 72 7.2.4 Health ................................................................................................................................... 77 7.3 Discussion ................................................................................................................................ 83

8. Interpretation .................................................................................................................... 85 8.1 Interpretation of the country-scale analysis.............................................................................. 85 8.2 Urban versus rural environment ............................................................................................... 88

9. Reflection ......................................................................................................................... 91 9.1 Technical validation.................................................................................................................. 91 9.2 Evaluation of the scope ............................................................................................................ 92

10. Conclusions and Recommendations ................................................................................... 95 10.1 Conclusions.......................................................................................................................... 95 10.2 Recommendations for future work ....................................................................................... 98

Bibliography ........................................................................................................................... 101

Page 19: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

xiii

Appendix A. Values of Water............................................................................................... 109 Appendix B. Selection of Indicators...................................................................................... 113 Appendix C. Data Screening ............................................................................................... 121 Appendix D. Explorative Correlation Analysis........................................................................ 133 Appendix E. Explorative Regression Analysis ....................................................................... 135 Appendix F. Validation Tables............................................................................................. 137 Appendix G. Interpretation Tables ........................................................................................ 139

Page 20: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

xiv

Figures and Tables

Figures

Figure 1-1: Location of Burkina Faso ...................................................................................................... 3 Figure 2-1: Framework for valuation ....................................................................................................... 9 Figure 3-2: Reservoir density distribution.............................................................................................. 29 Figure 4-1: Classification of dimensions of poverty............................................................................... 32 Figure 5-2: Conceptual model – the cause-effect diagram ................................................................... 47 Figure 6-1: Correlations between storage and income ......................................................................... 50 Figure 6-2: Correlations between storage and nutrition ........................................................................ 51 Figure 6-3: Correlations between storage and health ........................................................................... 52 Figure 6-4: Main correlations within the poverty-storage system.......................................................... 62 Figure C-1: Boxplot for indicators of income ....................................................................................... 122 Figure C-2: Boxplot for indicators of expenditures on health services................................................ 123 Figure C-3: Scatterplot ‘household size’ * ‘total value of expenditures on health services’ ................ 124 Figure C-4 left: Scatterplot ‘household size’ * ‘value of expenditures on nutritional products’............ 125 Figure C-4 right: Scatterplot ‘household size’ * ‘value of autoconsumption of nutritional products’.... 125 Figure C-5: Scatterplot ‘household size’ * ‘total income’ ..................................................................... 125 Figure C-6: Scatterplot ‘length of the child’ * ‘weight of the child’ ....................................................... 126 Figure C-7: Scatterplot ‘reservoir density’ * ‘population density’ ......................................................... 126

Tables

Table 3-1: Values of small multi-purpose reservoirs ............................................................................. 26 Table 4-2: Selected indicators for income............................................................................................. 34 Table 4-3: Selected indicators for nutrition............................................................................................ 35 Table 4-4: Selected indicators for health............................................................................................... 37 Table 5-1: Storage and poverty links..................................................................................................... 40 Table 6-1: Correlations for interdependencies: income ........................................................................ 54 Table 6-2: Correlations for interdependencies: nutrition ....................................................................... 55 Table 6-3: Correlations for interdependencies: health .......................................................................... 57 Table 6-4: Correlations external factors: external factors ..................................................................... 58 Table 6-5: Correlations external factors: income .................................................................................. 59 Table 6-6: Correlations external factors: nutrition ................................................................................. 60 Table 6-7: Correlations external factors: health .................................................................................... 61 Table 7-1: Regression model for reservoir density ............................................................................... 67

Page 21: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

xv

Table 7-2: Regression model for population density............................................................................. 67 Table 7-3: Regression model for total revenue (direct model) .............................................................. 68 Table 7-4: Regression model for total revenue (complete model) ........................................................ 68 Table 7-5: Regression model for income from (owned) resources ....................................................... 69 Table 7-6: Regression model for income from employment ................................................................. 70 Table 7-7: Regression model for income from entrepreneurship.......................................................... 70 Table 7-8: Regression model for level of education.............................................................................. 71 Table 7-9: Regression model for proximity of primary school ............................................................... 71 Table 7-10: Regression model for proximity of secondary school ........................................................ 71 Table 7-11: Regression model for occurrence of food insecurity (direct model)................................... 72 Table 7-12: Regression model for occurrence of food insecurity (complete model)............................. 72 Table 7-13: Regression model for value of expenditures on nutritional products ................................. 73 Table 7-14: Regression model for value of autoconsumption of nutritional products ........................... 73 Table 7-15: Regression model for access to stocks (of cereals) until next harvest .............................. 74 Table 7-16: Regression model for proximity of food market ................................................................. 74 Table 7-17: Regression model for livestock holding ............................................................................. 75 Table 7-18: Regression model for landholding...................................................................................... 75 Table 7-19: Regression model for prevalence of malnutrition (direct model) ....................................... 76 Table 7-20: Regression model for prevalence of malnutrition (complete model).................................. 76 Table 7-21: Regression model for recent prevalence of disease (complete model)............................. 77 Table 7-22: Regression model for chronic prevalence of disability....................................................... 78 Table 7-23: Regression model for recent prevalence of fever .............................................................. 79 Table 7-24: Regression model for recent prevalence of diarrhoea....................................................... 79 Table 7-25: Regression model for value of expenditures on health services ....................................... 80 Table 7-26: Regression model for access to (improved) toilets ............................................................ 81 Table 7-27: Regression model for access to (improved) garbage disposal.......................................... 81 Table 7-28: Regression model for proximity of potable water source................................................... 81 Table 7-29: Regression model for availability of (improved) potable water source .............................. 82 Table 7-30: Regression model for proximity of health services ............................................................ 82

Page 22: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in
Page 23: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in
Page 24: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in
Page 25: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

3

1. Introduction

Although poverty is often considered a matter of institutions, governance and infrastructure, water

resources play a vital role in economic growth, human health and reduction of poverty in the semi-arid

and sub-humid savannah areas of West Africa. Therefore, throughout Burkina Faso – and other parts

of West Africa – many (small) dams and reservoirs have been built. They are catalyst for change,

initiate income generating activities, allow people to cope better with hungry periods of the year and to

diversify their diets.

1.1 Context of the research

Most of Burkina Faso belongs to the Sahel zone, the agricultural region between the Sahara and the

coastal rain forests. The land is green in the south, with forests and fruit trees, and dessert in the

north. Large parts of central Burkina Faso are located on a savannah plateau, with fields, brush, and

scattered trees. The mean annual rainfall varies from 1200 mm in the south-western part of the

country to less than 600 mm in the north (Coche, 1998). The climate is tropical with hot, wet summers

and warm, dry winters during which the hot, dry and dusty Harmattan wind blows. The rainy season

lasts approximately four months – May/June to September – and is shorter in the north of the country.

Droughts are often a chronic problem, especially in northern regions.

Like in most developing countries, a large part of the

population depends on agriculture for their income.

This is reflected in the fact that the agricultural sector

is the most important in Burkina Faso, followed by live-

stock husbandry. Over 80% of the labour force is

working in agriculture, only a small fraction is involved

in industry and services. Irrigation development is

currently growing exponentially, especially small-scale

irrigation at community level. This growth is not only

reflected in an increasing total number of dams, but

also in the conversion of drinking and cattle water

storage reservoirs into irrigation supply basins (Van de

Giesen et al., 2000). Figure 1-1: Location of Burkina Faso

Page 26: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

4

People living in Burkina Faso face highly variable rainfall, experience droughts and floods and

resulting have insecure livelihoods. Not only the year-to-year rainfall has a high variance, more

important is the distribution of rain over the growing season (Van de Giesen et al., 2000). To

overcome dry periods, people have built small reservoirs. These reservoirs are an important socio-

economic asset to the poor rural communities. According to several sources there are between 1400

and 2000 reservoirs of all sizes in Burkina Faso (SRP, undated; Coche, 1998). They typically provide

water for irrigation, livestock watering, fishing and domestic use. But they also serve a purpose for

wildlife watering and recreational purposes. Even if the water levels in the reservoirs run low, the

reservoir’s bottom can be exploited in order to make bricks out of the clay-rich soil (Poolman, 2005).

Most reservoirs have a rather shallow depth and consequently a small storage capacity. For this

reason many of them are seasonal; they store water during the wet season, as to be used during dry

periods and (part of) the dry season. Most permanent reservoirs are concentrated in central Burkina

Faso (Coche, 1998). Compared to large reservoirs (> 100 ha), small surface water reservoirs are less

reliable and effective for water conservation. Their small volume does not allow for seasonal or annual

carry-over, and the high surface area to volume ratio leads to high evaporation losses (Keller et al.,

2000). Small reservoirs have the advantage to large reservoirs that they are – in many cases – a less

strong threat to society and environment. Additionally, they are operationally efficient; they response

rapidly to precipitation runoff, are flexible, close to the point of use, and require relatively few parties

for management.

1.2 Problem definition

In Burkina Faso, where rainfall is highly variable and droughts are frequent, storage of freshwater is

essential to secure livelihoods during the dry season. Therefore, many (small) dams and reservoirs

have been built. Unfortunately, the qualitative and quantitative impacts of (clusters of) small reservoirs

on these livelihoods are quiet unknown. This research focuses on the question how the use of small

multi-purpose reservoirs affects the well-being of poor rural livelihoods. The aim is to gain insight into

the relation between the presence – or absence – of small multi-purpose reservoirs and the state of

well-being of rural households living in the vicinity of those dams. So far, the extent and direction of

the relations between various dimensions of poverty and socio-economic values of small reservoirs

are based on assumptions. Knowledge of – positive and negative – impacts will be of significant value

for planning and management of these reservoirs.

Page 27: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

5

Accordingly, the main research question is stated as:

What is the relationship between the presence of small multi-purpose surface water

reservoirs and the state of well-being of rural households?

Research questions and approach

In order to answer this question, three sub-questions are formulated:

1. What are the socio-economic values of small multi-purpose surface water reservoirs to poor

rural households in Burkina Faso?

2. How can we define ‘poverty’ within the scope of this research?

3. Which dimensions of poverty are (in)directly related to the presence (or absence) of small

multi-purpose reservoirs? What is the statistical strength of these relations? Which external

factors play a significant role?

In this research poverty and well-being are considered as (inversely proportional) substitutes.

Scope of the research

This research provides an impact analysis of the presence of small multi-purpose reservoirs on the

well-being of people living in Burkina Faso on macro scale. It focuses on the complex system of

factors that describe the origin of the link between storage and poverty. Principally, the scale of

aggregation for this research is – determined by the data source – the household level (or attributed to

the household level). Therefore, micro scale effects as inter-household poverty distribution is not

considered, but the analysis neither goes very broad; it does not comprehend effects of e.g.

globalization and aid. Later in the report poverty is to be defined to fit the purpose of this research.

As the aim is to regard this system in a generalized way, the influence of temporal and spatial

dynamics can not be included. Although it is recognized that poverty should be regarded as an

(infinite) process wherein time is an important parameter, available data resources are static (i.e.

household survey).

Page 28: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in
Page 29: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

7

2. Approach

In this chapter the methodological considerations are drawn up and theoretical frameworks are

shaped. Section 2.1, 2.2 and 2.3 each describe how to come to an answer of respectively research

question one, two and three. Finally, this chapter gives the outline of the report.

2.1 Theory on socio-economic values

As stated in the Dublin Principles by the International Conference on Water and the Environment

1992, water should be regarded as an economic good. Later, at the 2nd World Water Forum 2000,

agreement was reached that the full resource value – economic, social, cultural and environmental –

should be recognized in water management decisions (Agudelo, 2001).

The most common reason for undertaking a valuation of ecosystems is to assess the contribution of

ecosystems to social and economic well-being. But purely economic valuation of water often overlooks

two important dimensions: (1) Environmental values, such as the role of water flows in maintaining

biodiversity and ecosystem integrity. (2) Social values, which – at its most basic – can mean simply

using water to grow food to eat (FAO, 2006). The environmental dimension of water is essential to

sustain the basis for economic development, growth and poverty reduction (Kemper et al., undated).

The aim of this research is to identify the socio-economic values of small multi-purpose reservoirs in

rural areas in Burkina Faso. Therefore, a literature review is conducted on the values of water

ecosystems in general, and small surface water reservoirs in particular. Hereby, the term ‘value’ is

used to describe the importance placed on these ecosystems by individuals, which includes not only

income generation due to the use of its goods and services, but also other benefits it provides for

human welfare. Not so much the (monetary) economic value of water is regarded, more the economic

characteristics; water as a natural asset that is used by agriculture and households, and so provides a

means for human well-being.

Parallel with wetlands

Since extensive research is being done on the values of wetlands, an argument to draw a parallel with

wetlands is noted;

Page 30: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

8

During the Convention of Wetlands, signed in Ramsar (Iran, 1971) wetlands are defined as: "Wetlands are areas of marsh, fen, peatland or water, whether natural or artificial, permanent or temporary, with water that is static or flowing, fresh, brackish or salt, including areas of marine water the depth of which at low tide does not exceed six meters" (Article 1.1). In addition Ramsar sites "may incorporate riparian and coastal zones adjacent to the wetlands, and islands or bodies of marine water deeper than six meters at low tide lying within the wetlands" (Article 2.1). Furthermore, there are human-made wetlands such as fish and shrimp ponds, farm ponds, irrigated agricultural land, salt pans, reservoirs, gravel pits, sewage farms, and canals (Barbier et al., 1997).

As the convention adopts an extremely broad approach in determining wetlands, and considering the

similarities on valuation concepts for wetlands and general ecosystems found in literature, the parallel

is considered admissible.

2.1.1 Framework for valuation

Foregoing research done on the valuation on ecosystems provides a wide range of approaches,

conceptual frameworks and terms. Terms as resources, attributes, functions and services are often

used in literature; however, authors give different definitions or use them in a different context. They

are useful in classifying the socio-economic and ecological values of water, if and only if, the

conceptual framework in which they are applied is clarified.

The concept of Total Economic Value

The concept of Total Economic Value (TEV) is a widely used framework for analysis of the utilitarian

value of ecosystems (Alcamo et al., 2003). This concept divides the values derived from ecosystems

into two main categories: use values and non-use values. Typically, use values involve some human

interaction with the resource whereas non-use values do not (Barbier et al., 1997). They can be

divided into direct use values – which arise from direct interaction with water resources – and indirect

use values that are linked with services provided by water resources, e.g. flood protection. There is

another form of indirect use values called (quasi)option value2. It deals with the current satisfaction

and future improved information. Non-use values are usually associated with existence values.

Turner et al. (2000) proposes a framework based on Total Economic Value that describes the

interface between ecology and economy, and integrates the ecological interdependencies. It shows

2 See Glossary

Page 31: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

9

that in order to value water resources one has to establish their functions – the link between the

ecosystem characteristics, structures and processes – and the goods and services they provide and

that are valued by society. It gives a summary of the complex relationship between economics and

ecology.

Conceptual framework developed for this research

Since none of the frameworks provided in literature fits the scope of this research – more specific the

purpose of this first question – we have combined the framework of Total Economic Value (derived

from Barbier et al., 1997) with the framework proposed by Turner et al. (2000), and designed a

valuation framework for the specific scope of this research: (non-monetary) valuation of small multi-

purpose surface water reservoirs.

The socio-economic value of small surface water reservoirs is divided into two categories: goods and

services. Goods refer to the natural products harvested or used by communities such as water supply

for domestic, agricultural and livestock purposes, and natural products as wildlife and fish. Services

support life by indirect use or existence. Which and how many goods and services a reservoir can

provide depend on its characteristics: the physical features, natural environment and internal

processes. It should be noted that the development and use of small surface water reservoirs may

change the natural environment, and thus the reservoir characteristics.

PhysicalFeatures

Reservoir character istics

ExternalEnvironment

InternalProcesses

Goods

Values of small reservoirs

Services

Figure 2-1: Framework for valuation

Page 32: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

10

2.2 Conceptualisation of poverty

What is poverty? Poverty is hunger. Poverty is lack of shelter. Poverty is being sick and not being able to see a doctor. Poverty is not having access to school and not knowing how to read. Poverty is not having a job, is fear for the future, living one day at a time. Poverty is losing a child to illness brought about by unclean water. Poverty is powerlessness, lack of representation and freedom (Website PovertyNet).

As a multi-dimensional phenomenon, poverty is defined and measured in various ways. The

formulation of the definition determines how we analyze poverty and understand its dimensions. While

the main understandings of the term include material and economic needs, increasingly, the notion of

what constitutes basic needs has expanded to encompass not only food, water, shelter and clothing,

but also access to other amenities such as education, credit, participation, security and dignity (Hulme

et al., 2001).

This second research question concerns the need to define poverty in a way to be relevant to the aim

and context of this research, and to be compatible with data availability. Therefore, a literature review

is conducted on the various definitions and concepts of poverty, and the related dimensions –

including the implications for the poverty indicators – in order to consider a broad range of dimensions

before formulating the definition of poverty. Further, it is recognized that poverty may be closely

related or correlated with inequity, vulnerability, underdevelopment and social exclusion. In order to

define poverty for this research the following steps are taken:

• Dimensions of poverty. Related to the concepts or approaches of poverty adapted, a range of

main- and sub-dimensions can be identified;

• Indicators of poverty. For every (sub)dimension one or more indicators or proxies can be

identified. These indicators are mainly determined by the data sources available.

2.2.1 Dimensions of poverty

The first step concerns the selection of relevant dimensions of well-being. Traditionally, poverty

assessment is focussed on monetary dimensions – income, expenditures and/or consumption – based

on the assumption that human well-being is determined by the material standard of living3. Nowadays,

3 The material standard of living includes the own production, since this is an important asset for most poor.

Page 33: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

11

it is increasingly recognised that poverty measures based on these money-metric dimensions reflect a

static concept, offering only a limited picture of household well-being (Falkingham et al., 2002). It

ignores non-monetary dimensions, like health, life expectancy, education, literacy and access to public

goods or common-property resources, as well as dimensions of human capabilities like deficient social

relations, insecurity and lack of empowerment. Although the introduction of non-income measures

complicates the analysis, it nevertheless provides a more complete assessment of poverty in its

different dimensions. Moreover, it permits deeper analysis of the causes of poverty (Lok-Dessallien,

undated b).

Three concepts of poverty: income (poverty lines), basic needs and human capabilities

Most human welfare and poverty dimensions can be grouped into three major concepts of poverty:

income/consumption lines, basic needs and human capabilities. As others, these concepts are derived

from perceived causes of poverty, divided into physiological and sociological deprivation.

Both the income (poverty lines) and the basic needs concept are based on physiological deprivations

(Lok-Dessallien, undated a). The basic needs approach considers dimensions of poverty that are most

related to physical survival. The main understandings of the term include material and economic

needs, however, in a broader interpretation it goes beyond food needs. The UNDP (1997) describes

poverty in the basic need perspective as: “Poverty is deprivation of material requirements for minimally

acceptable fulfilment of human needs, including food. This understanding of deprivation goes well

beyond the lack of private income; it includes the need for basic health and education and essential

community services. It also recognizes the need for employment and participation.”

While the basic needs approach is especially useful with respect to access to public non-marketable

goods and services, the income/consumption or poverty line approach has become the most used tool

to define poverty in terms of having resources to satisfy needs, normally placed in the sphere of

private consumption. It consists in attributing a monetary value to a set of basic goods and services,

and identifying as poor those whose income is lower than a defined minimum: the poverty line (Rocha,

1998). Note that using income as an indicator to measure the basic minimum requires a strong

assumption: different people have the equal needs and derive equal welfare from a given income.

Relative poverty lines go beyond basic necessities and concern also the distribution of assets, hence

inequity.

The human capability concept of poverty focuses on expanding people’s opportunities and spans both

physiological and sociological sides of deprivation (Lok-Dessallien, undated a). This concept of

poverty represents the absence of some basic capabilities to function. Concepts of poverty based on

sociological deprivations are based on underlying structural inequities. Meaning that access to

Page 34: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

12

external assets – such as credit, land, infrastructure and common property – and internal assets – as

health, nutrition and education – are not always equally distributed. Despite the clear distinction

between poverty and inequity, analysis of poverty often employs indicators of equity because of

inherent linkages between them (Lok-Dessallien, undated a).

2.2.2 Indicators of poverty

A considerable amount of literature exists on poverty indicators. By selecting indicator(s) for each

dimension of poverty one should take known characteristics of poverty in a given society and the

availability of data on the living conditions of the population into account (Rocha, 1998). Indicators

need to be direct, unambiguous and relevant. Attention needs to be paid to indicators that are

substitutes for each other, e.g. time and distance, expenditures and consumption.

Data gathering

The data available are in the form of a multi-topic household survey: the ‘Questionnaire des

Indicateurs de Base de Bien-être’ (QUIBB)4 performed between April and July 2003. This

questionnaire is written on behalf of the National Institute of Statistics and Demography (INSD) of

Burkina Faso, and is meant for gathering the data needed for the economic and social management of

the country. It is, in design, a known way for collecting information about households’ characteristics,

measures of access, usage and degree of satisfaction in matters of social service (INSD, 2003). The

questionnaire contains more than 54,000 individuals divided over 8,500 households. The complete

country of Burkina Faso is covered, divided into 425 zones. The main advantage of this type of survey

is that the variables can be correlated with reasonable accuracy, since the same sample is used for

different modules of the survey (Lok-Dessallien, undated c). Data on poverty are available on

household or individual scale.

4 Translated: questionnaire of indicators of basic well-being

Page 35: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

13

2.3 Methods for quantification of linkages

Basis for answering the third question is given by the qualitative analysis on dimensions of poverty

and socio-economic values of water5. All links between poverty and storage are to be visualized in a

cause-effect diagram, including their interdependencies and some external factors. The diagram

functions as the conceptual model that provides the basis for further research. Subsequently, the

hypothetical relations are tested by means of correlation and regression analysis. This section

describes the theory behind causality, the selection of the statistical techniques and the basic

assumptions they behold.

2.3.1 Cause-effect relations

In order to link indicators of storage and poverty a cause-effect diagram is drawn, which shows the

theoretical causal relations between and within both components. Causality always implies at least

some relationship of dependency between the cause and the effect, i.e. change in one variable is

assumed to result in change in another variable. A valid cause-effect diagram is essential for the

outcomes of this research, as it functions as the conceptual basis for testing. As Hair et al. (1998)

states: “The strength and conviction with which we can assume causation between two factors lies not

in the analytical methods chosen, but in the theoretical justification provided to support the analyses.”

Note that a causal diagram shows only the – positive or negative – type of direction of an impact; it

does not show the direction of the causality.

Dependence relationships are sometimes, but not always, hypothesized to be causal in nature. Causal

relationships are the strongest type of inference made in applying (multi)variate statistics. Therefore,

they can be supported only when the following conditions for causality exist (Hair et al., 1998):

• Covariance between cause and effect, as to indicate sufficient association between the two

variables;

• Temporal antecedence of the cause versus the effect, meaning the cause must occur before

the effect;

• Non-spurious association must exist between the cause and effect, hence lacking alternative

causal variables that explain away the relationship;

• Theoretical support must exist for the relationship between the cause and effect.

5 Hereinafter ‘storage’ refers to the socio-economic values of small reservoirs, and ‘poverty’ refers to the

(sub)dimensions of poverty.

Page 36: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

14

2.3.2 Selection of statistical techniques

To select the appropriate statistical techniques for testing we must return to the research question,

which requests the direction and extent of the relations. In addition, the quality of the available data

and the assumptions underlying the statistical techniques must be taken into account.

Data quality assessment

In general, as to select the appropriate statistical technique(s), the following pre-tests should be

performed: (1) outlier detection, and (2) missing data analysis.

Outliers are observations with a unique combination of characteristics identifiable as distinctly

different from the other observations. Typically, it is judged to be an unusually high or low value on a

variable, or a unique combination of values across several variables that make the observation stand

out from the others (Hair et al., 1998). Outlier detection deals with identification and, possibly, deletion

of these extreme values. Note that outliers should be considered within the context of the research;

hence, they can be problematic as well as beneficial. When beneficial, outliers may be indicative of

characteristics of the population. In contrast, problematic outliers are not representative for the

population and can seriously distort statistical tests.

The main concern of missing data analysis is to identify patterns and relationships underlying the

missing data, in order to maintain as close as possible the original distribution of values when any

remedy is applied. The extent to which missing data occurs is of second concern. Missing data are of

importance since they influence sample size and possibly bias results (when not missing at random).

Therefore, the first step in the analysis is to determine the type of the missing data, i.e. whether data

are (completely) missing at random or not missing at random. Dependent on the randomness a

remedy or imputation method can be chosen. Second step is to determine whether the extent of

missing data is low enough to not affect the results, even if it operates in a non-random matter (Hair et

al., 1998). Rule of thumb is that missing data under 10% for an individual case can generally be

ignored, except when the missing data occur in a specific non-random fashion. Additionally, the

number of cases with no missing data must be sufficient for the selected analysis technique if

replacement values will not be substituted (imputed) for the missing data (Hair et al., 1998).

Assessing underlying assumptions

The available data – more in particular selected indicators – should meet the requirements of the

selected statistical techniques. These are threefold:

Page 37: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

15

1. Selected data should satisfy the statistical assumptions underlying the (multi)variate

technique, e.g. normality, homoscedasticity and linearity6. Testing the data for compliance with

the underlying assumptions deals with the foundation upon which the technique provides

statistical inferences and results. Some techniques are less affected by violating certain

assumptions – which is termed robustness – but in all cases meeting some of the

assumptions will be critical to a successful analysis (Hair et al., 1998).

2. Selected data should satisfy the required level of measurement. A critical factor in selection

and application of statistical techniques is the measurement level of the dependent and

independent variables. Data can be classified into two categories – categorical or continuous

– based on their characteristics. The measurement level is critical in determining which

techniques are applicable, with considerations made for both independent and dependent

variables.

3. For each subject in the study there must be related pairs of scores, i.e. sets of measurements

are obtained on the same individuals or on pairs of individuals.

Selection

The hypothetical causal relations are tested by means of correlation and regression analysis. These

techniques are closely related, as correlation provides the basis for regression analysis. Two variables

are said to be correlated if changes in one variable are associated with changes in the other variable.

The concept of association, represented by the correlation coefficient is fundamental to regression

analysis by describing the relationship between two variables (Hair et al., 1998). Thus, the function of

determining correlations within this research is to pre-define interesting combinations of variables for

regression analysis. The higher the correlation between variables, the better the prediction by

regression will be.

A quick scan of the data tells us that the above mentioned requirements are not satisfied by nearly all

indicators of poverty, and therefore, it is concluded that – since the data do not meet the requirement

of normal distribution – so called non-parametric tests should be applied. Problem is that not for all

parametric tests a non-parametric alternative exists. Therefore, also parametric alternatives are

discussed in the sections here below.

6 See Glossary

Page 38: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

16

2.3.3 Correlation analysis

The first statistical test selected is (bivariate) correlation analysis. Correlation analysis is primarily

concerned with finding out whether or not a relationship exists. The correlation coefficient determines

the (quantitative) extent to which two variables are related, and whether there exists a positive or

negative relation – referred to as the type of the relationship. Note that correlation coefficients do not

indicate the direction of the causality.

Technically, the correlation coefficient represents the standardized measure of covariance (Field,

2005). Therefore, the value of the correlation coefficient varies between +1 and –1. Both of the

extremes represent perfect linear relationships between the variables, and zero represents the

absence of a linear relationship. The sign of the correlation coefficient indicates whether there exists a

positive or negative relation – referred to as the type of the relationship. This is not an indication of the

direction of the causality. For correlations involving dichotomous variables, the sign of the correlation

(positive or negative relation) depends entirely on the coding; hence this requires extra attention when

interpreting.

Correlation coefficients

There are alternative types of correlation coefficients. Relevant correlation coefficients are:

• Pearson product moment correlation coefficient;

• Spearman rank order correlation coefficient;

• Point-biserial correlation coefficient.

Pearson’s correlation (rp). This parametric statistic requires at least data of the interval level7 for it to

be an accurate measure of the linear relationship between two variables. However, in establishing

whether the correlation coefficient is significant, more assumptions are required: for the test statistic to

be valid data have to be normally distributed. In any case, if the data are non-normal or are not

measured at the interval level then non-parametric tests should be performed (Field, 2005).

Spearman’s correlation (rs). When data have been measured at only the ordinal8 level they are said

to be non-parametric and Pearson’s correlation is not appropriate. Therefore, in these cases

Spearman’s correlation coefficient is used. Spearman’s coefficient is a non-parametric statistic, and

so, can be used when the data violate parametric assumptions such as normal distribution. It is also

applied when data are classified to be of ordinal measurement level, i.e. the categories are ordered in

a meaningful way.

7 See Glossary

Page 39: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

17

Point-biserial correlation coefficient is used when one of the two variables is (discrete) dichotomous.

The point-biserial correlation is mathematically equivalent to the Pearson correlation, that is, cases

where we have one continuously measured variable and a dichotomous variable. Therefore, in these

cases it is common practice to apply the Pearson correlation.

2.3.4 Regression analysis

The objective of regression analysis is to predict a single dependent (criterion) variable form one or

more independent (predictor) variables. When the problem involves a single dependent variable, the

statistical technique is called simple regression. When the problem involves more dependent

variables, it is termed multiple regression (Hair et al., 1998). Multiple regression is more complicated

than simple regression, but the basic principle is the same; estimating the linear combination of

predictors that correlate maximally with the outcome variable. In itself, linear regression is a parametric

technique. Non-parametric alternatives are (multi-nominal) logistic regression or ordinal regression.

However, Hair et al. (1998) states that regression analysis has been shown to be quite robust even

when the normality assumption is violated.

Regression equation

In regression, each independent variable is weighted by the regression analysis procedure as to

ensure maximal prediction from the set of independent variables. In the case of multiple regression for

explanatory purposes – as is the scope of this research – all of the independent variables should be

on comparable scale, i.e. standardized. The standardized coefficients denote the relative contribution

of the independent variables to the overall prediction, although correlation among the independent

variables complicates the interactive process. The set of weighted independent variables form the

regression model: a linear combination of the independent variable that best predicts the dependent

variable (Hair et al., 1998). The model is fitted is linear, meaning it is based on a set of straight lines.

The mathematical technique used to establish the best fitting line is called method of least squares.

The general form of the standardized regression equation reads as follows:

β β β ε= + + + +1 1 2 2( ... )i n n iY X X X (eq. 2-1)

where Yi is the outcome, and Xi the predictor variable of the i-th case’s score. β1, …,βn are the

standardized regression coefficients that represent the number of standard deviations that the

outcome will change as a result of one standard deviation change in the predictor when all other

predictors are kept constant (Field, 2005). As they are directly comparable, they provide a better

insight into the relative importance of individual predictors. The intercept is omitted since the

Page 40: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

18

standardized regression model always goes through the origin (0;0), as the mean of all coefficients is

zero and their standard deviation one. There is a residual term εi, which represents the error of fit.

The line of best fit is found by ascertaining which line, of all possible, results in the least amount of

difference between observed data points and the estimated line (Field, 2005). We are interested the

residuals – that are the differences in the vertical – since we use the line to predict values of Y from

values of the X variable. As the sum of positive and negative residual values tend to cancel each other

out, the square of the differences is used, hence the sum of squares represents the accuracy of the

estimated line.

Assumptions and requirements

The basic assumptions of regression are equal to those of correlation analysis: linearity, normality and

homoscedasticity. As regression analysis is based on the concept of correlation, the linearity of the

relationship between dependent and independent variables is crucial. Regression analysis also poses

requirements upon the level of measurement and sample size. In case the dependent variable is

categorical, (multi-nominal) logistic regression8 or ordinal regression is appropriate. When the

independent variables are categorical, with more than two categories, they must be converted into a

set of dichotomous variables by dummy variable coding.

Additional assumptions underlying multiple regression are (Ho, 2006):

• Independence of error terms. In regression, it is assumed that the predicted value is not

related to any other prediction; hence each predicted value is independent. The Durbin-

Watson statistic informs about whether the assumption of independent errors is tenable;

• Normality of the error distribution. It is assumed that errors of prediction – differences between

the obtained and predicted dependent variable scores – are normally distributed. Violation of

this assumption can be detected by a visual check of the frequency distribution of residuals;

• No (almost) perfect multicollinearity. For unbiased multiple regression analysis there should be

no perfect linear relationship between predictors – i.e. no high correlations. Multicollinearity

can be diagnosed by assessing the correlation matrix and the collinearity diagnostics. The

latter provide some measures of whether there is collinearity in the data. Specifically, it

provides the variance inflation factor (VIF) and tolerance statistics.

8 In case of multiple categories of the dependent variable, multi-nominal logistic regression is applied.

Page 41: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

19

Overall model fit

Measures of the overall model fit are given by R and R-square. In case of several predictors the

correlation coefficient represents the correlation between the observed values of Y, and the values of

Y predicted by the multiple regression model. Consequently, the resulting R-square – also called

coefficient of determination – can be interpreted as follows: it is the amount of variation in the outcome

variable accounted for by the model.

However, the fact that the estimated model is significant does not imply the model to be a good

representation of reality. For this assumption, also the test statistic requires a significant confidence

level. The test statistic – that is again a measure of the variance, i.e. variance explained by the model

divided by the variance not explained by the model (Field, 2005) – is a measure of the goodness-of-fit

of the model. For regression analysis we review the t and F test statistics. Fischer (1991) described

this in his criterion that when this probability falls below .05, it gives sufficient confidence to assume

the value of test statistic is indicating that the model is generalizable; i.e. representing the population.

Model parameters

In order to assess the individual contribution of variables the model parameters – beta values – and

the significance of these values is assessed. Earlier is explained that the regression coefficients

β1,…,βn represent the change in the outcome resulting from a unit change in the predictor when all

other variables in the equation are kept constant. Also, if a predictor has a significant impact on the

prediction of the outcome then this coefficient (βi) should be different from zero. In multiple regression

the significance of the t-test indicates whether the regression coefficient is different from zero. As a

general rule, if this observed significance is less than 0.05, then the result reflects a genuine effect

(Field, 2005).

Logistic regression

In cases we aim to estimate regression relations where the outcome variable (dependent variable) is

dichotomous logistic regression is used. Logistic regression is multiple regression but with an outcome

variable that is a categorical dichotomy and predictor variables that are continuous or categorical

(Field, 2005). Logistic regression is limited, however, to prediction of only a two-group (binary)

dependent measure. Thus, in cases for which three or more groups form the dependent measure –

but in a non-ordered manner – multi-nominal logistic regression should be applied.

In logistic regression the overall model fit is assessed by the Hosmer & Lemeshow test. The chi-

square test statistic provides an indication of the goodness-of-fit of the model. It tests the hypothesis

that the observed data are significantly different from the predicted values from the model. So we want

Page 42: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

20

a non-significant value for this test – indicating that the model is a fair representation of the data. The

squared regression coefficient in linear regression is replaced by Nagelkerke R2. In terms of

interpretation this can be similarly interpreted to the R2 in linear regression; it provides a gauge of the

substantive significance of the model.

The assessment of the individual predictors is based on the Wald statistic. If the logistic coefficient B is

significantly different from zero then it can be assumed that the predictor is making a significant

contribution to the prediction of the outcome. Hence, the Wald statistic is basically identical to the t-

statistic in linear regression. The direction of the relation can be directly assessed from the sign of the

B-value, or indirectly from the exponentiated coefficients Exp(B); less than one are negative, greater

than one are positive. The magnitude is best assessed by the Exp(B), with the percentage change in

the dependent variable is shown by (Hair et al., 1998):

= −% ( ( ) 1) *100change Exp B (eq. 2-2)

For the aforementioned interpretation to be reliable the confidence interval of Exp(B) should not

include one. If the confidence interval ranges from less than one to more than one, then this would

limit the generalizability of the model parameter, i.e. the direction of this relationship may be unstable

in the population as a whole (Field, 2005).

2.4 Outline of the report

Chapter 1 gives the background of the research. It first illustrates the context of the research by

describing relevant features of livelihood in Burkina Faso and the role small reservoirs play in that

(Section 1.1). Secondly, it presents the problem definition, research questions and research

perspective (Section 1.2).

Chapter 2 describes the methodological considerations and theoretical frameworks applied in the

research. Section 2.1 describes the framework for identifying socio-economic values of small

reservoirs. Section 2.2 defines the context of poverty concepts – income (poverty lines), basic needs

and human capabilities – throughout which relevant dimensions and indicators are identified. Within

those concepts lies the definition of poverty for this research. Section 2.3 explores the theory behind

causality, as to enable design of a valid conceptual model for this research. This model functions as

the base for quantitative analysis. Sections 2.3.2 to 2.3.4 go deeper into the selection of appropriate

statistical techniques and the basic assumptions they behold.

Page 43: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

21

Chapter 3 elaborates on the framework for valuation given in Section 2.2. It gives a short overview of

the literature survey done on socio-economic values of water. Based on this, Section 3.3 provides a

description of goods and services provided by small reservoirs. However, which and how many are

provided depends on the reservoir characteristics, which are described in Section 3.4. Finally, Section

3.5 gives the indicators of storage used for this research.

Chapter 4 classifies the dimensions of poverty into the context of poverty concepts. Based on this it

gives the definition of poverty for this research. Subsequently, Section 4.4 gives proxies and indicators

for each sub-dimension of poverty.

Chapter 5 functions as the transition between the theoretical and statistical phase of this research. By

linking the socio-economic values to dimensions of poverty it provides the theory – visualized by

means of the cause-effect diagram – that will be the base for the application of statistical methods.

Moreover it describes detailed, in terms of (sub)dimensions and indicators, all relations that are

visualised in this conceptual model.

Chapter 6 deals with verification of proposed relations by means of bivariate correlation analysis. It

first gives the basic statistical knowledge needed prior to understand, implement and interpret the

analysis. The results of the correlation analysis are described and visualized in Section 6.2. The

chapter is closed with a short discussion (Section 6.3).

Chapter 7 analyzes the relative contribution of variables and their interaction effects by means of

multiple regression analysis. Again, the basic statistical knowledge needed prior to understand,

implement and interpret statistical test is discussed (Section 7.1). Section 7.2 describes the results of

the regression analysis split up into storage, income, nutrition and health. The chapter is closed with a

short discussion (Section 7.3).

Chapter 8 deals with the overall interpretation of the statistical tests. The aim is to get insight into the

complete system. Section 8.1 deals with the interpretation of the complete system on country scale.

And Section 8.2 additionally analyzes the system for separate environments: rural versus urban.

Chapter 9 provides a reflection on the performed research; it encompasses the validation of the

statistical tests by means of split-sample validation (Section 9.1) and looks deeper into the

authentication of the scope and assumptions of the research (Section 9.2).

Finally, Chapter 10 draws the overall conclusions, gives answers to the research questions and

provides recommendations for future work.

Page 44: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in
Page 45: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

23

3. Socio-Economic Values of Small Reservoirs

This chapter provides a short overview of the conducted literature review on the values of water and

discuss the different classifications used. Results of the complete literature review are found in

Appendix A. Next, values – the goods and services – are identified that are provided by small surface

water reservoirs in particular. Therefore, we use the framework for valuation of small reservoirs as

discussed in Chapter 2, Section 2.1.1. Finally, a list of selected values and their indicators is

presented.

3.1 Water as a natural asset

Surface water can be considered as a natural resource, the value of which resides in its ability to

create flows of goods and services over time. It provides a basis to many livelihood strategies, like

agriculture, fisheries and cattle farming, and thus has an essential socio-economic value. Not so much

the (monetary) economic value of water is regarded in this research, more the economic

characteristics; water as a natural asset that is used by poor rural households, and as such provides a

means for poverty reduction, social and economic development and food security.

3.2 Values and classifications found in literature

Since the early seventies, attempts have been made to provide a more systematic listing of the many

benefits of natural ecosystems to human society, and to design methods for assigning values to those

benefits (De Groot, 1992). The result is a large variety of frameworks and concepts for valuation.

Below, a brief summary of these frameworks and concepts is given. The complete overview of the

conducted literature review on the values of water can be found in Appendix A.

The classification of Barbier et al. (1997) is based on the framework of Total Economic Value (TEV),

and applied by the Ramsar Convention on Wetlands. It distinguishes use values from non-use values

based on the criteria of human interaction with an environmental resource. Non-use values refer to

current or future values associated with a resource which rely merely on its continued existence and

are unrelated to use. Use values are grouped according to whether they are direct or indirect. Direct

use values involve both commercial and non-commercial activities, with some of the activities often

being important for the subsistence needs of local populations in developing countries. The indirect

use values emanate from supporting or protecting activities. A special category of use values is option

Page 46: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

24

value, which arises because an individual may be uncertain about his/her future demand for a

resource and/or its availability in the future. Quasi-option value is simply the expected value of the

information derived from delaying exploitation and conversion of the resource today.

The classification of Roggeri (1995) is designed for valuation of tropical freshwater wetlands. It defines

functions, resources and attributes. Functions of wetlands are due to their role in many natural

phenomena and processes. Resources can be used in order to obtain products or services. Finally,

they provide attributes – or qualities – such as biological diversity. All functions, attributes and

resources are goods and services which have a value for human beings. Note that all values are

closely linked to the resource biological, chemical and physical characteristics, and to the interaction

between these. Therefore, not all possible goods and services are automatically provided by the

resource. Furthermore, the role a resource plays in a given process may vary considerably, both in

significance and quality.

The classification of De Groot (1992) is based on environmental function evaluation, thus includes not

only the harvestable goods (nature in the narrow sense) and land-use values, but also refers to other

benefits of the natural environment which are less tangible. Environmental functions are defined as the

capacity of natural processes and components to provide goods and services that – directly of

indirectly – satisfy physiological and psychological human needs. De Groot divides the environmental

functions into four classes:

• Regulation function. The capacity of natural and semi-natural ecosystems to regulate essential

ecological processes and life support systems, which contribute to the maintenance of a

healthy environment by providing clean air, water and soil;

• Information function. Natural ecosystems contribute to the maintenance of mental health by

providing opportunities for reflection, spiritual enrichment, cognitive development and

aesthetical experience;

• Carrier function. Implies that natural and semi-natural ecosystems provide space and a

suitable substrate or medium for many human activities such as habitation, cultivation and

recreation;

• Production function. Nature provides many resources, ranging from food and raw materials to

energy resources and genetic material.

None of the above classifications perfectly suits the purpose of this research, since the focus of this

research lies on small multi-purpose reservoirs in (semi)arid areas. Therefore, a new framework for

non-monetary valuation of small reservoirs in arid and semi-arid areas is designed, the content of

which is described in the next sections. The proposed framework combines the principles of Total

Economic Value (derived from Barbier et al., 1997) with the framework proposed by Turner et al.

(2000).

Page 47: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

25

3.3 Socio-economic values of small reservoirs

Many different classifications of ‘small reservoirs’ can be found in literature. According to Van de

Giesen et al. (2000) small reservoirs are defined as to have a surface smaller than 5 hectares,

reservoirs with a surface between 5 and 100 hectares are classified as medium and with surfaces

exceeding 100 hectares as large. For this research, reservoirs with a surface area till 100 hectares are

defined as ‘small’.

Small reservoirs are multi-purpose structures, whose uses include irrigation, livestock watering, brick

making, domestic use and recreation. Chapter 2, Section 2.1.1 presented the framework that is used

to identify the values – goods and services – of small surface water reservoirs. The socio-economic

value is divided into two main categories: goods and services. Goods refer to the natural products

harvested or used by communities such as water supply for domestic, agricultural and livestock

purposes, materials as clay and reed, and natural products as wildlife and fish. Services support life by

indirect use or existence, i.e. by functioning. Numerous processes and activities would be

considerably altered or disappear if the reservoir no longer exists or not performs properly. For

example, reservoirs protect downstream agricultural land by mitigating floods and droughts. Existence

might also have an intrinsic value based on ethical or aesthetical criteria. Goods and services are not

necessarily of socio-economic value, such value derives from the demand from human society. The

extent and scale at which each particular value is provided depends on several factors which include

quantity and quality of the water, productivity, accessibility, availability of alternative sources of water,

and communities’ network (Dinar et al., 1995). These factors are classified in the framework as the

reservoir characteristics. The framework leaves out the spatial and temporal dynamics, hence the

option and quasi-option values proposed by Barbier et al. (1997).

Below, a general list of reservoir goods and services at the household level is presented. It provides a

convenient starting point for identifying which values a reservoir is likely to contain. Note that it is often

inappropriate to assess a reservoir for all the values in a standard mode, since not all reservoirs

perform all functions to the same degree or magnitude, if at all (Smith et al., 1995).

Page 48: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

26

Table 3-1: Values of small multi-purpose reservoirs

Goods provided by small reservoirs Services provided by small reservoirs

Water supply for domestic use

basic consumption needs

brewing traditional beer

Water supply for agricultural use

communal agriculture

fruit trees and vegetable gardens

Water supply for animal use

livestock grazing and watering

wildlife cropping and harvesting

Raw material supply

building material

energy material

Food and nutrient supply

fisheries and aquaculture

gathering of natural products and medicines

Other uses

recreation and tourism

education and cultural heritage

Flood and drought mitigation

Groundwater recharge

Water quality improvement

nutrient retention

sediment retention

salinity control

water treatment

Micro-climate stabilization

Biodiversity maintenance

role in the life cycle of some species

Small reservoirs can be considered a socio-economic infrastructure, as it is used for many different

purposes. The multiple use of reservoir water has different positive and negative effects on human

health and rural development; hence they affect rural livelihoods by contributing to health, reducing

costs of living and even by providing additional income. A great many of the poor in urban and rural

settings base their livelihoods on informal activities; e.g. small-scale cropping, livestock rearing, agro-

processing and other micro-enterprises. In many of these activities adequate water supply is a crucial

enabling resource: used in, or necessary for, the activity itself; freeing time (by reducing time spent

collecting water); or as a key element in improved health that in turn enables people to do work

(Moriarty et al., 2004).

Page 49: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

27

Small reservoirs typically support – besides general irrigation – other agricultural purposes, such as

fruit trees and vegetable gardens and livestock watering. Livestock may depend directly on water from

small reservoirs, in addition to profiting from the higher availability of fodder from crop stubble. The

diversion of water to home gardens may contribute substantially to a varied diet or increase the

household income. These gardens – often the responsibility of women – may have vegetables, herbs,

and trees bearing nutritious fruit. Small reservoirs can also be a direct source of protein and micro-

nutrients in the form of aquatic plants, fish, and other organisms. Additionally, domestic purposes,

such as laundry, bathing, washing household utensils, soaking grains, cooking, drinking, house

cleaning, sanitation benefit from the presence of small reservoirs. Note that a small reservoir is not a

source for drinking water; that is extracted from the groundwater. However, in areas where rainfall is

very low, people have no other choice than to use surface water (Boelee et al., 2000).

Easier access to water can also contribute the development of local economic activities, be it small

scale and informal such as brick making, butcher shops, washing vehicles, pottery, mat weaving are

served. These rural industries may provide employment and contribute to local income generation

(Boelee et al., 2000). Besides, small reservoirs may supply a range of products that can be used as

fertilizer, energy and construction material.

Water resources development in general has often been blamed for negative impacts on human

health such as increased spread of malaria and schistosomiasis, however, in many cases the multiple

use of surface water reveals important contributions to improved health and livelihoods. While the

consumption of untreated surface water holds certain risks to human health, the higher availability of

water through the presence of irrigation systems may actually improve health. Water-washed and

even water-borne diseases are reduced with increased use of water for bathing and for consumption

(Boelee et al., 2000; Lipton et al., 2003).

3.4 Characteristics of small reservoirs

Reservoir characteristics describe the features of the ecosystem that the reservoir is part of, and that

determine the extent and scale at which values are provided to human society. They are classified into

three (interrelated) categories: physical features, natural environment and internal structure. Reservoir

characteristics are in turn influenced by the values they provide; development and use of small surface

water reservoirs may change the reservoir characteristics. Turner et al. (2000) defines characteristics

as “those properties that describe a wetland area in the simplest and most objective possible terms.

They are a combination of generic and site-specific features”. The interactions among hydrology and

geomorphology, saturated soil and vegetation determine the general characteristics and the

significance of the processes that occur in any given ecosystem. These processes also enable the

Page 50: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

28

development and maintenance of the ecosystem structure which in turn is the key to the continuing

provision of goods and services.

Small reservoirs – physical features

There exists an enormous variety in the physical features of small reservoirs. Many are earthen dams

constructed across a river or stream bed, so as to obstruct the natural stream flow and contain this

water in type of retention basin (Balasz, 2006). Therefore, the shape – slope and size – and volume of

reservoirs depend on the features of the stream channel beds and bottom valleys they are constructed

in. Especially in semi-arid areas, the presence and capacity of most small dams is variable – both

within seasons and between seasons – due to rainfall variability, evaporation, seepage, siltation and

water withdrawals. Small reservoirs often have one or two outlet structures. Their dams are usually

made out of earth with a clay core, in cases protected by rocks to prevent dam crest erosion. Often,

tropical rains cause overflowing and thereby damage the dam wall. The presence, dimensions and

quality of a spillway are of great importance for the dam’s durability (Liebe, 2002); i.e. the sustainability

and integrity of the dam. However, for this research, we assume reservoirs function as designed.

Small reservoirs – external environment

The external environment that is most close to the reservoir is the natural environment. This compiles

all natural conditions that influence the reservoir performance. Think of weather and climate

conditions, hydrology and geology. Besides the natural environment the social environment, economic

environment, political environment, etc can be recognized such as land-use, demography and

institutional setting.

Small reservoirs – internal processes

The internal processes are closely related to the services which a reservoir provides. They are

categorized into hydrological processes, chemical processes and biological processes. Under

hydrological processes are meant the – long and short term – storage of (sub)surface water and the

interaction with groundwater. Biological and chemical processes are the cycling of nutrients, removal

and retention of elements, and the export of organic carbon.

Page 51: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

29

3.5 Indicators of storage

Supply of goods and services can be quantified by measures of proximity and availability. Another

important indicator is water quality, since it determines the variety of activities for which the water can

be used for. Unfortunately, no data are available on reservoir water quality, thus, it is assumed the

water is appropriate for all types of uses with the exception of the use as potable water source.

Proximity to the reservoir is measured through assessing the average distance to a reservoir from

any point within the province that the reservoir belongs to. The provincial scale is appropriate since no

useful (georeferenced) data is available on the location of the households (or zones) that are

interviewed in the QUIBB. Another useful measure of reservoir accessibility is reservoir density at

provincial scale. Both indicators introduce an error in the real proximity of reservoirs, since borders of

the provinces are no physical boundaries in reality.

0 – 0.10

0.10 – 0.50

0.50 – 1.00

1.00 – 1.50

1.50 – 2.00

2.00 – 3.00

3.00 – 4.00

Reservoir density [#/100 km2]

Figure 3-2: Reservoir density distribution

Page 52: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

30

Water availability is not easy to measure. The most appropriate indicator is the reservoir volume or

stored volume of water at the end of the rainy season. Another indicator would be the reservoir

surface area – obtained from (digitalized) satellite images – that is input to empirical formulae to

estimate volumes. In that case images towards the end of the rainy season are required, in order to be

as accurate as possible. So far these images are not available. In any case, indicators face the

problem of seasonality. We have learned that most reservoirs are seasonal, since they have a rather

shallow depth and consequently a small storage capacity (Coche, 1998). The hypothesis would now

be that the larger the reservoir, the more impact it would have on (the reduction of) poverty. Here an

error is introduced, since not all reservoirs are seasonal or can be considered equally influenced by

seasonality. To overcome this problem we assume that seasonality does not exist; hence, all

reservoirs are always filled.

Page 53: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

31

4. Definition of Poverty

This chapter gives the definition of poverty for this research, meaning we explicitly define the

dimensions of poverty considered and the indicators that measure them. Before listing the dimensions

of poverty relevant in this study we give a short overview of the dimensions of poverty – or well-being

– found in literature.

4.1 Poverty in literature

The theme of the website of the World Bank says: “Working for a world free of poverty.” With poverty

described as: “Living below a minimum level of income, such as a US$1 per day per person, often

defines poverty. However, poverty is also a lack of adequate food, shelter, health, education, and

influence over decisions that affect one’s life. Of the six billion people on our planet, three billion live in

developing countries in conditions that fit those definitions of poverty” (Website World Bank).

Also the Human Development Report 1997 (UNDP, 1997) defines poverty beyond income: “Human

poverty is more than income poverty; it is the denial of choices and opportunities for living a tolerable

life.”

The International Fund for Agricultural Development states that: “Poverty is not only a condition of low

income and lack of assets. It is a condition of vulnerability, exclusion and powerlessness. It is the

erosion of their capability to be free from fear and hunger and have their voices heard” (Website

IFAD).

These are some illustrative examples of definitions of poverty by some NGO’s. Since we want to think

in terms of dimensions of poverty, we find here income, food, shelter and education. Looking further,

also more abstract dimensions as opportunities, vulnerability and powerlessness are found. The

coming sections will elaborate on the many different dimensions of poverty and the indicators that

support them.

4.2 Classification of dimensions

When regarding dimensions of poverty from the perspective of the three concepts of poverty

addressed in Chapter 2, Section 2.2.1, we can give the following broad list of (complementary)

Page 54: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

32

dimensions. Note that the following scheme is not the only interpretation possible, since borders of

concepts are flexible and poorly defined.

Poverty Lines Basic Human Needs Human Capabilities

Income

Consumption

Expenditures

Nutrition

Health

Education

Sanitation

Energy

Shelter

Potable water

Resources and assets

Public services

Food security

Employment

Literacy

Malnutrition

Sickness

Disability

Powerlessness

Vulnerability

Participation

Inequity

Choices and opportunities

Underdevelopment

Security

Rights and dignity

Social exclusion

Fear for the future

Livelihood sustainability

Figure 4-1: Classification of dimensions of poverty

Page 55: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

33

4.3 Relevant dimensions of poverty

From this list we can exclude dimensions that tend to go into the direction of the human capability

concept, since these fall outside of the scope of this research; that should consider the dimensions of

poverty that are (almost) directly influenced by the socio-economic values of small reservoirs, and

indicators are best to be measured at household level. These dimensions are fairly un-directly related

to the presence – or absence – of small reservoirs, and it can be stated that they are more directly

influenced by many other factors. Moreover, these dimensions are not easy to operationalizable and

measurable, neither indicators are available.

The dimensions that fall under the poverty lines and basic human needs concept can be seen as the

most basic and more directly influenced by water access, and therefore, are selected to represent

poverty. As income, health and nutrition being the main dimensions that explain poverty, their sub-

dimensions are supporting them. These sub-dimensions are measures of access (proximity and

availability), measures on health, nutrition and income levels and expenditures on resources and

assets.

Definition of poverty for this research

Based on the above analysis we are able to give a founded working definition of poverty for this

research:

Poverty is the lack of sufficient access to financial and material assets, and public and natural

resources, as to ensure being nutritioned and healthy.

So as well-being to be defined as not being poor.

4.4 Indicators of poverty

Further, this chapter will briefly elaborate on the (sub)dimensions of poverty and investigate indicators

that measure them, in parallel with the exploration of the available data. Data source is the

‘Questionnaire des Indicateurs de Base de Bien-être’ (QUIBB)9 performed between April and July

2003 on more than 10,000 households in Burkina Faso (INSD, 2003). The selected indicators are

given in a table for each main dimension. Note between brackets refers to the measurement scale of

the indicator. More detailed information on selected indicators is given in Appendix B.

9 Translated: Questionnaire of indicators of basic well-being

Page 56: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

34

Income

A common method used to measure poverty is based on income or consumption levels. In literature,

an extensive discussion about the choice between these monetary indicators of poverty can be found.

Most analysts argue that, provided the information on consumption obtained from a household survey

is detailed enough, consumption will be a better indicator of poverty measurement than income

(Coudouel et al., 2002). For this research we advocate for income as a main dimension of poverty, but

include expenditures on food and health in the analysis as proxies for consumption. Hereby, also the

potential weakness of monetary indicators in poor rural agrarian economies is covered.

Table 4-2: Selected indicators for income

Sub-dimension Indicator

Income diversication –

Total income Total revenue (continuous, CFA/yr)

Income diversication –

Income from employment

Salary from public sector (continuous, CFA/yr)

Salary from private sector (continuous, CFA/yr)

Income diversication –

Income from entrepreneurship

Revenue from rent (continuous, CFA/yr)

Revenue from interest (continuous, CFA/yr)

Income diversication –

Income form (owned) resources

Revenue from agriculture (continuous, CFA/yr)

Revenue from dairy farming (continuous, CFA/yr)

Education –

Education level Highest level of education reached (categorical, 4 items)

Education –

Access to schools

Proximity of primary school (categorical, 5 items)

Proximity of secondary school (categorical, 5 items)

Nutrition

Nutrition is often called an investment in development, since better nutrition has proven to increase

intellectual capacity and thus increases the (future) ability to obtain other types of assets that are

essential for development. Literature shows that foetal and infant undernutrition affects children’s later

school enrolment, educational attainment, cognitive ability, and lifetime earnings and labour

Page 57: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

35

productivity (Haddad, 2002). In this perspective, nutrition is closely related to the other dimensions of

poverty, i.e. health and income.

Multiple methods are available for measuring malnutrition. One alternative approach is measuring

nutritional outcomes by the investigation of child nutritional status based on anthropometric surveys.

Malnutrition is diagnosed when individuals' anthropometric measurements in terms of weight, height

and age fall below international reference standards. Poor growth in infants and children, as well as

underweight in adults may be the consequence of both inadequate food intake and poor absorption of

food caused by environmental factors, such as infections or inadequate parental care (FAO, 2003).

Three of the most reliable indices for malnutrition are stunting, wasting and underweight. Stunting is

an indicator of chronic malnutrition, the result of prolonged food deprivation and/or illness. Wasting is

an indicator of acute malnutrition, the result of more recent food deprivation or illness. Underweight is

used as a composite indicator, to reflect both acute and chronic malnutrition, although it cannot

distinguish between them (Nandy et al., 2003). Notice that these three indicators show considerable

overlap, for this research we argue to use one single variable that integrates them.

The dimension nutrition is divided into three sub-dimensions: (1) food security, (2) physical state

related to famine and (3) expenditures on food. Food security exists when all people, at all times, have

access to sufficient, safe and nutritious food to meet their dietary needs and food preferences for an

active and healthy life (Website FAO SPFS).

Table 4-3: Selected indicators for nutrition

Sub-dimension Indicator

Food security –

Occurrence of food insecurity

Occurrence of problems satisfying nutrition needs (categorical,

5 items)

Food security –

Access to food market Proximity of (local) food market (categorical, 5 items)

Food security –

Access to owned resources

Surface of landholding (continuous, # hectares)

Number of livestock (continuous, # cattle)

Food security –

Access to (direct) nutritional

resources

Access to stocks (of cereals) until the next harvest

(dichotomous, 2 items)

Value of autoconsumption of nutritional production (continuous,

CFA/month)

Page 58: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

36

Famine –

Malnutrition

Child is wasted (low weight for height) (dichotomous, 2 items)

Child is stunted (low height for age) (dichotomous, 2 items)

Child is underweight (low weight for age) (dichotomous, 2

items)

Child is stunted and/or wasted and/or underweight

(dichotomous, 2 items)

Food consumption –

Expenditures on food

Value of expenditures on nutritional products (continuous,

CFA/month)

Health

Beyond its intrinsic value to individuals, health clearly is a central issue in poverty reduction and

overall human development. Indeed, three of the Millennium Development Goals (MDGs) call for

global health improvements by 2015, among which the reduction of child mortality (OECD/WHO,

2003). The international community agreed upon that enjoying the highest attainable standard of

health is one of the fundamental rights of every human being. For poor people especially, health is

also a crucially important economic asset – i.e. form of human capital – that increases an individual

capability.

Waterborne diseases are due to the consumption of (with pathogenic micro-organisms) contaminated

water, or from species which existence – or lifecycle – is related to water (Website WHO).

Contaminated water that is used in food preparation can be the source of foodborne diseases.

According to the World Health Organization, diarrhoeal disease is responsible for the deaths of 1.8

million people every year. It is estimated that 88% of that burden is attributable to unsafe water supply,

sanitation and hygiene, and is mostly concentrated on children in developing countries (WHO, 2007).

Sanitary measures are the interventions – usually construction of facilities such as latrines – that

improve the management of excreta10 (WSSCC, 2005). It encompasses all services that maintain

hygienic conditions to prevent diseases, as the collection and treatment of wastewater, either

centralized or on-site by e.g. septic tanks and latrines. In a broader sense, sanitation also includes the

collection and disposal of solid wastes.

Of all water-related diseases malaria is the most severe one. Where Sub-Saharan Africa already

carries the highest per capita burden of disease in the world, malaria is the single most important

disease in this – being responsible for nearly one million deaths and 300 to 500 million clinical cases

every year (MARA/ARMA, 1998). In this research we use the occurrence of fever and diarrhoea as

10 See Glossary

Page 59: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

37

proxy for prevalence of waterborne diseases. Additionally, sanitation access levels are measured by

manner of garbage and toilet wastewater disposal.

The dimension health is explained by three sub-dimensions: (1) health security, (2) physical health

and (3) expenditures on health. Health security encompasses having sufficient access to (public)

health services, (reliable) potable water and sanitation facilities.

Table 4-4: Selected indicators for health

Sub-dimension Indicator

Health security –

Access to health services Proximity of (public) health service (categorical, 5 items)

Health security –

Access to potable water

Proximity of potable water source (categorical, 5 items)

Availability of (improved) potable water source (categorical, 3

items)

Health security –

Access to sanitation facilities

Access to (improved) toilets (categorical, 4 items)

Access to (improved) garbage disposal (categorical, 6 items)

Physical –

Disease and disability

Recent prevalence of disease (dichotomous, 2 items)

Chronic prevalence of handicap or injury (dichotomous, 2

items)

Physical –

Water-related disease

Recent prevalence of diarrhoea (dichotomous, 2 items)

Recent prevalence of fever (dichotomous, 2 items)

Health consumption –

Expenditures on health services

Value of expenditures on consultation (CFA/past month)

Value of expenditures on medical analysis (CFA/past month)

Value of expenditures on medicines (CFA/past month)

Value of expenditures on hospitals (CFA/past month)

Value of expenditures on other medical services (CFA/past

month)

Total value of expenditures on health services (CFA/past

month)

Page 60: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in
Page 61: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

39

5. Linking Storage to Poverty

The previous chapter identified income, nutrition and health as main dimensions that explain poverty.

This chapter determines the theoretical (inter)relations between the selected dimensions of poverty

and the socio-economic values of small reservoirs. By doing so, it provides the conceptual model that

will be the base for the application of statistical methods in the next phase of this research.

5.1 Storage as explanatory factor for poverty

Earlier (Chapter 3, Table 3-1) we identified the socio-economic values – goods and services – that

small multi-purpose reservoirs can provide to households living nearby them. We recall that goods

provided are domestic, agricultural and animal water supply, and food, nutrient and raw material

supply. It is hypothesized that these goods have a – more or less – direct and often positive relation to

health, nutrition or income; hence poverty. Services that can be provided by small reservoirs are flood

and drought mitigation, groundwater recharge, water quality improvement, micro-climate stabilization

and biodiversity maintenance. They support enhance alleviation by indirect use or functioning. The

availability of numerous goods would be considerably altered or disappear if the reservoir no longer

exists or performs its functions properly. For this reason services are addressed to have a secondary

impact on indicators of poverty.

In the table below a brief overview of possible impacts of small reservoirs on (sub)dimensions of

poverty. The table shows that socio-economic values of small reservoirs result in an impact on all

three main-dimensions of poverty. These relations are also present in the cause-effect diagram that is

given at the end of this chapter.

Page 62: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

40

Table 5-1: Storage and poverty links

Goods supplied by small reservoirs Impact on household poverty

Domestic water supply

Improved access to sanitation facilities

Improved access to nutritional resources by watering small

gardens

Improved access to nutritional resources by watering small

animals

Access to (better quality) potable water

Agricultural water supply Access to means of production of direct nutritional resources

Access to means of production to increase income

Animal water supply Access to means of production of direct nutritional resources

Access to means of production to increase income

Food and nutrient supply

Access to direct nutritional resources

Access to means of production of direct nutritional resources

Access to means of production to increase income

Raw material supply

Access to means of production of direct nutritional resources

Access to means of production to increase income

Access to materials to improve housing

5.2 Linking storage and poverty

In this section we describe the hypotheses on the linkages between storage and poverty, which are

visualized in the cause-effect diagram. We already, briefly, described the origin of the links between

storage and poverty (Table 5-1), but it is also recognize that the actual impact that can be achieved by

productive and household uses of small reservoirs will depend on interdependencies within the

storage-poverty interactive system and external factors. By interdependencies are meant the relations

between (sub)dimensions within either the storage or poverty component. There is a need to analyze

the interrelations; there can be a dependence of assets or uses onto each other, and competition can

occur between different assets (goods and services) or uses. As is seen from the cause-effect

diagram, within the poverty component, different (sub)dimensions or indicators are not independent

from each other; they are correlated. Furthermore, as there is a limited amount of water available, the

priority uses of water influence the overall effectiveness. Below, the assigned (inter)relations are

justified, but we start with evaluating the theory behind causality.

Page 63: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

41

Below, the theory and hypothesis on cause and effect between and within storage and poverty is

described exhaustively. Notes are made about the non-spuriousity and direction of hypothesized

relations. In most cases it is easy to determine which variable affects the next, however, in some

cases the temporal antecedence of cause and effect is difficult to determine. Only after application of

correlation analysis the ‘true’ existence of the causality is evaluated. This chapter deals with the

practical significance of correlations.

5.2.1 Links between storage and poverty

Storage - Income

Starting from the hypothesized links between storage11 and the income dimension of poverty: in the

cause-effect diagram income diversification is differentiated into three proxies that each have different

indicators – these are not mentioned in the diagram for simplicity reasons – see also Chapter 4. For

the link between storage and the indicator ‘total revenue’ it is hypothesized that in general a household

total income will increase with improved accessibility to water. Not only directly – through increased

access to means of production – but also indirectly; as time savings and improved health and

nutritional status will provide human capital12, hence enable employment.

Many scientists – like Bloom et al. (2000) – argue that the relation between health and income is one

of mutual reinforcement, i.e. the direction of the causality is not only from income to health, but runs

both ways. Better health leads to higher income, but there is also a positive feedback effect, giving rise

to a beneficial situation where health and income improvements are mutually reinforcing.

For the proxy income from (owned) resources (that are agriculture and dairy farming), the hypothesis

is that with increased access to water – assumed that there is sufficient availability – the income from

(owned) resources will increase. This is, in turn, based on the theory that production of agricultural and

dairy farming products depends on water access, and will increase with improved resources. However,

the increased production potential might not result in increased income since it might not be sold for

money, or can be used for own consumption (self sufficiency). Hence, the strength of the relation may

be diminished by interference of autoconsumption. Also external factors like the access to (local)

markets and milieu of residence (urban versus rural) are relevant in this concern.

11 Hereinafter the term ‘storage’ refers to the socio-economic values of small reservoirs measured by reservoir

density (province scale).

12 See Glossary

Page 64: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

42

The proxy income from entrepreneurship (composed of rent and interest revenues) is not directly

influenced by improved access to small reservoirs. However, storage may lead to increased total

income, better education and time savings, hence revenue from rent and interest as well.

The proxy income from employment (built up from salary from private and public sector) has a similar

indirect relation to storage; hence by means of considerable time savings and improved human

capital. With decreased distance to a source of water people spent less time fetching, and so the

hypothesis is that they would use their time for instance in employment. As already suggested before,

time savings give the opportunity to perform other activities; not necessarily employment. Obtaining

water often involves significant inconvenience of time spent in collection. Time spent collecting water

should be closely related to the distance to the source; the time required to collect water reduces the

time remaining for other activities such as cooking or farming, being employed or going to school

(DOW, 2001). Therefore, all activities that concur for time may diminish the strength of this relation.

More specifically, all indicators within the (main)dimension income are in a way related.

We classify education as sub-dimension of income. Many will argue that education to be a main

dimension of poverty; we subscribe the important role of education in the poverty reduction process,

however, we have chosen to give it the current status within this research. The hypothesis is that there

is a positive relation between storage and education level. Like income diversification, education will

benefit from improved health and nutritional status; hence, illness will negatively influence education

level. A more direct relation to storage would be due to time savings, as one spent less time on

fetching water it is possible to attain education. Many literature sources mention the drawback water

fetching has on education – particularly that of girls (Boelee et al, 2000; Website Unicef). Influential to

this relation would be the access to education facilities and other time consuming activities.

Storage - Nutrition

The sub-dimension food security is measured by four proxies of which occurrence of food insecurity is

seen as the most meaningful. The hypothesis is that the better access to water resources, the fewer

problems satisfying nutritional needs occur. This is based on the assumption that small reservoirs

provide a means to practice (small holder) agriculture and dairy farming, and to gather nutritional

products from in and around the reservoir.

Two other indicators within this dimension are ‘value of autoconsumption of nutritional products’ and

‘access to stocks of cereals until next harvest’, where the latter is seen as a proxy for stocks of

nutritional products in general. Both indicators have a causal relation to the satisfaction of nutritional

needs. With increased value of autoconsumption and sufficient stocks it is more likely that less

problems satisfying nutritional problems occur. All is – of course – relative to the number of household

Page 65: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

43

members. For the relation to storage it is hypothesized that as (agricultural) production increases both

autoconsumption levels as well as the availability of stocks will benefit from this. Unfortunately, no

direct or suitable indicators for production levels are available for this research.

The second sub-dimension – famine – is a result of food insecurity. It encompasses the

anthropometric measures related to weight, length and age of children under five. It is widely accepted

that poor nutrition in a child is reflected in failure to grow sufficiently (Nandy et al., 2003). The indicator

‘malnutrition’ is used as an aggregate measure of the indices wasted, stunted and underweight.

Food consumption levels are included by the indicator ‘expenditures on nutritional products’. These

are indirectly influenced by storage. Storage may lead to increased income levels, and so, people may

tend to spend more on nutritional products. It is a known phenomenon that as incomes rise, food

habits change in favour of more nutritional and more diversified diets (Molden, 2007). Note that, we

should verify this relation for autoconsumption levels; in case these are sufficient to ensure food

security, expenditures on food are less likely to increase. Moreover, access to (local) food markets is a

limiting factor for expenditures on food. Note that measures on expenditures are relative to the number

(and age) of household members.

Finally, indicators for the proxy access to owned resources are ‘surface of landholding’, ‘number of

livestock’. As the density of reservoirs increases local communities have access to sources for

livestock watering and irrigation water supply. Therefore, they are able to sustain a larger number of

livestock and find more beneficial agriculture practices. Naturally, the more of these resources

available the more input for income generation and food security the household beholds. Indirectly,

even measures of health security should improve. It is hypothesized that income from (owned)

resources and access to (direct) nutritional products show the most direct relations with access to

owned resources. While income from other sources (than agriculture and dairy farming) as well as

education levels may diminish.

Storage - Health

Likewise to nutrition, we have defined health security as a sub-dimension of health. It encompasses

the access to (public) health services, potable water and sanitation facilities. Only access to (public)

health services is considered as external influencing factor, since it is not related to the presence – or

absence – of small reservoirs. One may argue that access to potable water should be regarded as

external. We recall the assumption that the reservoirs’ water is appropriate for all types of uses with

the exception of the use as potable water source. However, literature reveals that in cases reservoir

water will be used for drinking water purposes (Boelee et al., 2000), or that the presence of a reservoir

will locally raise the groundwater table (Savadogo, 2006), accommodating improved digging and

Page 66: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

44

pumping of (potable) water. Therefore, we include a direct link between storage and the proxy access

to potable water.

The proxy access to sanitation facilities is not considered to be influenced by the presence of a water

source. Two indicators represent this proxy: ‘access to (improved) toilets’ and ‘access to (improved)

garbage disposal; thus, measures of hygiene that require water are not included.

A clear sub-dimension of health is physical health. We already mentioned several times the impact

physical health has on other dimensions of poverty. But how does storage influence physical health?

Actually, here we find the only (possibly) negative relation in the system. In literature, many examples

can be found where it can go wrong with the human interaction with surface water, called water-

related diseases. One important issue is malaria. While malaria prevalence is highly dependent on the

presence of (stagnant) surface water, many other factors will play a role, e.g. rainfall and temperature

(Website MARA/ARMA). In this research, fever is used as an indicator for malaria prevalence. The

hypothesis is that with higher reservoir densities the number of people suffering from malaria

increases. Secondly, sources of water-related diseases are unsafe drinking water and lack of

sanitation and hygiene. Main symptom is considered to be diarrhoea (WHO, 2007), this is used as a

second indicator for the treat storage has it.

Note that the above hypotheses are quite interesting from the research point of view, however, they

are influenced by a variety of factors that are kept outside the scope of this research (therefore, not

included in the conceptual model). This implies that we should be most careful in drawing any

conclusions on this issue. Moreover, malaria prevalence is highest just after the rainy season (Website

MARA/ARMA; Beiersmann et al., 2007), while the QUIBB interviews are performed just before and at

the start of the rainy season.

Health consumption is measured by the indicator ‘value of expenditures on health services’. This is

indirectly influenced by storage. Either when physical health status increases or decreases this has an

influence on expenditures on e.g. consultation, medicines and medical analysis. Also with improved

welfare and income levels, people might make use of health services in a more moderate way.

Therefore, access to (public) health services is a determining factor in this relation. Be aware that

measures of expenditures are relative to the household size.

5.2.2 External explanatory factors

The actual benefits of small reservoirs will depend on other constraints faced by poor people like

availability of labour, skills, infrastructure, equipment, presence and knowledge of markets for products

Page 67: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

45

and services, transport, quality control standards (Moriarty, 2003). All these factors originate from

outside the conceptual model (represented by the cause-effect diagram). They mainly influence the

direction and extent of impacts, rather than the existence of the impact itself. Possible factors should

be identified and assessed on their influence onto the proposed causal relations. The reservoir

characteristics – described in Chapter 3, Section 3.4 – serve as the framework for identifying external

factors. Recall that reservoir characteristics are the physical features, external environment and

internal processes. Some relevant external factors are weather and climate conditions, land-use,

hydrology and geology, demography and community composition, and the quality and presence of

different (social-economic) infrastructures. Many of these indicators are not included in the analysis, as

they fall outside the scope of this research.

Some external factors are present in the cause-effect diagram, since they are seen as proxy within

(sub)dimensions of poverty; they are supporting the definition of the (sub)dimension. This set of

relations can be considered as a first verification on the extent of the relations between storage and

poverty. Availability and proximity of (communal) services and resources is important to many poor

households. For instance, access to a (local) food market is relevant to food security; the amount of

expenditures on nutritional products and income from (owned) resources depend on it. Since trade

may not be done in monetary terms, also the relation to nutritional satisfaction should be tested. Also

access to schools determines people’s possibilities to overcome literacy, and so provide capabilities.

Health security includes the external factor access to (public) health service; with improved proximity

to hospitals and clinics, physical health is likely to increase and expenditures on health may increase

as well.

Other external factors are not included in the diagram because they represent a more generic

influence on the system. Examples of the latter are ‘age of the individual’ and ‘number of household

members’, which are used to test for the relativeness of expenditures, income and consumption levels.

Main controlling external variables are ‘population density’, ‘milieu of residence’, and ‘proximity of

(public) transport’.

As regards the relation between poverty and population density it is not surprising to find that the

direction of the relationship is neither obvious nor simple. World Bank economist Pritchett (1997) is

bolder: “Solid evidence that population growth is a cause or even an exacerbating condition of poverty

cannot be given because there is none". For this research, it is hypothesized that population density

generally positively influences the access to (communal) services and resources, as exploitation

becomes more beneficial due to higher demand. However, the hypothesis is that the relation between

storage and population density is one of mutual enforcement. One may reason as above: where

population density is higher the demand for water resources increases, and thus, it is beneficial to

build a reservoir. The opposite reasoning is also cogent: people tend to settle there were water

Page 68: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

46

resources are available. Hence, the temporal antecedence between cause and effect is not at all

clear. Aim is to find evidence for the leading direction of this relation, whereby other factors influencing

the relation are taken into account.

For the external factor ‘milieu of environment’, so whether a household is situated in a rural or urban

environment, it is hypothesized that in more urbanized areas – mostly situated in central Burkina Faso

– access to (communal) services and resources is higher. Furthermore, indicators that are related to

livelihood strategies may show differences, depending on environmental factors.

The influence of access to (public) transport is hypothesized to be mainly on access to services and

resources. This is supported by literature; the availability of physical infrastructure (roads) improves

the access and availability of food, access to health services and education facilities (Thimm, 1993).

Consequently, roads play a major role in achieving food security.

Explanation of Figure 5-2:

The figure shows the conceptual framework for the storage-poverty interactive system. The arrows

indicate the hypothesized relations, type of relation and direction of causality. A plus in the diagram

means that when the value of the independent variable increases, the value of the dependent variable

increases as well; hence, this does not always mean there is a positive effect. Once again, the main

dimensions of poverty are divided into sub-dimensions. Proxies and indicators are given in the light-

coloured boxes. Their item scales can be found in Appendix B.

Page 69: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

access toprimary/

secondaryschool

access to(local)food

market

access to(public)healthservice

level ofeducation

totalrevenue

recentprevalence of disease

value ofexpenditures

on healthservices

occurrenceof problemssatisfyingnutritional

needs

value ofexpenditureson nutritional

product

Health security

Income diversication

Physical health

Food security

Education

Food consumption

Health consum

ption

HEALTH NUTRITION INCOME

recentprevalenceof diarrhoea

recentprevalence

of fever

chronicprevalenceof handicap

or injury

access to(improved)garbagedisposal

access to(improved)

toilets

reservoirdensity

(provincescale)

STORAGE

incomefrom

employment

incomefrom

(owned)resources

incomefrom

entrepre-neurship

-

+

+

-

+

+ +

+

+ +

-

+

+

-+

-

access toowned

resources

+

Famine

prevalence ofmalnutrition

-

+

access to(improved)source ofpotablewater

-

+

access tostocks (ofcereals)until nextharvest

value ofautocons-umption

of nutritionalproducts

+

-

+

-

+

+

-

+

Figure 5-2: Conceptual model – the cause-effect diagram

Page 70: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in
Page 71: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

49

6. Verification of Relationships

This chapter aims to verify the hypotheses as stated in the previous chapter. By applying correlation

analysis on each of the linkages visualized in the conceptual model the existence of a (statistical)

significant relation is verified; hence, hypotheses are accepted or rejected. The correlation coefficient

also determines the (quantitative) extent to which two variables are related, and whether the type of

relationship is positive or negative. Note that correlation coefficients do not indicate the temporal

antecedence of the cause versus the effect; this should be pre-defined by theory (see Section 5.2).

This chapter first gives some basic knowledge that is required to interpret the outcomes of the

correlation analysis. Main outcome of this chapter is the test results of bivariate correlation, visualized

by means of the cause-effect diagram. Further, it will briefly discuss each test individually. Later –

Chapter 8 – will look at the model as a whole, integrate all statistical tests performed and give answers

to the research questions. Additionally, a more exploratory approach is applied by testing relations that

are not hypothesized by theory as direct (see Appendix D).

6.1 Model specification

We are interested in testing the experimental hypothesis (or prediction) that there exists a relation

between specific indicators of poverty and storage. The reverse possibility – stating that no relation

exists – is called the null-hypothesis. As to confirm or reject the null-hypothesis, inferential statistics

are reviewed. Basically, we calculate the probability that the estimated coefficient is not occurring by

chance. As this probability decreases, greater confidence is gained that the experimental hypothesis is

actually correct, and that the null-hypothesis can be rejected. Literature sources (Hair et al., 1998;

Field, 2005; Fischer, 1991) suggest that only when the probability of a genuine result – i.e. the result is

not found by change – is 95% or more it can be accepted as being true. This implies that the

probability of the estimation being found by change is 5% or less. In statistical terms this is called a

statistically significant finding.

From the data screening process (see Appendix C) we have learned that none of the indicators are

normally distributed. Theory prescribes that in case of violation of the normality assumption, the

Spearman’s correlation coefficient should be selected. In cases that one of the variables is

dichotomous Pearson correlation analysis is applied. The Pearson correlation coefficient is denoted as

rp, the Spearman correlation coefficient is denoted as rs.

Page 72: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

50

incomefrom

employment

incomefrom

(owned)resources

level ofeducation

totalrevenue

proximityof primary/secondary

school

Income diversication

Education

INCOME

reservoirdensity

(provincescale)

STORAGE

incomefrom

entrepre-neurship

+.173+.147

+.269

-.144

+.194

Note all correlations are tested one-tailed, unless mentioned otherwise. Also, all ordinal variables are

considered as of interval measurement level, this requires extra attention when interpreting (the sign

of) the correlation coefficient.

6.2 Bivariate correlation analysis

The strategy for testing is to build the model bottom-up. Hence, it starts with assessing direct links

between storage and poverty, followed by the indirect linkages and interrelations (that lead to the

indirect links). Finally, correlations of external factors with the storage-poverty interactive system are

estimated.

6.2.1 Direct links between storage and poverty

Storage - Income

The hypothesis is that between storage and

income diversication there should be a positive

relation of some importance. So when the

reservoir density increases, the total income will

increase; hence a positive correlation. The test

confirms this hypothesis: rs = +.173 (p<.01). The

analysis provides the indication that storage has

a negative relation to income from (owned)

resources: rs = -.144 (p<.01). Of which the

correlation with revenue from agriculture and

from dairy farming are respectively rs = -.114

(p<.01) and rs = -.088 (p<.01). Indicating that

with higher reservoir density, the income from

(owned) resources will decrease. This is the

opposite from the hypothesis that – from all

other revenues – these revenues are the most

influenced by the presence of small reservoirs.

All other sources of income show a positive

relation to storage.

Figure 6-1: Correlations between storage and income

Page 73: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

51

proximityof (local)

foodmarket

occurrenceof problemssatisfyingnutritional

needs

prevalence ofmalnutrition

value ofexpenditureson nutritional

product

Food security

Famine

Food consumption

NUTRITION

reservoirdensity

(provincescale)

STORAGE

access tostocks (ofcereals)until nextharvest

value ofautocons-umption

of nutritionalproducts

+.229

(+).041

+.058

(-).082

-.254

surface of landholding

number oflivestock

-.318

-.214

Remarkable is the relatively high positive correlation between storage and income from employment:

rs = +.269 (p<.01). Therefore, the interpretation may be that the benefits of reservoirs are highly

influenced by time savings and improved human capital. A similar conclusion can be drawn for the

relation between storage and education level. Where we hypothesize a positive relation, the test

confirms this: rs = +.194 (p<.01).

Storage - Nutrition

An important hypothesis within this research is

that food security should improve due to the

proximity of small reservoirs. Hence, the

hypothesis is that the better access to water

resources – i.e. higher density – the fewer

problems satisfying nutritional needs the

household faces. However, the indicator

‘occurrence of problems satisfying nutritional

needs’ does not show to have a high direct

correlation to storage, moreover, the correlation

is positive: rs = +.058 (p<.01), suggesting that

with higher densities the household faces more

often problems satisfying her nutritional needs.

In a way this outcome is supported by the

following results: the indicator ‘value of

autoconsumption’ shows a relatively high, but

negative correlation: rs = -.254 (p<.01), indicating

that with increasing reservoir density, the value

of autoconsumption is decreasing. This is

neither consistent with the hypothesis. Also the

indicator ‘access to stocks until the next harvest’

shows a high, and negative correlation to

storage: rp = (-).082 (p<.01). The interpretation is

that stocks are not sufficient due to higher

reservoir densities.

Figure 6-2: Correlations between storage and nutrition

Page 74: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

52

chronicprevalenceof handicap

or injury

recentprevalenceof diarrhoea

recentprevalenceof disease

recentprevalence

of fever

value ofexpenditures

on healthservices

Health security

Physical health

Health consum

ptionHEALTH

reservoirdensity

(provincescale)

STORAGE

access to(improved)garbagedisposal

access to(improved)

toilets

+.118

.000

(+).011

availability of(improved)

potablewatersource

proximity ofpotablewatersource

proximityof (public)

healthservice

+.201

+.056

For both indicators of access to owned resources the correlation coefficient is relevant, but negative;

suggesting that with higher reservoir densities the surface of owned land and number of owned cattle

decreases. And, although the correlation between storage and ‘prevalence of malnutrition’ does not

show to be important, also this correlation is not consistent with the hypothesis that nutritional status

should improve due to storage. Only ‘expenditures on nutritional products’ are positively influenced by

storage: rs = +.229 (p<.01), suggesting that expenditures increase with higher reservoir density.

Summarizing, no evidence is found that storage positively influences measures of food security, nor

direct sources of nutritional products. Again we should apply tests for controlling factors as household

size and access to (local) food markets. For the value of autoconsumption a controlling variable would

be production levels, however, there are no indicators available to perform a partial correlation test.

Storage - Health

An interesting hypothesis concerns the human

interaction with small reservoirs. The hypothesis

is that with higher reservoir density the

prevalence of waterborne diseases will increase.

Main symptoms for these diseases are

considered diarrhoea for unreliable drinking

water, sanitation and hygiene, and fever for the

main water related disease malaria; hence we

use these as indicators for the threat storage

has on physical health. Tests show that the

correlation between storage and ‘prevalence of

diarrhoea’ is insignificant, and the correlation to

‘prevalence of fever’ is unimportant: rp = (+).011

(p<.01). The conclusion may be that the direct

negative impact of storage on (physical) human

health is negligible. The hypothesis is that

storage has a positive effect on access to

potable water. It may be the case that the

proximity and availability of a potable water

source improves with more proximate

reservoirs. The analysis shows that the

correlations are respectively rs = +.056 (p<.01),

Figure 6-3: Correlations between storage and health

Page 75: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

53

and rs = +.201 (p<.01), suggesting that reservoir density has no important impact on the proximity of

potable water, but that the availability of an (improved) potable water source generally increases.

As storage has no real negative effect on the prevalence of waterborne diseases and a positive effect

on access to potable water we assume that the relation between storage and ‘expenditures on health’

is positive. This hypothesis is rejected by the statistical test: rs = +.118 (p<.01), suggesting that with

increased reservoir density also expenditures on health increase. Note that measures on health

security and physical health are not the only determining factors influencing the value of expenditures

on health; additional tests need to be performed.

6.2.2 Indirect links (and interdependencies) between storage and poverty

Income

Education is hypothesized to have a positive impact on all sources of income. The test shows this is

true for the relation to total revenue, income from employment and income from entrepreneurship.

Again income from (owned) resources is violating this hypothesis: rs = -.251, suggesting that

individuals with a lower education level have more income from (owned) resources, and less income

from employment. Overall, the higher level of education, the more income is obtained.

It is hypothesized that, in general, total revenue will not only increase directly from improved

accessibility of water, but also indirectly as storage potentially leads to improved health and nutritional

status; hence will have a positive effect on human capital. The relation between income and nutrition

is one of mutual dependence. As income levels increase the expenditures on nutritional products may

increase as well. This is supported by the statistical test: rs = +.442. Consequently, increased food

consumption levels are likely to lead to improved food security and nutritional status. In turn, these

improvements may have a positive impact on income and education levels, due to time savings and

improved human capital. This is confirmed by the correlation analysis between ‘occurrence of

problems satisfying nutritional needs’ and ‘total revenue’: rs = -.180, suggesting that with less

nutritional problems income levels increase. The same positive effect is seen for the relation between

‘occurrence of problems satisfying nutritional needs’ and ‘level of education’: rs = -.106.

Likewise, the relation between income and health is one of mutual dependence. As increased income

level may enable higher expenditures on health. Unfortunately, the latter are mainly determined by

other factors, and only limited by income. Nevertheless, the analysis shows there is a positive relation:

rs = +.144 (p<.01). The hypothesis is now that, due to improved nutrition and higher expenditures on

health, the physical health of individuals will improve. However, the correlation analysis does not show

Page 76: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

54

interesting correlations from indicators of prevalence of diseases and disability to ‘total revenue’.

Neither education levels are influenced by disease and disability.

Table 6-1: Correlations for interdependencies: income

Independent variable Dependent variable Coefficient

Total revenue rs = +.256 (p<.01)

Income from (owned) resources rs = -.251 (p<.01)

Income from employment rs = +.402 (p<.01)

Level of education

Income from entrepreneurship rs = +.168 (p<.01)

Expenditures on nutritional products rs = +.442 (p<.01) Total revenue

Expenditures on health services rs = +.144 (p<.01)

Occurrence of problems satisfying

nutritional needs Total revenue rs = -.180 (p<.01)

Recent prevalence of disease Total revenue Insignificant

Chronic prevalence of handicap or injury Total revenue rp = (-).009 (p<.01)

Nutrition

Although storage does not seem to have a positive direct impact on measures of nutrition, we are

interested in the interrelations within this dimension. According to the conceptual model, the proxy

access to owned resources has a positive relation to ‘value of autoconsumption of nutritional products’

and ‘access to stocks until next harvest’. All test results show that with increasing land and livestock

holding both autoconsumption levels as access to stocks are positively influenced. Additionally,

access to owned resources is hypothesized to have a positive effect on income from (owned)

resources. The analysis confirms this hypothesis as correlations are relatively high and positive.

It is hypothesized that the occurrence of food insecurity reduces due to higher autoconsumption

levels. Unfortunately, the analysis does not show an interesting correlation. Remarkable is the

relatively high positive contribution of ‘access to stocks (of cereals) until next harvest’: rp = (-).375. For

completeness we note that the correlation between ‘value of autoconsumption of nutritional products’

and ‘access to stocks (of cereals) until next harvest’ is relatively high, and positive: rp = (+).170 (2-

tailed). Conclusion may be that autoconsumption does not have a direct relation to ‘occurrence of

problems satisfying nutritional needs’, but that ‘access to stocks until next harvest’ has. We apply a

controlling test for the impact of ‘expenditures on nutritional products’. The analysis suggests that

more a higher level expenditures on food lead to less food insecurity.

Page 77: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

55

Note that there is a trade-off between expenditures on food and autoconsumption levels. The

correlation coefficient rs = -.350 (2-tailed), suggests that when expenditures increase, autoconsumption

levels decreases. The overall conclusion may be that stocks (of cereals) are mostly formed by own

production than from marketed products. Consequently, we may conclude that expenditures on

nutritional products have a major impact on food security, relative to value of consumed own food

production. Note that we should apply tests to control for access to local markets, measures of income

(later in this section), household size and milieu of residence.

Table 6-2: Correlations for interdependencies: nutrition

Independent variable Dependent variable Coefficient

Value of autoconsumption of nutritional

products rs = +.525 (p<.01)

Access to stocks (of cereals) until next

harvest rp = (+).161 (p<.01)

Surface of landholding

Income from (owned) resources rs = +.392 (p<.01)

Value of autoconsumption of nutritional

products rs = +.487 (p<.01)

Access to stocks (of cereals) until next

harvest rp = (+).137 (p<.01)

Number of livestock

Income from (owned) resources rs = +.532 (p<.01)

Value of autoconsumption of nutritional

products

Occurrence of problems satisfying

nutritional needs rs = -.023 (p<.01)

Occurrence of problems satisfying

nutritional needs rp = (-).375 (p<.01)

Access to stocks (of cereals) until next

harvest

Value of autoconsumption of nutritional

products rp = (-).170 (p<.01)

Occurrence of problems satisfying

nutritional needs rs = -.129 (p<.01)

Value of autoconsumption of nutritional

products rs = -.350 (p<.01)

Value of expenditures on nutritional

products

Access to stocks (of cereals) until next

harvest rp = (+).039 (p<.01)

Occurrence of problems satisfying

nutritional needs Prevalence of malnutrition rp = (-).054 (p<.01)

Recent prevalence of disease rp = (-).063 (p<.01) Prevalence of malnutrition

Chronic prevalence of handicap or injury rp = (-).031 (p<.01)

Value of expenditures on nutritional

products Value of expenditures on health services rs = +.218 (p<.01)

Page 78: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

56

As ‘occurrence of problems satisfying nutritional needs’ is considered to be the main indicator of food

security it should have a (positive) relation to famine. Recall that as indicator for famine we use the

aggregate measure malnutrition that represents whether or not a child is wasted and/or stunted and/or

underweight13. The relation between ‘occurrence of problems satisfying nutritional needs’ and

‘prevalence of malnutrition’ is hypothesized to be direct: with little or no nutritional problems the

prevalence of malnutrition should diminish. Unfortunately, the correlation is rather weak and the sign

of the coefficient rejects the hypothesis. Neither other indicators of food security show interesting

correlations with ‘prevalence of malnutrition’. Overall conclusion is that all correlations from food

security and food consumption to famine are weak.

Naturally, measures of nutrition are related to measures of health. When assessing the relation

between nutritional status and heath status we find that both indicators are weak related to

malnutrition. Therefore, we can draw the conclusion that no real relation between measures of famine

and physical health exists. Note that all indicators of both are dichotomous, so this interpretation is

biased by the measurement scale. Neither other measures of nutrition show important correlations to

measures of physical health and health consumption. For the relation from ‘expenditures on nutritional

products’ to ‘expenditures on health services’: rs = +.218, suggesting that the increase of food

consumption leads to increasing healthcare consumption, which is a rather unexpected result. Overall

we can conclude that the relation between food security and physical health are not very strong in this

research. We also recognize that improved nutritional status (‘prevalence of malnutrition’) and – more

indirectly – food insecurity (‘occurrence of problems satisfying nutritional needs’) are factors

influencing an individuals’ health status.

Health

Many external factors play a role in health security, however, the proxy access to sanitation facilities is

considered to be internal. It is hypothesized that with improved access to sanitation facilities –

measured by manner of garbage and toilet wastewater disposal – the prevalence of diarrhoea will

decrease. However, the analysis shows no relations. Besides improved sanitation also indicators of

access to potable water are hypothesized to influence the prevalence waterborne diseases. Again, the

results do not allow conclusions since correlations are dispensable. We have already seen that the

direct relation between storage and waterborne diseases is also of low extent; this might explain the

weak relations that we find here.

13 See Glossary

Page 79: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

57

While it is hypothesized that storage has a direct impact to the prevalence of waterborne diseases,

many other diseases are mainly due poor nutritional status – primarily by famine, secondarily by food

security and food consumption – or external factors. Prevalence of disease is largely related to the

prevalence of water related diseases. No important correlations are found between waterborne

diseases and ‘chronic prevalence of handicap or injury’.

Table 6-3: Correlations for interdependencies: health

Independent variable Dependent variable Coefficient

Access to (improved) toilets Recent prevalence of diarrhoea Insignificant

Access to (improved) garbage disposal Recent prevalence of diarrhoea rp = (-).010 (p<.05)

Recent prevalence of diarrhoea Insignificant Proximity of potable water source

Recent prevalence of fever rp = (-).016 (p<.01)

Recent prevalence of diarrhoea rp = (-).011 (p<.01) Availability of (improved) potable water

source Recent prevalence of fever rp = (-).012 (p<.01)

Recent prevalence of diarrhoea Recent prevalence of disease rp = (+).377 (p<.01)

Recent prevalence of fever Recent prevalence of disease rp = (+).663 (p<.01)

Recent prevalence of disease Value of expenditures on health services rp = (+).086 (p<.01)

Chronic prevalence of handicap or injury Value of expenditures on health services rp = (+).023 (p<.01)

Access to (improved) toilets Value of expenditures on health services rs = +.143 (p<.01)

Access to (improved) garbage disposal Value of expenditures on health services rs = +.079 (p<.01)

The third sub-dimension is health care consumption. Expenditures on health care are hypothesized to

be indirectly influenced by storage: due to chancing physical health status expenditures on e.g. health

case, medicines and medical analysis change in the opposite direction. Also with improved welfare

and income levels, people might use health services in a more moderate way. The overall influence of

storage on health may be negative, however, we should be careful to conclude upon this since the

above analysis did not indicate strong negative relations between storage and (waterborne) diseases,

the relation is hypothesized to be indirect and many other (external) factors should be considered. The

correlation analysis suggests that health care consumption levels are not strongly influences by

disease and disability.

Page 80: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

58

6.2.3 Links with external factors

The selected paths include factors that origin from outside the storage – poverty interactive system

(represented by the causal relation diagram). They mainly influence the direction and extent of

impacts, rather than the existence of the impact itself. Some factors are present in the causal relation

diagram, since they are seen as indicator within (sub)dimensions of poverty; they are supporting the

definition of the (sub)dimension. For the underlying theory we direct to Chapter 5, Section 5.2.

External factors - Storage

Three external factors influence storage – reservoir density – that are population density, type of the

environment (urban versus rural) and the proximity of (public) transport. So far, the direction of the

causal relation between storage and those factors is not always clear; e.g. high reservoir density may

cause more people to find a living around them, but on the other hand, high population densities may

urge communities to built small reservoirs. The correlation analysis shows the following results:

Table 6-4: Correlations external factors: external factors

Independent variable Dependent variable Coefficient

Population density Reservoir density rs = +.751 (p<.01)

Milieu of residence Reservoir density rp = (-).485 (p<.01)

Proximity of (public) transport Reservoir density rs = +.103 (p<.01)

Suggesting that with higher reservoir densities also population densities increase, or, that with higher

population densities also reservoir densities increase. This relation is expected, however, the question

remains which of the two factors acts as driving force. Further, the analysis gives that reservoirs are

more densely in rural environments, and that where reservoirs are densely distributed, (public)

transport is more easily reached.

External factors – Income

The external factor that is included in the dimension income is access to primary/secondary schools.

Naturally, the main impact should be on ‘level of education’ (is a more direct impact). The analysis

shows that proximity of primary schools has the least impact; however, both results suggest that with

improved access to schools education levels increase. Secondly, the hypothesis is that with increased

access to schools (both primary as secondary) measures of income should increase. The correlation

analysis to ‘total revenue’ confirms the hypothesis with the following correlations. Differentiated tests

Page 81: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

59

show that the main impact is on income from employment, while there is a negative impact on income

from (owned) resources.

Other external factors influencing the dimension income are population density and access to (public)

transport. The hypothesis is that population density leads to increasing proximity of schools, and

indirectly, to higher income and education levels. This is confirmed by the correlation analysis. Also

access to transport positively influences the above measures of income.

Table 6-5: Correlations external factors: income

Independent variable Dependent variable Coefficient

Level of education rs = +.260 (p<.01)

Total revenue rs = +111 (p<.01)

Income from employment rs = +.286 (p<.01)

Proximity of primary school

Income from (owned) resources rs = -.319 (p<.01)

Level of education rs = +.385 (p<.01)

Total revenue rs = +.271 (p<.01)

Income from employment rs = +.431 (p<.01)

Proximity of secondary school

Income from (owned) resources rs = -.311 (p<.01)

Proximity of primary school rs = +.297 (p<.01)

Proximity of secondary school rs = +.409 (p<.01)

Total revenue rs = +.224 (p<.01)

Population density

Level of education rs = +.272 (p<.01)

Proximity of primary school rs = +.487 (p<.01)

Proximity of secondary school rs = +.621 (p<.01)

Total revenue rs = +.164 (p<.01)

Proximity of (public) transport

Level of education rs = +.269 (p<.01)

External factors – Nutrition

One external factor for nutrition is included in the causal relation diagram, namely access to (local)

food market. It is hypothesized that this indicator will influence measures of food security,

expenditures on nutritional products and income from (owned) resources. The test for bivariate

correlations suggest that with increased proximity of food markets, the value of autoconsumption

decreases, but the expenditures on food increases. Additionally, it is hypothesized that with more

Page 82: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

60

proximate food markets, income from (owned) resources increases, however the test shows the

opposite. Remarkably is that, when we test the relation between the proximity of a food market and

‘total revenue’ we do find a positive correlation: rs = +.120.

External factors influencing the access to a (local) food markets are population density, milieu of

residence and access to (public) transport. The hypothesis is that population density, in general,

positively influences the access to (communal) services and resources. The correlation analysis

confirms this hypothesis with respect to the proximity of food markets: rs = +.274. Additionally, in an

urban environment food markets are more proximate.

Table 6-6: Correlations external factors: nutrition

Independent variable Dependent variable Coefficient

Occurrence of problems satisfying

nutritional needs rs = -.095 (p<.01)

Value of autoconsumption of

nutritional products rs = -.247 (p<.01)

Value of expenditures on nutritional

products rs = +.284 (p<.01)

Income from (owned) resources rs = -.291 (p<.01)

Proximity of (local) food market

Total revenue rs = +.120 (p<.01)

Population density Proximity of (local) food market rs = +.274 (p<.01)

Milieu of residence Proximity of (local) food market rp = (-).320 (p<.01)

Proximity of (public) transport Proximity of (local) food market rs = +.601 (p<.01)

Occurrence of problems satisfying

nutritional needs rs = +.088 (p<.01)

Household size

Total food consumption rs = +.421 (p<.01)

Naturally, we should control indicators of nutrition for household size. With increasing number of

household members it is more likely that the household faces problems, hence score high on

‘occurrence of problems satisfying nutritional needs’, and have higher food consumption levels. The

analysis shows that the occurrence of nutritional problems is not highly correlated to the number of

household members, even though, the result suggests that with increasing household size problems

satisfying needs occur more frequently. Food consumption levels (summated) are more sensitive to

‘household size’: rs = +.421.

Page 83: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

61

External factors – Health

As configured in the causal relation diagram, we assume better physical health with increased access

to (public) health services. The correlation analysis confirms the positive impact of access to (public)

health services to indicators of health security, physical health and health consumption. Correlations to

measures of physical health are tested negligible.

Table 6-7: Correlations external factors: health

Independent variable Dependent variable

Expenditures on health services rs = +.123 (p<.01)

Recent prevalence of disease rp = (+).019 (p<.01)

Proximity of (public) health services

Chronic prevalence of handicap or injury Insignificant

Proximity of (public) transport Proximity of (public) health services rs = +.631 (p<.01)

Proximity of (public) health services rs = +.361 (p<.01)

Proximity of potable water source rs = +.145 (p<.01)

Availability of (improved) potable water

source rs = +.198 (p<.01)

Access to (improved) toilets rs = +.397 (p<.01)

Population density

Access to (improved) garbage disposal rs = +.203 (p<.01)

The hypothesis is that population density leads to increasing access to (communal) services and

resources. For each of the above indicators higher population densities have a positive effect.

Explanation of Figure 6-4:

The figure shows the correlations within the storage-poverty system. The arrows show the hypothetical

relations and the results of the correlation analysis. The colours indicate whether the hypothesis is

confirmed (blue) or rejected (red) by the statistical test.

Page 84: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

+.095

proximityof (local)

foodmarket

proximityof (public)

healthservice

totalrevenue

recentprevalenceof disease

value ofexpenditures

on healthservices

occurrenceof problemssatisfyingnutritional

needs

value ofexpenditureson nutritional

products

Health security

Income diversication

Physical health

Food security

Education

Food consumption

Health consum

ption

HEALTH NUTRITION INCOME

recentprevalenceof diarrhoea

recentprevalence

of fever

chronicprevalenceof handicap

or injury

access to(improved)garbagedisposal

access to(improved)

toilets

STORAGE: reservoir density (province scale)

incomefrom

employment

incomefrom

(owned)resources

incomefrom

entrepre-neurship

Famine

prevalence ofmalnutrition

access tostocks (ofcereals)until nextharvest

value ofautocons-umption

of nutritionalproducts

-

number oflivestock

proximityof primary

school

proximityof secondary

school

level ofeducation

surface of landholding

availability of(improved)

potablewatersource

proximityof potable

watersource

.000+.218 +.144 +.442

-.129

(-).054

(-).375

-.023

(+).170 -.350

-.095

+.525+.487

+.284

-.291

+.392

+.532

+.201

+.056

-.214

-.318

-.106 -.180-.251

+.402

+.168

+.123

+.143

.000 .000

.000

.000

(+).023

(+).086

.000

+.385

+.260

-.144 +.269 +.147

+.173

+.194

.000

Figure 6-4: Main correlations within the poverty-storage system

Page 85: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

63

6.3 Discussion

In this chapter we have seen that:

• Most correlations between storage and poverty and within poverty are representing small to

medium effects. Throughout the analysis we regard the relative strength of the correlation

coefficient; hence, correlations above 0.10 are referred to as interesting or important. There

are two possible causes for finding relatively low correlations:

Measurement error: the degree to which the observed values are not representative of

‘true’ values. Sources can be data entry errors, imprecision of the measurement (e.g.

representation of concepts) and the inability of respondents to accurately provide

information. Thus all variables used must be assumed to have some degree of

measurement error (Hair et al., 1998). The measurement error results more deviation

around the linear relationship, and thus, in less strong correlation coefficients;

Different scales of aggregation: variables are disaggregated to lower scale for testing.

Hereby they loose part of their variance which results in less strong correlation

coefficients;

• Accuracy of the chosen indicators for physical health and famine may be low since we do not

find many – hypothesized or not hypothesized – interesting correlations to other indicators.

This phenomenon might be due to their measurement level, measurement scale or extent of

missing data. However, due to the measurement scale (individual) – and related sample size

(N = 54035) – of measures of physical health, one may expect to find significant correlations

easier. As Hair et al. (1998) states: “With regard to sample size we need to be aware that at

any given alpha level, increased sample sizes always produce greater power of the statistical

test”;

• So far, the analysis did not go deeper into the effect of milieu of residence (urban versus rural)

onto the storage-poverty interactive system. However, the current analysis suggests there

may be important effects of both (external) factors. Therefore, a split-sample test is applied in

a later stage of this research – see Chapter 8.

Page 86: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in
Page 87: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

65

7. Quantification of Relationships

So far, the existence of relationships between two variables is verified and quantified. However, the

conceptual model (Figure 5-2) shows that in most cases more than one variable is affecting one other.

As the preceding analysis encompasses only one-to-one relations it does not tell anything about the

contribution of several causes on one effected variable, nor does it take relations between causes into

account. In this research, multiple regression analysis is applied to assess the relative contribution of

different independent variables on one dependent variable and involve interaction effects between

independent variables. Note that linear regression is a parametric technique. Non-parametric

alternatives are (multi-nominal) logistic regression or ordinal regression. However, Hair et al. (1998)

states that regression analysis has been shown to be quite robust even when the normality

assumption is violated.

7.1 Model specification

A good strategy to adopt for multiple regression modelling is to include predictor variables for which

there are sound theoretical reasons for expecting them to predict the outcome. Field (2005) suggests

running a regression analysis in which all predictors are entered into the model and examine the

output to see which predictors contribute substantially to the models ability to predict the outcome.

Once established which variables are important, re-run the analysis including only the important

predictors and use the resulting parameter estimates to define the regression model. However, in this

research we face three problems to this approach:

1. The complete model containing all explanatory variables – about 40 for this research – may

become too complex. Chen et al. (2004) even suggest this can result in inaccurate estimation

of the parameters and instability of the model structure.

2. The proposed conceptual model shows several relations of mutual dependence that require

iterative testing. This can not be embraced in only one multiple regression model, but requires

sequences of models. Statistical techniques are available for estimating systems of multiple

regressions; however, the available data violate all assumptions underlying those techniques.

3. Theory shows there is no single indicator representing ‘poverty’; hence, we know multiple

outcome variables. Fortunately, the anterior correlation analysis does provide sufficient insight

into the interactive system to provide a convenient starting point for regression analysis.

Page 88: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

66

The modelling strategy is now to estimate an overall regression model for each main-dimension of

poverty and to dissect for its included variables. Additionally, a regression model for storage is

estimated. Note that the type of regression analysis is determined by the measurement level of the

dependent variable.

Method of entering

A great deal of care should be taken in selecting predictors for the model because the values of the

regression coefficients depend upon the variables in the model (Field, 2005). Therefore, the predictors

included, and the way in which they are entered into the model can be of great impact. Only when

predictors are all completely uncorrelated the order of variable entry has a minor effect on the

parameters calculated. Often, actually usually, this is not the case.

Field (1995) suggests that in case there is a sound theoretical literature available, the model should be

based on this past research. As a general rule, the fewer predictors the better and certainly include

only predictors for which a good theoretical grounding is available. Note to respect also the minimum

required sample size of 20 cases per included predictor variable. Since the aim of this research is to

explain (inter)dependencies between storage and poverty, forced entry (or enter method) is applied. In

this method predictors are all forced into the model simultaneously. This method relies on theoretical

reasons for including predictors. However, no decision about the order of entry needs to be made. In

this research, both the causal relation diagram as the correlation analysis provides the theoretical

base for including predictor variables.

7.2 Multiple regression analysis

In this section the results of multiple regression analysis are given. For each main dimension of

poverty – income, nutrition and health – a representative indicator is chosen, for which a more

extensive model is estimated. For all variables included in this ‘complete’ regression model only the

relative contribution of only the direct effects are assessed. For a broad overview of the results of the

regression analysis see Appendix E.

7.2.1 External factors

Reservoir density

Throughout this research the indicator used to represent proximity of small reservoirs is ‘reservoir

density’. In most relations storage is a determining parameter, however, there are also variables that

determine storage. The hypothesis is that both population density and income determine storage.

Page 89: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

67

Note that the hypothesis does not explicitly state that the direction of the relation is from population

density towards storage; however, this direction is used during for this analysis.

The preceding analysis has already show reservoir density and population density are highly

correlated. This result is supported by the regression analysis as the explained variance by the

variables model is 79.6%. The relative contribution of population density is 89.8%, while income has a

small negative contribution (β= -3.6%). Clearly, population density is an important determining factor

behind storage.

Table 7-1: Regression model for reservoir density

Dependent variable Independent variable βi Sig.

Reservoir density Population density +.989 .000

Total revenue -.036 .000

Population density

Factors within this research that possibly influence population density are reservoir density – as

people tend to settle there were water resources are available – and the difference between rural and

urban environments.

Table 7-2: Regression model for population density

Dependent variable Independent variable βi Sig.

Population density Reservoir density +.787 .000

Milieu of environment (-).216 .000

From the above analysis the temporal antecedence between cause and effect remains unclear (as

expected). However, the hypothesis that population density generally positively influences the access

to (communal) services and resources, as exploitation becomes more beneficial due to higher demand

remains to be tested below.

7.2.2 Income

Total Revenue

Naturally, total revenue is directly influenced by its diversified sources – entrepreneurship,

employment and (owned) resources – but also by level of education, physical health, food insecurity

and storage play a role. The total variance (R2) explained by these factors is 76.4%. Clearly, the

Page 90: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

68

differentiated measures of income sources (1st three variables) determine the most variance on the

dependent variable. Prevalence of disease and disability does not have an effect on the total revenue.

Table 7-3: Regression model for total revenue (direct model)

R2= 76.4%

Dependent variable Independent variables βi Sig.

Total revenue Income from (owned) resources +.575 .000

Income from employment +.443 .000

Income from entrepreneurship +.448 .000

Level of education +.037 .000

Recent prevalence of disease .000 .955

Chronic prevalence of handicap and injury (+).002 .414

Occurrence of problems satisfying nutritional needs -.029 .000

Reservoirs density +.047 .000

Factors that are indirectly related to the total revenue are access to schools, food markets and

transport, and population density. After re-estimating the regression model including all variables the

explained variance on the dependent variable increases to 77.2%. The indicators of physical health

remain insignificant.

Table 7-4: Regression model for total revenue (complete model)

R2= 77.2%

Dependent variable Independent variables βi Sig.

Income from (owned) resources +.584 .000

Income from employment +.430 .000

Income from entrepreneurship +.448 .000

Level of education +.009 .000

Recent prevalence of disease (-).002 .436

Chronic prevalence of handicap and injury (-).002 .346

Occurrence of problems satisfying nutritional needs -.013 .000

Proximity of (local) food market +.017 .000

Total revenue

Proximity of primary school -.014 .000

Page 91: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

69

Proximity of secondary school +.058 .000

Reservoir density -.066 .000

Proximity of (public) transport +.026 .000

Population density +.113 .000

Cleary, income from agriculture and dairy farming has the largest and positive contribution to total

revenue (β= 58.4%), followed by income from employment (β= 43%) and income from

entrepreneurship (β= 44.8%). Population density has a small relative contribution (β= 11.3%), while

the relative contribution of reservoir density is negative (β= -6.6%). The low regression coefficient for

level of education (β= 0.9%) was not expected based on the correlation analysis. Therefore, it is

interesting to know which variables determine most of the variance on the three diversified sources of

income.

Income from owned resources

Livestock holding has the highest relative contribution to this source of income (β= 26.8%),

especially when compared to the contribution of land holding (β= 2.5%). Reservoir density,

education level and proximity of (local) food market are negatively related to income from owned

resources. This is consistent with the results of the correlation analysis, however, inconsistent with

the hypotheses.

Table 7-5: Regression model for income from (owned) resources

R2= 10.5%

Dependent variable Independent variables βi Sig.

Proximity of (local) food market -.063 .000

Surface of landholding +.025 .000

Number of livestock +.268 .000

Reservoir density -.053 .000

Level of education -.033 .000

Occurrence of problems satisfying nutritional needs -.046 .000

Recent prevalence of disease (+).005 .272

Income from

(owned) resources

Chronic prevalence of handicap and injury (-).002 .544

Page 92: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

70

Income from employment

The analysis shows that chronic prevalence of disability does not play a role in the explanation of

income from employment. Most determining is level of education, followed by the positive

contribution of storage. Furthermore, the role of food insecurity is positive, and larger than on any

other source of income.

Table 7-6: Regression model for income from employment

R2= 16.7%

Dependent variable Independent variable βi Sig.

Level of education +.302 .000

Occurrence of problems satisfying nutritional needs -.108 .000

Recent prevalence of disease (+).011 .007

Chronic prevalence of handicap and injury (-).003 .492

Income from

employment

Reservoir density +.165 .000

Income from entrepreneurship

The total explained variance by included variables is 1.9%, thus far too low to draw any valid

conclusions upon.

Table 7-7: Regression model for income from entrepreneurship

R2= 1.9%

Dependent variable Independent variables βi Sig.

Level of education +.094 .000

Occurrence of problems satisfying nutritional needs -.022 .000

Recent prevalence of disease (+).003 .476

Chronic prevalence of handicap and injury (+).004 .387

Income from

entrepreneurship

Reservoir density +.069 .000

Level of education

All included variables contribute significantly to the explained variance on ‘level of education’.

Likewise to the correlation analysis the largest contribution is from ‘proximity of secondary school’

(β= 30.8%), while the relative contribution of ‘proximity of primary school’ is much smaller (β=

Page 93: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

71

4.3%). Reservoir density is the second determining variable, moreover, the coefficient is positive

(β= 16.1%) – suggesting the hypothesis that storage contributes to considerable time savings.

Table 7-8: Regression model for level of education

R2= 19.4%

Dependent variable Independent variable βi Sig.

Proximity of primary school +.043 .000

Proximity of secondary school +.308 .000

Occurrence of problems satisfying nutritional needs -.089 .000

Recent prevalence of disease (-).004 .336

Chronic prevalence of handicap and injury (-).016 .000

Level of education

Reservoir density +.161 .000

Proximity of primary and secondary schools

The proximity of schools is hypothesized to be influenced by two external variables: population

density and proximity of (public) transport. The analysis shows that the influence of both is

positive, but that the influence of proximity of (public) transport has a 50% larger effect.

Table 7-9: Regression model for proximity of primary school

R2= 27.8%

Dependent variable Independent variables βi Sig.

Proximity of (public) transport +.443 .000 Proximity of

primary school Population density +.187 .000

Table 7-10: Regression model for proximity of secondary school

R2= 47.6%

Dependent variable Independent variables βi Sig.

Proximity of (public) transport +.554 .000 Proximity of

secondary school Population density +.298 .000

Page 94: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

72

7.2.3 Nutrition

In the conceptual model, the indicator ‘prevalence of malnutrition’ is the end-indicator for the

dimension nutrition. However, food insecurity (indicator ‘occurrence of problems satisfying nutritional

needs’) is seen as a more overall measure for nutritional status. Therefore, this indicator is used as

starting point for the regression analysis on nutrition.

Occurrence of problems satisfying nutritional needs

Directly, food insecurity is influenced by the value of autoconsumption and expenditures on nutritional

products, access to stocks (of cereals) until next harvest and (local) food markets. Consistent with the

correlation analysis the contributions of expenditures on food and access to stocks are largest and

positively influencing food security.

Table 7-11: Regression model for occurrence of food insecurity (direct model)

R2= 16.9%

Dependent variable Independent variables βi Sig.

Value of expenditures on nutritional products -.126 .000

Value of autoconsumption of nutritional products +.042 .000

Access to stocks (of cereals) until next harvest (-).389 .000

Occurrence of

problems satisfying

nutritional needs

Proximity of (local) food market -.066 .000

Indirectly, also income, land and livestock holding, reservoir density, and external factors population

density and access to (public) transport play a role. The relative contribution of food markets becomes

insignificant. Important determining variables are population density (β= -34.7%) and reservoir density

(β= 32.9%). Although their extent is equal, the sign of population density indicates a positive impact on

food insecurity – i.e. improvement towards food secure livelihood – while reservoir density has a

negative impact. As expected based on the correlation analysis, access to stocks is an important

determining factor (β= 38%).

Table 7-12: Regression model for occurrence of food insecurity (complete model)

R2= 20.9%

Dependent variable Independent variables βi Sig.

Value of expenditures on nutritional products -.064 .000 Occurrence of

problems satisfying Value of autoconsumption of nutritional products +.047 .000

Page 95: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

73

Access to stocks (of cereals) until next harvest (-).380 .000

Proximity of (local) food market -.024 .054

Total revenue -.083 .000

Number of livestock -.053 .000

Surface of landholding -.005 .640

Reservoir density +.329 .000

Population density -.347 .000

nutritional needs

Proximity of (public) transport -.069 .000

Again, differentiated regression models are estimated to investigate the determining factors for each of

the indicators, wherein only direct relations to the dependent variable are included.

Value of expenditures on nutritional products

Clearly, income has the highest relative importance to the dependent variable (β= 38.8%),

followed by the proximity of (local) food markets (β= 14.6%). The mutual dependency with direct

nutritional products is – as hypothesized – negative, however, of low relevance.

Table 7-13: Regression model for value of expenditures on nutritional products

R2= 19.1%

Dependent variable Independent variables βi Sig.

Total revenue +.388 .000

Proximity of (local) food market +.146 .000

Value of autoconsumption of nutritional products -.060 .000

Value of expenditures

on nutritional

products

Access to stocks (of cereals) until next harvest (-).035 .000

Value of autoconsumption of nutritional products

Important determining parameters are – compatible with the correlation analysis – surface of

landholding (β= 25.1%) and number of livestock (β= 14.3%). Unfortunately, also in this analysis

the coefficient for storage is negative (β= -9.8%); thus rejects the hypothesis that small reservoirs

support self sufficiency.

Table 7-14: Regression model for value of autoconsumption of nutritional products

R2= 15.8%

Page 96: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

74

Dependent variable Independent variables βi Sig.

Access to stocks (of cereals) until next harvest (+).101 .000

Value of expenditures on nutritional products -.041 .000

Surface of landholding +.251 .000

Number of livestock +.143 .000

Value of

autoconsumption of

nutritional products

Reservoir density -.098 .000

Household has access to stocks (of cereals) until next harvest

As the dependent variable is dichotomous the regression model is estimated by logistic

regression. Assessment of the χ2-statistic reveals that the model is not a good fit of the data (p=

.000); i.e. the hypothesis that the observed data are significantly different from the by the model

predicted values is confirmed. Also the variance explained by the model is low with Nagelkerke R2

.072 (7.2%). Hence, no genuine conclusions upon this indicator are possible.

Table 7-15: Regression model for access to stocks (of cereals) until next harvest

R2= 7.2%

Dependent variable Independent variables B Sig. Exp(B)

Value of expenditures on nutritional products .000 .348 1.000

Value of autoconsumption of nutritional products .000 .000 1.000

Surface of landholding +.007 .000 1.013

Number of livestock +.013 .000 1.007

Access to stocks (of

cereals) until next

harvest

Reservoir density +.280 .902 1323

Proximity of (local) food market

Access to (local) food markets is influenced by two external factors: population density and access

to (public) transport. The relative contribution of population density is 10.2%, and access to

(public) transport is the largest determining factor: β= 56.9%.

Table 7-16: Regression model for proximity of food market

R2= 36.7%

Dependent variable Independent variables βi Sig.

Proximity of (local) Proximity of (public) transport +.569 .000

Page 97: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

75

food market Population density +.102 .000

Number livestock

The only direct determining variable for number of livestock in the conceptual model is reservoir

density. This variable explains 2.8% of the variance of the dependent variable; the standardized

regression coefficient (β) is -.167. As the explaining power of this only predictor variable is low, a

more exploratory approach is applied though including ‘total revenue’ and ‘population density’ in

the equation. Herein the regression coefficient for ‘reservoir density’ becomes insignificant, but the

additional variables contribute to the explained variance; that increases to 4.6% (still too low to

draw any valid conclusions upon). The analysis shows that, as expected, population density has a

negative contribution (β= -19.5%) and income has a positive contribution (β= 11.5%).

Table 7-17: Regression model for livestock holding

R2= 4.6%

Dependent variable Independent variable βi Sig.

Reservoir density -.008 .745

Total revenue +.115 .000

Number of livestock

Population density -.195 .000

Surface of landholding

Again, the only determining variable for surface of landholding in the conceptual model is reservoir

density. This variable explains 9.5% of the variance of the dependent variable; the standardized

regression coefficient (β) is -.309. Again, a more exploratory approach is applied, including ‘total

revenue’ and ‘population density’ in the equation. The inclusion of these variables leads to an

explained variance of 10%, wherein the role of income is insignificant. Both remaining variables

relate negatively to the surface of landholding.

Table 7-18: Regression model for landholding

R2= 10.0%

Dependent variable Independent variable βi Sig.

Reservoir density -.187 .000

Total revenue -.016 .137

Surface of

landholding

Population density -.134 .000

Page 98: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

76

Prevalence of malnutrition

Only occurrence of food insecurity is hypothesized to have a direct relation to the prevalence of

malnutrition. Recall, the indicator malnutrition represents whether or not the child (younger than 5

years) is stunted, wasted or undernutritioned14. Since the dependent variable is a dichotomous dummy

variable, the model is estimated using binary logistic regression. The assessment of the χ2-statistic

reveals that the model is a good fit of the data (p= .086). However, the variance explained by the

model is negligible as Nagelkerke R2 is .007 (0.7%). Therefore, no conclusions are drawn upon this

model.

Table 7-19: Regression model for prevalence of malnutrition (direct model)

R2= 0.7%

Dependent variable Independent variable B Sig. Exp(B)

Prevalence of

malnutrition Occurrence of problems satisfying nutritional needs -.182 .000 .833

Although the correlation analysis does not show any relevant correlations with the prevalence of

malnutrition, a more exploratory regression analysis is applied including all indictors of the dimension

nutrition, storage, income, indicators of physical health, access to (improved) toilets and access to

potable water, plus external factors population density and access to (public) transport. This broader

analysis reveals that disease and disability are most (positive) influential to malnutrition (χ2-statistic

insignificant, R-square 4.2%).

Table 7-20: Regression model for prevalence of malnutrition (complete model)

R2= 4.2%

Dependent variable Independent variable B Sig. Exp(B)

Occurrence of problems satisfying nutritional needs -.177 .000 .837

Access to stocks (of cereals) until next harvest +.124 .255 1.132

Value of expenditures on nutritional products .000 .261 1.000

Value of autoconsumption of nutritional products .000 .166 1.000

Surface of landholding -.001 .527 .999

Prevalence of

malnutrition

Number of livestock -.003 .101 .997

14 See Glossary

Page 99: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

77

Proximity of (local) food market -.051 .215 1.052

Reservoir density +8.871 .396 7119

Total revenue .000 .030 1.000

Recent prevalence of disease +.766 .000 2.152

Chronic prevalence of handicap or injury +1.020 .029 2.772

Availability of (improved) potable water source -.027 .888 .974

Proximity of (potable) water source +.058 .311 .944

Access to (improved) toilets +.390 .003 .677

Population density .000 .983 1.000

Proximity of (public) transport +.074 .054 .928

7.2.4 Health

Theoretically, ‘recent prevalence of disease’ is a good overall indicator for the dimension health.

However, in the preceding correlation analysis we have seen that this indicator does not correlate high

to other indicators within the conceptual model, and that there are doubts about the reliability of the

indicator. Therefore, we also apply a complete analysis – including indirect relations and external

factors – for the indicator ‘value of expenditures on health services’.

Recent prevalence of disease

Besides the water-related diseases (with indicators fever and diarrhoea) this indicator encompasses

also eye, nose, ear, throat and skin problems. Chronic prevalence of disability, recent prevalence of

water-related disease, expenditures on health services and their proximity, and the prevalence of

malnutrition are – consistent with the conceptual model – directly influencing the prevalence of

diseases. Indirectly related factors are access sanitation facilities and access to potable water

sources, total income, storage, population density and access to (public) transport. Added to the

model the χ2-statistic reveals that the model is a good fit of the data (p= .727). Also the variance

explained by the model is satisfying as Nagelkerke R2 is .787 (78.7%). The analysis shows that

prevalence of malnutrition and chronic prevalence of disability are determining factors for disease.

Table 7-21: Regression model for recent prevalence of disease (complete model)

R2= 78.7%

Dependent variable Independent variable B Sig. Exp(B)

Recent prevalence of Chronic prevalence of handicap or injury -1.492 .047 .225

Page 100: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

78

Total value of expenditures on health

services .000 .001 1.000

Prevalence of malnutrition +.646 .033 1.907

Recent prevalence of fever -24.876 .990 .000

Recent prevalence of diarrhoea -24.742 .992 .000

Proximity of (public) health service +.071 .390 .931

Access to (improved) toilets -.090 .663 1.095

Access to (improved) garbage disposal +.068 .413 .934

Availability of (improved) potable water

source +.465 .152 1.592

Proximity of potable water source +.041 .746 .959

Total revenue .000 .725 1.000

Reservoir density +8.153 .666 3474

Population density +.002 .277 1.002

disease

Proximity of (public) transport -.023 .764 1.023

Chronic prevalence of handicap or injury

Chronic prevalence is – according to the conceptual model – determined by prevalence of water-

related diseases, expenditures and proximity of health services and prevalence of malnutrition. As

we fit a logistic regression model to these data the χ2-statistic reveals that the model is a good fit

of the data (p= .401). The variance explained by the model is 1.7% according to Nagelkerke R2.

Although this value is too low to really interpret, it can be noticed that the prevalence of

malnutrition is the only significant indicator in the model, with B is -5.582 (p= .016) and Exp(B)

3.025.

Table 7-22: Regression model for chronic prevalence of disability

R2= 1.7%

Dependent variable Independent variable B Sig. Exp(B)

Recent prevalence of disease -1.371 .065 .254

Recent prevalence of fever -24.836 .990 .000

Recent prevalence of diarrhoea -24.660 .992 .000

Total value of expenditures on health services .000 .000 1.000

Chronic prevalence of

handicap or injury

Proximity of (public) health service +.176 .003 .839

Page 101: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

79

Prevalence of malnutrition +.633 .029 1.883

Recent prevalence of fever

As the outcome variable is dichotomous, a logistic regression model is estimated. The χ2-statistic

reveals that the model is not a good fit of the data (p= .000). The variance explained by the model

is (Nagelkerke R2) 0.8%. Hence, no valid conclusions can be drawn upon the outcomes to the

regression analysis.

Table 7-23: Regression model for recent prevalence of fever

R2= 0.8%

Dependent variable Independent variable B Sig. Exp(B)

Total value of expenditures on health services .000 .000 1.000

Proximity of (public) health service +.025 .238 .975

Access to (improved) toilets +.114 .032 .892

Access to (improved) garbage disposal -.047 .029 1.049

Availability of (improved) potable water source +.066 .451 13068

Proximity of potable water source +.066 .083 .936

Recent prevalence of

fever

Reservoir density +.316 .904 1.372

Recent prevalence of diarrhoea

Likewise to fever, the prevalence of diarrhoea is determined by access to reservoirs and potable

water, access to sanitation facilities and health services, expenditures on health services. The χ2-

statistic reveals that the model is not a good fit of the data (p= .029). The variance explained by

the model is (Nagelkerke R2) 0.5%. Again, no valid conclusions can be drawn upon the outcomes

to the regression analysis.

Table 7-24: Regression model for recent prevalence of diarrhoea

R2= 0.5%

Dependent variable Independent variable B Sig. Exp(B)

Total value of expenditures on health services .000 .000 1.000

Proximity of (public) health service -.020 .592 1.020

Access to (improved) toilets +.077 .430 .926

Recent prevalence of

diarrhoea

Access to (improved) garbage disposal -.075 .050 1.077

Page 102: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

80

Availability of (improved) potable water source -.437 .004 .646

Proximity of potable water source -.012 .838 1.012

Reservoir density -2.924 .546 .054

Value of expenditures on health services

The value of expenditures on health services is hypothesized to be directly influenced by physical

health, the proximity of (public) health services and income. These parameters together explain 4.0%

of the variance. Indirectly, water-related diseases, improved sanitation and potable water, malnutrition

and food insecurity, storage and external factors population density and proximity of transport are of

influence. The explained variance increases to 5.5%, thus still is considered too low. Both variables

from the dimension nutrition (‘malnutrition’ and ‘food insecurity’) do not have significant regression

coefficients. The analysis shows that the three most important determining factors for expenditures on

health are total income (β=11.4%), chronic prevalence of disability (β=10%) and availability of

(improved) potable water source (β=7.2%). Hence, the expenditures on health care increase due to

these factors. Note that due to the low explained variance we should be careful drawing conclusions

upon the results.

Table 7-25: Regression model for value of expenditures on health services

R2= 5.5%

Dependent variable Independent variable βi Sig.

Recent prevalence of disease (+).100 .000

Chronic prevalence of handicap or injury (+).018 .000

Proximity of (public) health service +.032 .000

Total income +.115 .000

Recent prevalence of fever -.018 .003

Recent prevalence of diarrhoea -.020 .000

Access to (improved) toilets +.020 .000

Access to (improved) garbage disposal +.028 .000

Availability of (improved) potable water source +.072 .000

Proximity of potable water source -.040 .000

Occurrence of problems satisfying nutritional needs +.003 .445

Reservoir density +.024 .012

Total value of

expenditures on health

services

Population density +.052 .000

Page 103: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

81

Proximity of (public) transport +.007 .226

Access to (improved) toilets

In turn, access to improved sanitary facilities – i.e. improved toilets – is hypothesized to be

dependent on income and population density.

Table 7-26: Regression model for access to (improved) toilets

R2= 27.3%

Dependent variable Independent variable βi Sig.

Population density +.427 .000 Access to (improved)

toilets Total revenue +.232 .000

Access to (improved) garbage disposal

The only factors influencing garbage disposal are income and population density (within this

research). The explained variance is 8.0% with regression coefficients respectively β=4.8% and

β=27.1%. Clearly, income plays a minor role.

Table 7-27: Regression model for access to (improved) garbage disposal

R2= 8.0%

Dependent variable Independent variable βi Sig.

Population density +.271 .000 Access to (improved)

garbage disposal Total revenue +.048 .000

Proximity of potable water source

We are interested in the impact of storage and income, and external factors population density

and proximity of (public) transport on the proximity of potable water sources. The proximity of

(public) transport is the most influential variable (β= 30.8%), followed by population density (β=

13.2%). Both income as storage have a negative influence (respectively β= -2.7% and β= -7.4%).

Table 7-28: Regression model for proximity of potable water source

R2= 11.1%

Dependent variable Independent variable βi Sig.

Page 104: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

82

Reservoir density -.074 .001

Population density +.132 .000

Proximity of (public) transport +.308 .000

Proximity of potable

water source

Total revenue -.027 .011

Availability of (improved) potable water source

As we test the impact of storage, income and population density on the availability of improved

potable water sources we get the following results: the relative contribution of population density is

largest with 35.5%, followed by the positive contribution of income 20.5% and the negative

coefficient for storage (β= -5.3%).

Table 7-29: Regression model for availability of (improved) potable water source

R2= 16.1%

Dependent variable Independent variable βi Sig.

Reservoir density -.013 .546

Proximity of potable water source +.131 .000

Population density +.255 .000

Proximity of (public) transport +.164 .000

Availability of

(improved) potable

water source

Total revenue +.194 .000

Proximity of (public) health service

Within the scope of this research the proximity of (public) health services is determined by

population density and proximity of (public) health services. The total variance explained by these

variables is 46.4% and the relative contribution is respectively 24.7% and 56.9%.

Table 7-30: Regression model for proximity of health services

R2= 27.3%

Dependent variable Independent variable βi Sig.

Population density +.247 .000 Proximity of (public)

health service Proximity of (public) transport +.569 .000

Page 105: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

83

7.3 Discussion

• In general, the explained variance by the indicators included in the regression model is low to

medium. Exceptions are the complete models for storage, income and health, where R-square

is almost 80%. Hence, we should be aware that most of the dimension nutrition is not largely

explained by variables included in the model.

• Main assumptions underlying linear regression analysis are linearity, normality and

homoscedasticity. From the data screening process we have learned that none of the

indicators are normally distributed. However, Hair et al. (1998) states that regression analysis

has been shown to be quite robust even when the normality assumption is violated.

Assessment of the standardized residual plot shows that in many cases the assumption of

linearity is violated. Hence, the regression equation is not determined by a set of straight lines.

Due to the aforementioned violations heteroscedasticity is occurring as well.

• Additional assumptions are independence of error terms and normality of the error distribution.

The Durbin-Watson statistic informs about whether the assumption of independent errors is

tenable. Residual terms should be uncorrelated; i.e. independent. In most cases the

requirement of independent errors is satisfied. On the contrary, the normality of error

distribution is often violated

• One important assumption for unbiased multiple regression analysis is the requirement of no

perfect multicollinearity15. There should be no perfect linear relationship between predictors –

i.e. no high correlations. Multicollinearity can be diagnosed by assessing the correlation matrix

and the collinearity diagnostics. Fortunately, in most cases the assumption of no perfect

multicollinearity is satisfied. However, as reservoir density and population density are highly

correlated they are a source of collinearity. Possible solution is to exclude one out of two

variables from the analysis – here population density explains reservoir density at high level

(90%) thus can be excluded as explanatory variable – or summarize both variables in one

other variable

• Assessment of the goodness-of-fit indicates that all linear regression models are a good

representation of the population; hence, the models are generalizable. Note that this

conclusion is lessened by low values of R-square.

15 See Glossary

Page 106: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in
Page 107: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

85

8. Interpretation

In this chapter we provide the overall interpretation of the results from the correlation and regression

analyses. The aim is to get insight into the overall storage – poverty interactive system. As we have

mentioned before – discussion on correlation analysis – from the estimated correlation model on

country scale we learned that the variable ‘milieu of residence’ showed high correlations with many

other variables. This may implicate that the overall system – and driving forces – in a rural

environment shows to be different than when we regard the urban environment. This hypothesis can

only be tested by comparing the model for rural and urban sub-sets. Moreover, the research question

refers to rural households.

8.1 Interpretation of the country-scale analysis

Income

As hypothesized, the impact of small reservoirs on the total income – thus from all possible sources –

is positive. This total income is highly determined by its differentiated sources, wherein income from

(owned) resources has the largest stake. Access to small reservoirs seems to have a large positive

impact on education levels and employment rates (income from employment). This leads to the new

hypothesis that when small reservoirs are more proximate this leads to considerable time savings.

Obtaining water often involves significant inconvenience of time spent in collection. This reduces the

time remaining for other activities such as cooking or farming, being employed or going to school

(DOW, 2001).

Adversely, a negative relation of storage to income from (owned) resources – agriculture and dairy

farming – is found, while it was expected that this relation would be positive as well. It is found that

livestock holding has the highest contribution to income from (owned) resources. In turn, education

has a positive effect on all sources of income, with the exception of income from (owned) resources.

The proximity of schools leads to higher education levels, and consequently to higher income rates

(with the exception of income from agriculture and dairy farming). Remarkable is that the contribution

of secondary schools to education levels is much larger than that of primary schools.

Nutrition

Access to small reservoirs does not seem to have a positive – nor highly negative – impact on any

measure of the dimension nutrition. Food insecurity nor autoconsumption levels nor land and livestock

holding are positively related to reservoir density. Hence, there is no evidence found that small

Page 108: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

86

reservoirs have a positive direct impact on food supply from (irrigated) agriculture, dairy farming or

aquaculture – that are considered (throughout this research) as the main socio-economic values of

small reservoirs. Explanation may be that most small reservoirs are non-carryover; hence lacking to

secure water supply until the end of the dry season to sufficiently water livestock and serve basic

domestic needs or to provide in cash-crop opportunities.

Autoconsumption levels are highly determined by land and livestock holding (owned resources). This

supports the hypothesis that in poor rural societies food security is dependents on own production

(self suffiency). However, the indicator for food insecurity – occurrence of problems satisfying

nutritional needs – does not seem to benefit from autoconsumption nor from land and cattle holding,

since coefficients are low. Moreover, the importance of stocks (of cereals) and consumption of

marketed nutritional products occur to be most important. In turn, the total income is the determining

factor for expenditures on food; hence can be seen as driving force behind food security. We clearly

see there is a positive feedback loop from income to expenditures on nutritional products to food

insecurity (occurrence of problems satisfying nutritional needs) back to income. Unfortunately, the

analysis does not make clear what determines the availability of stocks.

Although access to owned land and livestock does not show to be an important factor in the alleviation

of food insecurity, it relates positively to both autoconsumption levels, as well as to income from

agriculture and dairy farming; consequently the total income. The role of local food markets in this

process is not clear; higher access to food markets leads to higher levels of expenditures on food, and

thus, contributes to food security, while when markets are at further distance values of

autoconsumption increase. However, income from (owned) resources does not seem to benefit from

more proximate markets.

The analysis reveals that disease and disability are most influential to malnutrition, while it does not

show malnutrition is related to measures of food security and food consumption levels. Explanation

may be that food access alone does not yield food security; food adequacy – quality besides quantity

– and physical ability to absorb nutrients (usually affected by disease) are determining factors (POST,

2006; Mwaniki, undated)

Health

The concern – and hypothesis – is that the presence of small reservoirs would cause higher

prevalence of water-related diseases. Fortunately, the analysis shows that this concern is

dispensable; there is no evidence found that reservoir density relates to gauges of water-related

diseases (fever and diarrhoea). However, better sanitation – improved latrines and garbage

evacuation – nor availability of improved potable water sources contribute to the reduction of water-

Page 109: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

87

related diseases. In fact, the prevalence of water-related diseases is not explained by factors within

the conceptual model.

Small reservoirs positively influence the availability of improved potable water sources – as they

possibly support groundwater raising (Savadogo, 2006) – and the option to have improved latrines.

However, compared to the impact of high population density the influence of storage on potable water

is low. Hence, driving force behind improved sources of potable water and, besides, behind improved

sanitation is population density.

Remarkable is that food consumption levels, access to (potable) water resources and improved

sanitation do not lead to a reduction of healthcare consumption. Moreover, no effects are found from

these factors onto prevalence of disease, disability or famine. Factors that importantly increase the

expenditures on health are income level, proximity of health services, and prevalence of disease and

disability; hence as people can access health services better, the use tends to increase.

Overall

Most likely population density is an important driving force behind the presence of small reservoirs,

while income is in the current analysis insignificant. Generally, the analysis shows that in more densely

populated areas resources (food market, potable water) and services (schools, health services)

become better accessible. Also access to (public) transport contributes significantly to the accessibility

of resources and services.

It can be concluded that the impact of small reservoirs on poverty is originating from income

generation and education – thus partly by time saving and improved human capital. Both food and

health security improve as total incomes increases, since more is spent on nutritional products and

health services. The extent of this relation is in turn influenced by measures of access to (local) food

markets and (public) health services, and consequently by access to (public) transport and population

density.

Question that comes out of this analysis is the relation that small reservoirs have with food security

and income from agriculture and dairy farming. As storage has a negative impact on autoconsumption

levels – opposite from the hypothesis – we may conclude that reservoir resources are not so much

used for own consumption, rather than to market products. However, as additionally land and livestock

holding, and income from these resources are found to have a negative relation to small reservoirs

this analysis does not provide evidence that established socio-economic values of small reservoirs are

properly provided, obtained or used.

Page 110: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

88

8.2 Urban versus rural environment

Possibly, driving forces differ between the rural and urban milieu of residence. It is hypothesized that

in a rural environment people mainly practice agriculture and dairy farming, therefore have higher

levels of autoconsumption and have more income from these resources. And so, storage possibly

plays a major role in sustaining their livelihoods, hence alleviate poverty levels. Where in the urban

environment expenditure levels are higher and income is not so much from agriculture and access to

services and assets is better.

Income

In the rural environment most income is obtained from agriculture and dairy farming. Besides other

sources of income also income from agriculture and dairy farming are positively related to the

presence of small reservoirs; the latter the most. However, land and livestock holding is not positively

influenced by the presence of small reservoirs.

In the urban environment most income is obtained from employment and entrepreneurship. The

overall contribution of small reservoirs on all sources of income is positive, however, with the

exception of income from owned resources. Income from employment relatively has the highest

contribution to the total income. Consequently, the role of education in employment is large – also

compared to the rural environment. Employment rates and education levels are more (positively)

influenced by food security, while physical health does not seem to have any relevance.

Nutrition

Neither in the rural nor in the urban environment access to small reservoirs contributes to the

alleviation of food insecurity, level of autoconsumption of nutritional products nor access to stocks (of

cereals). The conclusion that small reservoir do not have a positive direct impact on food attainment is

strengthened by this analysis.

In both environments access to stocks and expenditures on food are determining the state of food

(in)security, whereby the latter is mainly determined by income levels and market access. Basically,

the impact of income levels on nutrition (food insecurity and consumption) is higher in the urban

environment. Note the disputable role of autoconsumption; the regression analysis shows a negative

effect in both cases, while the correlation analysis shows a minor positive effect in the rural

environment. Likewise to the overall analysis, autoconsumption levels are highly determined by

access to owned resources (land holding more than livestock husbandry). The difference in

importance between the rural and urban environment herein is negligible. Only in the urban

Page 111: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

89

environment livestock and land holding are positively influenced by the presence of small reservoirs,

but negatively by higher population densities. In the rural environment income is the (positive)

determining factor for the ownership of these resources.

Health

A clear distinction between the rural and urban environment are the determining factors for the

prevalence of disease. While in the rural environment disease is mainly related to disability and

famine, in the urban environment access to sanitation facilities are more important. Unfortunately, the

prevalence of water-related diseases is not explained by the variables within the model.

In the urban environment, access to improved sanitation facilities and improved potable water are

most determined by income and population density. In the rural environment income plays a minor

role in access to improved sanitation facilities and improved potable water. There proximity of

(potable) water and transport are the most important factors.

Neither in the rural environment nor the urban environment we can draw unquestionable conclusion

upon the interaction between small reservoirs, potable water, sanitation facilities and the prevalence of

(water-related) diseases; no substantial correlations nor high explained variances are reached.

External factors

In both environments the presence of small reservoirs is highly explained by population density; in

neither environment income levels play a role in this. Remarkably is that – compared to the urban and

overall system – in the rural environment access to small reservoirs is less explained by both

parameters.

In the urban environment population density has higher impact on the access resources (food market,

potable water) and services (health centres, schools), including sanitation facilities. However, the role

of transport remains the main determining factor – and is even more important in the rural

environment.

Finally, two findings should be noted:

• In the urban environment total income increases with increased population density, however,

in the rural environment there is no relation between both.

• In the rural environment there is a less strong negative influence of population density to the

amount of livestock and land holding by a household.

Page 112: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in
Page 113: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

91

9. Reflection

This chapter does not only give the (technical) validation of the statistical test, additionally the

research as a whole is evaluated by discussing the validity of the scope. This evaluation concerns –

amongst others – the width of the perspective, the assumptions made and whether the approach used

is influencing the validity of outcomes.

9.1 Technical validation

In order to assess the accuracy of the statistical models we apply validation by data splitting. This

approach involves randomly splitting of the dataset (in half), re-computing the model and then

comparing the resulting models. If a model can be generalized, then it must be capable of accurately

predicting equal outcomes to different samples (Field, 2005). Note that the requirement of minimum

sample size should still be satisfied (see Chapter 7).

The split-sample validation of the correlation analysis shows that the absolute difference between

datasets is on average only 0.01 (1%), with a maximum of 0.1 (10%). These results show that the

accuracy of the tests is confirmed. As regression analysis is founded on correlation analysis there is

no need to apply additional split-sample validation of the correlation analysis.

Benchmarking the storage-indicator

In order to assess the accuracy and validity of the indicator ‘reservoir density (province scale)’ we

compare the outcomes of the correlation analysis for relations with storage for the identical relation

using the indicator ‘average distance to a reservoir’. The analysis shows that the absolute difference

between outcomes using different indicators is on average 0.02 (2%), with a maximum of 0.31 (31%).

Page 114: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

92

9.2 Evaluation of the scope

• Limitations of the indicators for storage

The indicator used to represent storage – socio-economic values of small reservoirs – reservoir

density knows two drawbacks:

As the indicator is measured at the province level this implicates that province borders act like

physical borders in the analysis, while in reality province borders do not function as physical

borders – e.g. people can easily cross the province border to obtain assets and resources.

Hence, the measurement scale of the indicator introduces an error;

The principal scale of aggregation for this research is – determined by the data source – the

household level. Hence, the reservoir density at province scale is attributed to all households

in a certain province. Again a spatial error is introduced, as the real proximity of reservoirs is

not equal for each household within the province.

• Limitations of the definition of poverty

As explained in Chapter 4, concepts of poverty that tend to go into the direction of the human

capability concept, since these fall outside of the scope of this research. Many experts on poverty

will debate to take into account – at least – socio-demographic variables as sex, age and

household headship. Cavendish (1999) argues that collection and use of environmental resources

is also strongly linked to the sex of the individual. Additionally, different (quantity and quality)

resources are used by different individuals and households at different ages. Further, resource

use can also be affected by household structure; an imperfect proxy here is to stratify the data by

household headship. Ommittance of inter-household poverty distribution assumes less complexity

of the poverty process, while in reality this is a significant factor in the succeeding of poverty

alleviating measures (e.g. small reservoirs) as not all household members benefit equally.

• Imperfect representation of concepts

It should be recognized that in some cases the indicator(s) used to represent a concept do not suit

perfectly. A good example is income from entrepreneurship is proxy is composed of revenue from

rent and interest, while it is preferred to include income that is obtained from small enterprises or

informal businesses. However, these measures are not available. Other examples are imperfect

measures of water-related diseases – as more direct measures are unavailable – and no

measures of production levels – as these are best to measure the benefits from small reservoirs.

Page 115: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

93

• Limited inclusion of spatial dynamics

The current research discards most cases of spatial dynamics as it analyses the relation between

water storage and poverty at country scale, and additionally between the rural and urban

environment. Reasons for this limited scope are:

Assessment of the outcomes of the analysis for four different climate zones did not give any

clear results; possibly due to inaccuracy of the definition of the zones;

Poverty mapping is not applied.

The aim of this research goes beyond community differences. As to assess local differences,

fieldwork trips are needed.

• Exclusion of temporal dynamics

The current research discards all temporal dynamics both between years as within the year.

Consequence is that poverty is less depicted as a process and that the influence of seasonality in

reservoir water supply is assumed not to exist. Moreover, there is a discrepancy in time between

the household survey of 2003 and the reservoir database of 2004.

• Limited inclusion of external factors

So far, the external factors included are population density, milieu of residence (rural versus

urban) and access to (public) transport. This limited number is mainly due to data availability

constraints. However, ideally factors derived from hydrology and climate, livelihood diversification,

and local demographics, economics and socio-politics should be included. It should be recognized

that both macro (including globalisation effects) and micro processes play a role. For example,

good infrastructure does not only make food markets better accessible – and so enable a

household to increase their income and food security – it also influences the prices of inputs and

outputs, improves levels and efficiency of use of inputs, and can even change the composition of

the labour market by creating opportunities for non-farm employment (Thimm, 1993).

Page 116: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in
Page 117: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

95

10. Conclusions and Recommendations

10.1 Conclusions

Although poverty is often considered a matter of institutions, governance and infrastructure, water

resources play a vital role in economic growth, human health and the reduction of poverty in the

savannah areas of West Africa. Therefore, throughout Burkina Faso – and other parts of West Africa –

many (small) dams and reservoirs have been built. They are an important source of water for many,

mostly poor, rural communities.

This research focuses on the question how the use of small multi-purpose reservoirs affects the well-

being of poor rural livelihoods. The aim is to give insight into the interdependences between the

presence – or absence – of small multi-purpose surface water reservoirs and the state of well-being of

rural households living nearby them. As we have knowledge of the extent and direction of the relations

between various dimensions of poverty and socio-economic values of small surface water reservoirs

their planning and management will be more sufficient. By answering the research question the driving

forces behind change and the role of small reservoirs in that is explained. But first we aim to answer

the sub-questions:

1. What are the socio-economic values of small multi-purpose surface water reservoirs to poor rural

households in Burkina Faso?

The reason for undertaking a valuation of small reservoirs is to assess their overall contribution to

social and economic well-being. Hereby, the term value is used to describe the importance placed

on the ecosystem by individuals, which includes not only income generation due to the use of its

goods and services, but also other benefits it provides for human welfare. Hence, not so much the

(monetary) economic value of water is regarded, moreover the economic characteristics; water as

a natural asset that is used by agriculture and households, and so provides a means for

livelihoods.

The concept of ‘total economic value’ (TEV) is a widely used framework for analysis of the

utilitarian value of ecosystems. We have combined this framework (derived from Barbier et al.,

1997) with the framework proposed by Turner et al. (2000), and designed a valuation framework

for the specific scope of this research: non-monetary valuation of small reservoirs. In this, the

socio-economic value of small reservoirs is divided into goods and services. Goods refer to the

Page 118: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

96

natural products harvested or used by communities, while services support life by indirect use or

existence – hence by functioning. Which and how many goods and services a reservoir can

provide depend on its characteristics: the physical features, natural environment and internal

processes.

Identified socio-economic values – goods and services – that small multi-purpose reservoirs can

provide to household living nearby them are water supply of domestic, agricultural and animal use,

raw material, food and nutrient supply, and other uses like recreation and education; Water from

small reservoirs is used in and around the house, e.g. for cleaning, bathing, washing, cooking.

Additionally, general agriculture and other agricultural purposes – such as fruit trees and

vegetable gardens – are served. Livestock may depend directly on water from small reservoirs, in

addition to profiting from the higher availability of fodder from crop stubble. The diversion of water

to home gardens may contribute substantially to a varied diet or increase the household income.

Easier access to water can also contribute the development of local economic activities, be it

small scale and informal such as brick making, beer brewing, and mat weaving. Generally, it is not

a source for drinking water; that is extracted from the groundwater; however, in areas where

rainfall is very low people may have no other choice than to use reservoir water (Boelee et al.,

2000).

Clearly, indicators of storage are water quality, water availability and proximity of small reservoirs.

Due to data restriction only the latter indicator is used in this research. Hence, the actual proxy for

the values of small reservoirs is represented by reservoir density at province scale.

2. How can we define ‘poverty’ within the scope of this research?

As a multidimensional phenomenon, poverty is defined and measured in various ways. The

formulation of the definition determines how we analyze poverty and understand its dimensions.

While the main understandings of the term include material and economic needs, increasingly, the

notion of what constitutes basic needs has expanded to encompass not only food, water shelter

and clothing, but also access to other assets such as education, credit, participation, security and

dignity (Hulme et al., 2001).

As to determine relevant dimensions of poverty for this research we analyzed literature on three

concepts of poverty: income (poverty lines) approach, basic needs approach and human capability

concept. The dimensions that fall under the poverty lines and basic human needs concept can be

seen as the most basic and more directly influenced by the presence – or absence – of small

reservoirs, and therefore, are selected to represent poverty. The analysis leads to a founded

Page 119: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

97

working definition of poverty: Poverty is lacking sufficient access to financial and material assets,

and public and natural resources, as to ensure being nutritioned and healthy.

As income, health and nutrition being the main dimensions that explain poverty, their sub-

dimensions are supporting them. These sub-dimensions are measures of access and availability,

measures on health, nutrition and income levels and expenditures on resources and assets.

3. Which dimensions of poverty are (in)directly related to the presence (or absence) of small multi-

purpose reservoirs? What is the statistical strength of these relations? Which external factors play

a significant role?

Statistical analysis - correlation analysis and multiple regression analysis - shows that the main

(positive) direct effect of small reservoirs is income generation and education. Mainly employment

rates and education levels benefit from storage. This leads to the new hypothesis that when small

reservoirs are more proximate this leads to considerable time savings.

Overall, food security benefits from the presence of small reservoirs. However, the direct relation

between the presence of small reservoirs and (sources of) nutrition is relevant, though not

unambiguous; the role of reservoirs differs between rural and urban environments. For example,

only in the urban environment livestock and land holding are positively influenced by the presence

of small reservoirs. In the rural environment income is the (positive) determining factor for the

ownership of these resources. Food insecurity is mainly alleviated by the presence of stocks (of

cereals) and consumption of marketed nutritional products. In turn, income is the determining

factor for expenditures on food; hence can be seen as driving force behind food security.

The analysis reveals that disease and disability are most influential to malnutrition, while it does

not show malnutrition is related to measures of food security and food consumption levels.

Explanation may be that food access alone does not yield food security; food adequacy – quality

besides quantity – and physical ability to absorb nutrients (usually affected by disease) are

additional determining factors.

Small reservoirs positively influence access to improved drinking water sources. No evidence is

found that the presence of small reservoirs relates to gauges of water-related diseases (fever and

diarrhoea). However, better sanitation – improved latrines and garbage evacuation – nor

availability of improved (potable) water sources contribute to the reduction of (water-related)

diseases.

Page 120: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

98

We can conclude population density is an important motivation for improved accessibility of water

storage (small reservoirs), and other resources (food market, potable water) and services

(schools, health services) become better accessible. Moreover, access to (public) transport

contributes significantly to the accessibility of resources and services.

What is the relationship between the presence of small multi-purpose surface water reservoirs and the

state of well-being of rural households?

We can conclude that the impact of small reservoirs on the poverty system is originating from income

generation – thus partly by time saving and improved human capital. Nearby reservoirs enable to

engage other activities as farming, going to school or developing small scale industries. Income is the

driving force behind food and health security, which – in turn – lead to improved human capital.

Population density is an important determining factor in access to small reservoirs. In more densely

populated areas resources (markets, water) and services (schools, health services, public transport)

become better accessible.

The applied statistical techniques puzzle out the existence and strength of relations, however, the

direction remains only given in by theory. In general, the strength of the relations and the explained

variance by the regression models is low to medium; hence, we should be careful drawing strong

conclusions. Relations are weak due to (1) measurement error and the use of proxies to represent

concepts, and (2) disaggregation of variables. Possible solutions are given below.

10.2 Recommendations for future work

• Improvement of the storage data

So far – in this research – the presence of small reservoirs is represented by the proximity of

those reservoirs. No (sufficient) data on water availability and water quality are available.

However, more powerful assessment of the relation between water storage and poverty can be

reached as estimates of reservoir volumes (in time) are included. The hypothesis would than be

that larger reservoirs have more impact on poverty alleviation. Furthermore, as to be able to

assess the impact of seasonality, water volumes over time – and at least at the beginning and end

of the rainy season – should be involved, using either satellite images or ground surveys.

Apart from extending the number of features covered, additionally, the current database should be

up-dated and attributed with the correct number and names of provinces and departments.

Page 121: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

99

• Improvement of the poverty data

It is recognized that poverty should be regarded as a process in which different dimensions – e.g.

income, nutrition and health – have mutual relations wherein the temporal antecedence of the

cause versus the effect is not always clear. The current research provides a rather static

representation of this poverty process; as multiple regression analysis does not enable iterative

testing. However, statistical techniques as structural equation modelling do allow to regard the

system as a whole and to include iterations. Unfortunately, the current data do not allow applying

this technique due to their measurement level and statistical quality; i.e. the data should be of

interval measurement level and satisfy the assumptions underlying many multi-variate techniques

(e.g. linearity, normality and homoscedasticity). Therefore it is recommended to upgrade the data

by re-formulating survey questionnaires as to obtain data of (at least) interval level16. Additionally,

re-formulating should lead to more reliable responds (reduce measurement error) and more

accurate indicators (to omit imperfect representation of concepts).

• Geo-referencing of the household survey

As mentioned in the reflection (Chapter 9), in the current research the indicator for the proximity of

small reservoirs – reservoir density – is at province scale. Therefore, an error is introduced in the

real proximity of small reservoirs; as the real proximity of reservoirs is not equal for each

household within the province. This error can be abolished in case the household survey would be

geo-referenced. In that case – by means of GIS tools – the real distance between the household

residence or community and small reservoirs can be assessed.

• Poverty mapping

The proposed geo-referencing of (future) household surveys additionally allows poverty mapping.

Poverty mapping is comparing the spatial distribution of poverty indicators with data from other

assessments, such as access to resources and services, visualized on maps. For the follow-up of

this research the contribution of poverty mapping is that regional differences can be shown – thus

spatial dynamics is introduced in the research – multiple dimensions can be displayed in one map

and the relation between cause and effect can become clearer from the visualisation by mapping.

It should be noted that, although poverty mapping can serve as a useful (exploratory) tool in

establishing and presenting the spatial relationship between indicators, it does not proof causal

relations between indicators; this should be assessed by the appropriate (statistical) analysis

techniques.

16 See Glossary

Page 122: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in
Page 123: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

101

Bibliography

Agudelo, J.I. (2001) “The economic valuation of water: Principles and methods”, Value of

water research report series No.5, IHE Delft, The Netherlands

Alcamo, J. et al. (2003) “Ecosystems and human well-being: A framework for assessment”,

Millennium Ecosystem Assessment, World Resources Institute,

Washington, USA

http://www.millenniumassessment.org/

Balazs, C. (2006) “Rural livelihoods and access to resources in relation to small

reservoirs: A study in Brazil’s Preto River Basin”, Masters Project,

University of California, Berkeley, USA

http://www.smallreservoirs.org/

Barbier, E.B. et al. (1997) “Economic valuation of wetlands, A guide for policy makers and

planners”, Ramsar convention bureau, Gland, Switzerland

http://www.ramsar.org/lib/lib_valuation_e.pdf

Beiersmann, C. et al. (2007) “Malaria in rural Burkina Faso: Local illness concepts, patterns of

traditional treatment and influence on health-seeking behaviour”,

Malaria Journal 2007 Volume 6 Article 106

http://www.malariajournal.com/content/6/1/106

Bloom, D.E. et al. (2000) “The health and wealth of nations”, Science 18 February 2000 Volume

287 No. 5456 pp 1207-1209

http://www.sciencemag.org/cgi/content/summary/287/5456/1207

Boelee, E. et al. (2000) “Multiple use of irrigation water in dry regions of Africa and South-

Asia”, Communication Texts, Volume 1, Session 1B-51-58,

International Conference Water and Health, Ouaga 2000, Health and

nutritional impacts of water development projects in Africa, November

21-24, 2000, Ouagadougou, Burkina Faso

Cavendish, W. (1999) “Empirical regularities in the poverty-environment relationship of

African rural households”, TH Huxley School, Imperial College,

London, UK

http://www.csae.ox.ac.uk/workingpapers/pdfs/9921text.pdf

Chen, C.K. et al. (2004) “Using ordinal regression model to analyse student satisfaction

questionnaires”, Association for Institutional Research, IR applications

Volume 1

Page 124: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

102

Coche, A.G. (1998) “Supporting aquaculture development in Africa”, Research network on

Integration of Agriculture and Irrigation, CIFA Occasional Paper No.

23, Rome, Italy

http://www.fao.org/docrep/X5598E/X5598E00.htm

Coudouel, A. et al. (2002) “Chapter 1: Poverty Measurement and Analysis”, in A sourcebook for

Poverty Reduction Strategies: Part 1 Core techniques and cross-

cutting issues, World Bank, Washington DC, USA

http://povlibrary.worldbank.org/files/5467_chap1.pdf

De Groot, R.S. (1992) “Functions of nature: Evaluation of nature in environmental planning,

management and decision making”, Wolters-Noordhoff, Groningen,

The Netherlands

Dinar, A. et al. (1995) “Restoring and protecting the world’s lakes and reservoirs”, World

Bank technical paper No. 289, World Bank, Washington, USA

DGIRH (2004) Official list of reservoirs in Burkina Faso February 2004, Ministère de

l’Envinronnement et de l‘Eau, Direction Générale de l’Inventaire des

Ressources Hydrauliques, Ougadougou, Burkina Faso.

DOW (2001) “Rural water demand: The case of Eastern Africa”, Lessons from the

Drawers of Water II Study

http://webworld.unesco.org/water/wwap/pccp/cd/pdf/educational_tools

/course_modules/reference_documents/water/ruralwaterdemand.pdf

Falkingham, J. et al. (2002) “Measuring health and poverty: A review of approaches to identifying

the poor”, DFID Health Systems Resource Centre, London, UK

FAO (2003) “Measurement and Assessment of Food Deprivation and

Undernutrition”, Proceedings of the International Scientific

Symposium, 26-28 June 2002, Rome, Italy

http://www.fao.org/DOCREP/005/Y4249E/y4249e00.htm

FAO (2006) “What’s water worth?” Agriculture 21 Magazine March 2006, FAO,

Rome, Italy

http://www.fao.org/ag/magazine/0603sp1.htm

Field, A. (2005) “Discovering statistics using SPSS”, Second edition, Sage

Publications, London, UK

Haddad, L. (2002) “Nutrition and Poverty”, in ACC/CSN (2002), Nutrition: A foundation for

development, Geneva, Switzerland

http://www.ifpri.org/pubs/books/intnut/intnut.pdf

Hair, J.F. et al. (1998) “Multivariate data analysis”, Fifth edition, Prentice Hall, New Jersey,

USA

Page 125: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

103

Ho, R. (1996) “Handbook of univariate and multivariate data analysis and

interpretation with SPSS”, Central Queensland University, Rock

Hampton, Australia

Hulme, D. et al. (2001) “Chronic poverty: Meanings and analytical framework”, CPRC

Working Paper 2, Chronic Poverty Research Centre, Manchester, UK

http://www.chronicpoverty.org/resources/working_papers.html

INSD (2003) "Enquête Burkinabè sur les conditions de vie des ménages: Première

phase", Manuel de l’agent enquêteur, Institut national de la statistique

et de la démographie, Ouagadougou, Burkina Faso

INSD (2004) “Projections de population du Burkina Faso (2004)”, Ministère du

l’Economie et du Développement, SecrétariatGénéral et Institut

National de la Statistique et de la Démographie (INSD),

Ouagadougou, Burkina Faso

http://www.insd.bf/actualites/Publications/f_Projections_de_population

.pdf

Keller, A. et al. (2000) “Water scarcity and the role of storage in development”, Research

Report 39, IWMI, Colombo, Sri Lanka

Kemper, K. et al. (undated) “The global water challenge”, World Bank global issues seminar series

http://siteresources.worldbank.org/EXTABOUTUS/Resources/WaterP

aper.pdf#search=%22The%20global%20water%20challenge%20kem

per%22

Liebe, J. (2002) “Estimation of water storage capacity and evaporation losses of small

reservoirs in the upper east region of Ghana”, Diploma thesis,

Geographische Institute der Rheinischen Friedrich-Wilhelms-

Universität, Bonn, Germany

http://www.smallreservoirs.org/

Lipton, M. et al. (2003) “Preliminary review of the impact of irrigation on poverty: with special

emphasis on Asia”, Land and Water Development Division, FAO,

Rome, Italy

Lok-Dessallien, R. (undated a) “Review of poverty concepts and indicators”, SEPED series on

poverty reduction

http://www.undp.org/poverty/publications/pov_red/

Lok-Dessallien, R. (undated b) “Poverty profile: Interpreting the data”, SEPED series on poverty

reduction

http://www.undp.org/poverty/publications/pov_red/

Lok-Dessallien, R. (undated c) “The data: Where to find them”, SEPED series on poverty reduction

http://www.undp.org/poverty/publications/pov_red/

MARA/ARMA (1998) “Towards an atlas of malaria risk in Africa”, Durban, South Africa

Page 126: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

104

Molden, D. (2007) “Water for food Water for life: A comprehensive assessment of water

management in agriculture”, Earthscan, London, UK and International

Water Management Institute, Colombo, Sri Lanka

Moriarty, P. et al. (2003) “The productive use of domestic water supplies: How water supplies

can play a wider role in livelihood improvement and poverty

reduction”, Thematic overview paper, IRC International Water and

Sanitation Centre, Delft, The Netherlands

Moriarty, P. et al. (2004) “Beyond domestic: Case studies on poverty and productive uses of

water at the household level”, Technical paper series No 41, IRC

International Water and Sanitation Centre, Delft, The Netherlands

Mwaniki, A. (undated) “Achieving food security in Africa: Challenges and issues”, Cornell

University, Ithaca, USA

http://www.un.org/africa/osaa/reports/Achieving%20Food%20Security

%20in%20Africa-Challenges%20and%20Issues.pdf

Nandy, S. et al. (2003) “Poverty, food and health in welfare: Current issues, future

perspectives”, International Conference on Poverty, Food and Health

in Welfare, Lisbon, Portugal

Newcome, J. et al. (2005) “The economic, social and environmental value of ecosystem

services: A literature review”, Final report for the Department for

Environment, Food and Rural Affairs, Eftec, London, UK

OECD/WHO (1993) “Poverty and Health”, DAC Guidelines and Reference Series,

Organisation for economic co-operation and development in

cooperation with World Health Organisation, Paris, France

http://whqlibdoc.who.int/publications/2003/9241562366.pdf

Pallant, J. (2001) “SPSS survival manual: A step by step guide to data analysis using

SPSS”, Open University Press. Chicago, USA

Pearce, D.W. et al. (1993) “World Without End”, Oxford University Press, Oxford, USA

Poolman, M. (2005) “Developing small reservoirs: A participatory approach can help“,

Masters Thesis, Delft University of Technology, Delft, The Netherlands

http://www.smallreservoirs.org/

POST (2006) “Food security in developing countries”, Postnote December 2006, No.

274, Parliamentary Office of Science and Technology, London, UK

http://www.imf.org/external/pubs/ft/scr/2005/cr05338.pdf

Pritchett, L. (1997) “Review of Robert D. Kaplan’s The Ends of the Earth”, Finance and

Development March 1997, International Monetary Fund and the

International Bank for Reconstruction and Development, Washington

DC, USA

http://worldbank.org/fandd/english/0397/mar97.htm

Page 127: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

105

Rocha, S. (1998) “On statistical mapping of poverty social reality: Concepts and

measurement”, Texto para discussao No. 553, Rio de Janeiro,

Argentina

Roggeri, H. (1995) “Tropical freshwater wetlands: A guide to current knowledge and

sustainable management”, Kluwer Academic, Dordrecht, The

Netherlands

Savadogo, A.S. (2006) “Water resources management in Burkina Faso: A case study on the

potential of small dams”, WaterAid, Ouagadougou, Burkina Faso

http://www.wateraid.org/documents/plugin_documents/burkina_faso_fi

eldwork_report__cwrm.pdf

SRP (undated) “Project proposal”, Small Reservoirs Project

http://www.smallreservoirs.org/

Smith, R.D. et al. (1995) “An approach for assessing wetland functions using hydrogeomorphic

classification, reference wetlands, and functional indices”, Wetlands

research program technical report WRP-DE-9, US Army Corps of

Engineers, Washington, USA

Thimm, H.U. (2003) “Interdisciplinary evaluation of the role of infrastructure”, Proceedings

for the International Symposium on Regional Food Security and Rural

Infrastructure, 3-6 May 2003, Giessen-Rauischholzhausen, Germany

Turner, R.K. et al. (2000) “Ecological-economic analysis of wetlands: Scientific integration for

management and policy”, Ecological Economics Volume 35, special

issue, Elsevier

Turner, R.K. et al. (2004) “Economic valuation of water resources in agriculture: From the

sectoral to a functional perspective of natural resources

management”, FAO water reports No. 27, FAO, Rome, Italy

http://www.fao.org/docrep/007/y5582e/y5582e00.HTM

UNDP (2003a) “Human Development Report 2003: Human development indicators”

http://hdr.undp.org/reports/global/2003/pdf/hdr03_HDI.pdf

UNDP (2003b) “Human Development Report 2003: Technical note 1”,

http://hdr.undp.org/reports/global/2003/pdf/hdr03_backmatter_2.pdf

UNDP (2005) “Human Development Reports: Burkina Faso Country Sheet”

http://hdr.undp.org/statistics/data/countries.cfm?c=BFA

UNDP (1997) “Human Development Report 1997”, Oxford University Press, New

York, USA

http://hdr.undp.org/reports/global/1997/en/

Van de Giesen, N.C. et al. (2000)

“The Glowa Volta project: Integrated assessment of feedback

mechanisms between climate, land-use and hydrology”, Wengen-

Page 128: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

106

2000 Workshop, Climatic Change: Implications for the Hydrological

Cycle and for Water Management, Wengen, Switzerland

http://www.glowa-volta.de/publications/printed/wengen2000.pdf

Website FAO SPFS (February 2007)

http://www.fao.org/spfs/

Website IFAD (November 2006)

http://www.ifad.org/sf/

Website MARA/ARMA (September 2007)

http://www.mara.org.za/

Website PovertyNet (June, 2007)

http://www.worldbank.org/poverty/

Website Unicef (September 2007)

http://www.unicef.org/wes/

Website WHO (August 2007)

http://www.who.int/water_sanitation_health/diseases/en/index.html

Website World Bank (November 2006)

http://web.worldbank.org/WBSITE/EXTERNAL/EXTABOUTUS/0,,cont

entMDK:20040565~menuPK:1696892~pagePK:51123644~piPK:3298

29~theSitePK:29708,00.html

Website World Concern (June 2007)

http://www.worldconcern.org/NETCOMMUNITY/Page.aspx?&pid=567

&srcid=414

WHO (2007) “Combating waterborne diseases at the household level”, The

international network to promote household water treatment and safe

storage, Geneva, Switzerland

http://www.who.int/water_sanitation_health/diseases/burden/en/index.

html

World Bank (2004) “Poverty monitoring guidance note 1: Selecting indicators”,

Washington, USA

http://poverty2.forumone.com/library/view/15138

World Bank (2005) “Data and statistics on Burkina Faso”

http://web.worldbank.org/WBSITE/EXTERNAL/COUNTRIES/AFRICA

EXT/BURKINAFASOEXTN/0,,menuPK:343902~pagePK:141132~piP

K:141109~theSitePK:343876,00.html

WSSCC (2005) “Sanitation and hygiene promotion: Programming guidance”, Water

supply and sanitation collaborative council and World health

organization, Geneva, Switzerland

Page 129: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in
Page 130: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in
Page 131: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

109

Appendix A. Values of Water

Values and classifications found in literature

The following list gives a good impression of the values of water ecosystems found in literature. One

can see that different authors use different conceptual frameworks – e.g. Turner et al. (2000 & 2004),

Barbier et al. (1997) and Newcome (2005) versus De Groot (1992) versus Alcamo (2003) versus

Roggeri (1995) – and/or give different meanings to terms as goods, services, use values and non-use

values, functions, etc. Note that this list is not an exhaustive review on literature, many more have

published on this subject. In addition, it must be realised that there are probably many unknown values

that are not recognised yet, but which may have considerable (potential) benefits (De Groot, 1992).

The classification used by Turner et al. (2000 & 2004), Barbier et al. (1997) and Newcome (2005) is

based on the framework for Total Economic Value (TEV). It provides a framework for grouping values,

and there is an increasing consensus that it is the most appropriate one to use. Simply put, total

economic valuation distinguishes between use values and non-use values, the latter referring to those

current or future (potential) values associated with an environmental resource which rely merely on its

continued existence and are unrelated to use (Pearce et al., 1993). Typically, use values involve some

human interaction with the resource that is grouped according to whether they are direct or indirect.

Direct use values involve both commercial and non-commercial activities. The indirect use values

derive from supporting or protecting activities. A special category of use values is (quasi)option value,

which is uncertainty on its future use and information function.

Table A-1: Values of water combined from Barbier et al., 1997 and Newcome et al., 2005

Direct use values Indirect use values Option and quasi-option Non-use values

livestock and

cultivation

fisheries

agriculture

fibre and fuelwood

recreation

transport

wildlife harvesting

and hunting

peat/energy

aesthetic value

sediment and

nutrient cycling

flood water storage

and stream flow

regulation

groundwater

recharge

external ecosystem

support

• micro-climate

stabilisation

potential future direct

and indirect uses of

goods and services

future value of

information

biodiversity

bequest values

(value on the

conservation of

wetlands for future

generations)

existence

cultural knowledge

and traditions

Page 132: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

110

Turner et al. (2000) goes one step further, by explicitly linking economic valuation to ecological

characterisation. This, he labels as going from wetland functioning to wetland uses. Economic values

will always relay upon the wetland performing functions that are somehow perceived as valuable by

society. Functions in themselves are therefore not necessarily of economic value – such value derives

from the existence of a demand for wetland goods and wetland services due to these functions.

Table A-2: Values of water derived from Turner et al., 2000 & 2004

Goods Services Use values Non-use values Other values

agriculture

fisheries

forestry

non-timber

forest

products

water supply

recreation

flood control

groundwater

recharge

nutrient

removal

toxics

retention

biodiversity

maintenance

consumptive

recreational

aesthetical

educational

indirect use

values

existence

bequest

philanthropic

option

quasi option

The classification used by De Groot (1992) is based on environmental function evaluation, and thus

includes not only the land-use values and harvestable goods (nature in the narrow sense), but also

refers to other benefits of the natural environment which are less tangible. Environmental functions are

defined as “the capacity of natural processes and components to provide goods and services that –

directly of indirectly – satisfy physiological and psychological human needs”. De Groot (1992)divides

the environmental functions in to four classes:

• Regulation function relates to the capacity of natural and semi-natural ecosystems to regulate

essential ecological processes and life support systems, which contributes to the maintenance of

a healthy environment by providing clean air, water and soil.

• Information function, natural ecosystems contribute to the maintenance of mental health by

providing opportunities for reflection, spiritual enrichment, cognitive development and aesthetical

experience.

• Carrier function implies that natural and semi-natural ecosystems provide space and a suitable

substrate or medium for many human activities such as habitation, cultivation and recreation.

• Production function, nature provides many resources, ranging from food and raw materials to

energy resources and genetic material.

Page 133: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

111

Table A-3: Values of water derived from De Groot, 1992

Due to regulation

function

Due to information

function

Due to carrier

function

Due to production

function

storage and

recycling of

nutrients, human

and organic waste

groundwater re- &

discharge

flood control & flow

regulation

erosion control

salinity control

water treatment

climatic

stabilization

maintenance of

biodiversity

education and

monitoring

cultural heritage

agriculture

stock farming

(grazing)

wildlife cropping

energy

production

transport

tourism and

recreation

human

habitation and

settlement

water

food

fuel wood

medicinal resources

raw materials for

building and

industrial use

genetic resources

The framework proposed by Alcamo et al. (2003) as a first product of the Millennium Ecosystem

Assessment (MA) – a four-year international work program designed to meet the needs of decision-

makers for scientific information on the links between ecosystem change and human well-being –

places human well-being as the central focus for assessment, while recognizing that biodiversity and

ecosystems also have intrinsic value and that people take decisions concerning ecosystems based on

considerations of well-being as well as intrinsic value. Ecosystem services are the benefits people

obtain from ecosystems. These include provisioning, supporting, regulating, and cultural services,

which directly affect people, and supporting services needed to maintain the other services. Changes

in these services affect human well-being through impacts on security, the basic material for a good

life, health, and social and cultural relations. These constituents of well-being are, in turn, influenced

by and have an influence on the freedoms and choices available to people.

Page 134: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

112

Table A-4: Values of water derived from Alcamo et al., 2003

Supporting services Provisioning services Regulation services Cultural services

nutrient cycling

soil formation

primary production

food

fresh water

wood and fibre

fuel

climate regulation

flood regulation

water purification

aesthetic

spiritual

educational

recreational

The classification of Roggeri (1995), designed for tropical freshwater wetlands. Functions are due to

their role in many natural phenomena and processes. Their resources can be used in order to obtain

products or services. Finally, they have attributes such as biological diversity. These functions,

attributes (or qualities) and resources are goods and services which have a value for human beings.

Note that they are closely linked on the one hand to the wetlands’ biological, chemical and physical

characteristics, and on the other hand to the interaction of these characteristics. Therefore, wetlands

do not automatically provide all the goods and services as mentioned below. Furthermore, the role

that wetlands can play in a given process may vary considerably, both in significance and quality.

Table A-5: Values of water derived from Roggeri, 1995

Resources Functions Attributes

agriculture, forestry, forage

production

wildlife or fish production

aquaculture

natural products

water supply

energy production

transport

tourism, recreation

research and education

nutrient retention, export

groundwater dis/recharge

flood mitigation

sediment retention

erosion control

salinity control

water treatment

climate stabilization

ecosystem stability

biological diversity

cultural or historic value

aesthetic value

Page 135: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

113

Appendix B. Selection of Indicators

B.1 Selection of indicators for poverty

Data source available on poverty indicators is the ‘Questionnaire des Indicateurs de Base de Bien-

être’ (QUIBB)17 performed between April and July of 2003. This questionnaire is written on behalf of

the National Institute of Statistics and Demography and Demography (INSD) of Burkina Faso, and is

meant for gathering the data needed for the economical and social management of the country. It is –

in design – a known way for collecting information about household characteristics, measures of

access, usage and degree of satisfaction en matters of social service. The questionnaire uses a

methodology developed by a group of donators and institutions, e.g. World Bank, BIT, UNICEF, and

PNUD (INSD, 2003).

The questionnaire divides the country of Burkina Faso in 425 zones (representing communities), in

each of which 24 households are interviewed. The questionnaire is composed of the following

sections:

Section A Administration of the interview place, time and date

Section B Information about the informant

Section C Education

Section D Health

Section E Employment

Section F Assets of the household

Section G Resources to the household

Section I Information on children younger than 5 years

Section J Expenses on education, health and agriculture

Section K Information on agricultural activities

Section L Information on entrepreneurial activities

Section M Information on nutritional products consumed

Section N Information on non-food products consumed

Section O Information on source of income

Section P Information on different types of services

17 Translated: Questionnaire of indicators of basic well-being

Page 136: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

114

Sections B till E, I and L concern information on individual level, and sections F, G, J, K and M till P

information on household level. Section I is information on individual level concerning the children

younger than five years. Note that some available variables do not belong to any section mentioned

above. Also, the included manuals for interviewers and controllers do not mention these ‘other’

indicators neither do they mention section J till P.

The tables below show the selected indicators on the main dimensions of poverty – health, nutrition

and income – that are selected from the broad list of variables available from the QUIBB

questionnaires. They are selected as to be the most feasible proxies for the (sub)dimensions of

poverty. The tables show – ordered per dimension – the name of the indicator, description of the

indicator, the measurement level and scale, and the item scale. This last characteristic of the data is

relevant since specific statistical procedures appoint requirements upon the minimum level of

measurement. Basically, four levels of measurement are commonly used in statistics as to describe

the nature of the information contained within numbers assigned to objects and, therefore, within the

variable:

• Nominal – the nominal measurement level is considered the lowest. It assigns numerical

values as labels to identify categorical data.

• Ordinal – in case of ordinal scales categorical data can be ordered or ranked in relation to the

amount of the attribute possessed. However, the scale is really non-quantitative, because it

indicates only relative positions in an ordered series.

• Interval – represents quantitative data with a constant unit of measurement, that have an

arbitrary zero point. Therefore, it is not possible to state that any value on an interval scale is a

multiplication of any other value on the scale.

• Ratio – represents the highest form of measurement precision because they possess the

advantages of all lower scales plus an absolute zero point (Hair et al., 1998).

Page 137: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

115

Table B-1: Selected indicators for income

Name Indicator Measurement

scale

Measurement

level Item scale

Revagric Revenue from agriculture Household Ratio CFA/yr

Reveleva Revenue from dairy farming Household Ratio CFA/yr

Salpubli Salary form public sector Household Ratio CFA/yr

Salprive Salary form private sector Household Ratio CFA/yr

Revloyer Revenue form rent Household Ratio CFA/yr

Transf Revenue from interest Household Ratio CFA/yr

Revtotal Total revenue Household Ratio CFA/yr

Revownresources1 Income from (owned) resources

(∑ revagric + reveleva) Household Ratio CFA/yr

Revemployment1 Income from employment

(∑ salpubli + salprive) Household Ratio CFA/yr

Reventrepreneur1 Income from entrepreneurship

(∑ revloyer + tranf) Household Ratio CFA/yr

Niveduc Highest level of education reached Individual Ordinal2

1: Not at all

2: Primary

3: Secondary

4: Higher

G7d Proximity of primary school Household Ordinal2

1: > 1 hour

2: 45-59 min.

3: 30-44 min.

4: 15-29 min.

5: 0-14 min.

G7e Proximity of secondary school Household Ordinal2

1: > 1 hour

2: 45-59 min.

3: 30-44 min.

4: 15-29 min.

5: 0-14 min. 1 Composed aggregate measure of variables available from INSD (2003). 2 Variable beholds an underlying continuum and therefore is treated as of interval level in the analysis.

Page 138: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

116

Table B-2: Selected indicators for nutrition

Name Indicator Measurement

scale

Measurement

level Item scale

F3 Surface of landholding Household Ratio # hectares

F8 Number of large cattle owned Household Ratio # cattle

F10 Number of small cattle owned Household Ratio # cattle

Cattle1 Number of livestock

(Σ F8 + F10) Household Ratio # cattle

F14 Occurrence of problems satisfying

nutritional needs Household Ordinal2

1: Never

2: Rarely

3: Sometimes

4: Often

5: Always

F20 Access to stocks (of cereals) until the

next harvest Household Ordinal

0: No

1: Yes

G7b Proximity of (local) food market Household Ordinal2

1: > 1 hour

2: 45-59 min.

3: 30-44 min.

4: 15-29 min.

5: 0-14 min.

Autoali Value of autoconsumption of nutritional

products Household Ratio CFA/month

Achali Value of expenditures on nutritional

products Household Ratio CFA/month

Depali Total value of consumption of nutritional

products Household Ratio CFA/month

Wasted Child has low weight for height Infant Dichotomous 0: No

1: Yes

Stunted Child has low height for age Infant Dichotomous 0: No

1: Yes

Underweight Child has low weight for age Infant Dichotomous 0: No

1: Yes

Malnutrition1 Child is wasted and/or stunted and/or

underweight Infant Dichotomous

0: No

1: Yes

Length Length of the child Infant Ratio # cm

Weight Weight of the child Infant Ratio # kg

1 Composed aggregate measure of variables available from INSD (2003). 2 Variable beholds an underlying continuum and therefore is treated as of interval level in the analysis.

Page 139: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

117

Table B-3: Selected indicators for health

Name Indicator Measurement

scale

Measurement

level Item scale

D3 Chronic prevalence of handicap or injury Individual Dichotomous 0: No

1: Yes

D4 Recent prevalence of disease Individual Dichotomous 0: No

1: Yes

D5a Recent prevalence of fever Individual Dichotomous 0: No

1: Yes

D5b Recent prevalence of diarrhoea Individual Dichotomous 0: No

1: Yes

G3 Access to (improved) potable water

source Household Ordinal2

1: River or lake

2: (Drilled) well

3: Inside tap

G4 Access to (improved) toilets Household Ordinal2

1: In nature

2: Ordinary

latrine

3: Latrine with

ventilated put

4: Flush with

septic put

G7a Proximity of potable water source Household Ordinal2

1: > 1 hour

2: 45-59 min.

3: 30-44 min.

4: 15-29 min.

5: 0-14 min.

G7f Proximity of (public) health service

(hospital or clinic) Household Ordinal2

1: > 1 hour

2: 45-59 min.

3: 30-44 min.

4: 15-29 min.

5: 0-14 min.

G11 Access to (improved) garbage disposal Household Ordinal2

1: Street

2: Bag

3: Put

4: Individual

garbage pile

5: Public

garbage pile

6: Garbage bin

(emptied by a

service)

Page 140: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

118

J31 Value of expenditures on consultation Household Ratio CFA/past

month

J32 Value of expenditures on medical

analysis Household Ratio

CFA/past

month

J33 Value of expenditures on medicines Household Ratio CFA/past

month

J34 Value of expenditures on hospitals Household Ratio CFA/past

month

J35 Value of expenditures on other medical

services Household Ratio

CFA/past

month

Tothealth1 Value of expenditures on health services

(∑ J31 + J32 + J33 + J34 + J35) Household Ratio

CFA/past

month 1 Composed aggregate measure of variables available from INSD (2003). 2 Variable beholds an underlying continuum and therefore is treated as of interval level in the analysis.

B.2 Selection of external indicators

Source: INSD (2003), Questionnaire des Indicateurs de Base de Bien-être, National Institute of

Statistics and Demography and Demography (INSD), Ougadougou, Burkina Faso

Table B-4: Selected indicators for external factors

Name Indicator Measurement

scale

Measurement

level Item scale

Urbrur Milieu of residence Household Dichotomous 0: Urban

1: Rural

Hhsize Number of household members Household Ratio # people

B5 Age (of last birthday) Individual Ratio # of years

G7c Proximity of (public) transport Household Ordinal2

1: > 1 hour

2: 45-59 min.

3: 30-44 min.

4: 15-29 min.

5: 0-14 min. 1 Composed aggregate measure of variables available from INSD (2003). 2 Variable beholds an underlying continuum and therefore is treated as of interval level in the analysis.

Page 141: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

119

Source:

A. INSD (2004), Projections de population du Burkina Faso, Ministère du l’Economie et du

Développement, Secrétariat général, et Institut National de la Statistique et de la

Démographie (INSD), Ouagadougou, Burkina Faso.

B. DGIRH (2004), Official list of reservoirs in Burkina Faso February 2004, Ministère de

l’Envinronnement et de l‘Eau, Direction Générale de l’Inventaire des Ressources

Hydrauliques, Ougadougou, Burkina Faso.

Table B-5: Population and reservoir density

No. Province PopulationA

[#]

AreaB

[km2]

Population

density A [#/km2]

ReservoirsB

[#]

Reservoir

density B [#/100*km2]

31 Balé 198765 4543 43.75 13 0.286

1 Bam 250144 4073 61.42 66 1.621

32 Banwa 261140 5825 44.83 2 0.034

2 Bazéga 236316 3874 61.01 73 1.885

3 Bougouriba 84775 3420 24.79 8 0.234

4 Boulgou 478576 6371 75.12 39 0.612

5 Boulkiemde 461393 4254 108.46 102 2.398

6 Comoé 292479 15305 19.11 28 0.183

7 Ganzourgou 305556 4131 73.96 43 1.041

8 Gnagna 370533 8585 43.16 31 0.361

9 Gourma 255906 11212 22.82 35 0.312

10 Houet 818471 11630 70.38 21 0.181

33 Ioba 176641 2634 67.05 18 0.683

11 Kadiogo 1204346 2930 411.00 94 3.208

12 Kénédougou 245618 8336 29.47 12 0.144

34 Komandjari 57965 5175 11.20 2 0.039

35 Kompienga 57388 7048 8.14 1 0.014

13 Kossi 268002 7490 35.78 0 0.000

36 Koulpélogo 221476 5395 41.05 19 0.352

14 Kouritenga 289763 2802 103.41 58 2.070

Page 142: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

120

37 Kourwégo 135227 1657 81.63 21 1.268

38 Léraba 106632 3076 34.67 10 0.325

39 Loroum 134457 3656 36.78 32 0.875

15 Mouhoun 284220 6898 41.20 13 0.188

16 Nahouri 129778 3814 34.03 46 1.206

17 Namentenga 293466 6320 46.44 15 0.237

40 Nayala 150417 3698 40.68 5 0.135

41 Noumbiel 56452 2709 20.84 2 0.074

18 Oubritenga 237096 2790 84.97 57 2.043

19 Oudalan 161415 9957 16.21 83 0.834

20 Passoré 307668 3940 78.08 55 1.396

21 Poni 198585 7379 26.91 14 0.190

22 Sanguié 272599 5144 52.99 80 1.555

23 Sanmatenga 538068 9406 57.20 50 0.532

24 Séno 242846 6984 34.77 21 0.301

25 Sissili 183107 7105 25.77 30 0.422

26 Soum 306846 12565 24.42 41 0.326

27 Sourou 216955 6080 35.68 16 0.263

28 Tapoa 301431 14803 20.36 15 0.101

42 Tuy 206309 5717 36.08 13 0.227

43 Yagha 150058 6461 23.23 9 0.139

29 Yatenga 514218 6587 78.07 74 1.123

44 Ziro 144095 5251 27.44 35 0.667

45 Zondoma 148983 2218 67.18 15 0.676

30 Zoundwéogo 240820 3858 62.43 34 0.881 A Derived from source A. B Derived from source B.

Page 143: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

121

Appendix C. Data Screening

In this appendix we perform the following:

• Test of extreme values by visual inspection. A ‘quick-and-dirty’ scan as to select outlier

variables that need further investigation. The goal of this inspection is actually to retain as

many entries as possible, but be aware of their existence and possible influence;

• Test of missing values. Provides information on the extent of missing data for a single

variable. In case of insignificant missing data (less than 10%) any of the imputation methods

can be applied, or the missing data can even be ignored (Hair et al., 1998);

• Test of underlying assumptions. Prevents potential distortions and biases that occur when the

assumptions are violated. The complexity of the analysis and results may mask the indicators

of assumption violations.

By examining the data we obtain a better understanding of the basic characteristics of the data and

relationships between variables. This will provide us with a better perspective on the complexity of the

statistical techniques, as to be able to better interpret the results. Also, we ensure that the data

underlying the analysis meet all of the requirements for its application.

C.1 Data quality assessment

Outlier detection

Outliers are observations with a unique combination of characteristics identifiable as distinctly different

from other observations (Hair et al., 1998). They need to be considered since they can have an effect

on any type of empirical analysis, and they must be viewed in the perspective of their

representativeness for the population. Outlier detection concerns identifying and possibly deleting

these extreme values. Outliers can be identified from a univariate, bivariate, or multivariate perspective

which of as many as possible should be utilized (Hair et al., 1998).

Univariate detection examines the distribution of observations for each (continuous) variable through

boxplots, and selects as outliers those cases falling at the outer range of the distribution. Univariate

outlier detection can be supported by examining the standard-scores (z-scores). For larger samples (n

> 80) the threshold value is four. Higher standard scores indicate extreme values (Hair et al., 1998).

The Missing Value Analysis (MVA) option in SPSS also provides an indication of the number of

extreme values, based on range of standard deviation. Bivariate detection assesses specific pairs of

variables that have a dependence relationship through scatterplots with confidence intervals at a

Page 144: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

122

specified alpha level. Since the effort needed becomes more substantial with an increasing amount of

variables, only specific relationships are considered.

When outliers are identified and examined, we must decide upon deleting or retaining them. In this we

should seek the balance in representativeness for the population. Below, the detection procedure and

possible deletion of values for this research is described.

Step 1 – Univariate detection (only continuous variables)

• Reservoir density – Density of small reservoir (province) [household dataset]. One clear high

extreme can be detected from the visual examination. However, no values are deleted.

• Reservoir distance – Average distance to small reservoir (province) [household dataset]. Two

clear high extremes can be detected from the visual examination. However, no values are

deleted.

• Population density – Density of the population (province) [household dataset]. One clear high

extreme can be detected from the visual examination. However, no values are deleted.

• Measures of income [household dataset]. For the measures on revenues from different

sources the standardized scores indicate between 20 and 110 outliers per variable. The

boxplot of total revenue (revtotal) shows one extreme value due to an extreme value on

revenue from interest (revtransf). This value is deleted from the analysis.

Figure C-1: Boxplot for indicators of income

Page 145: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

123

• Measures of food consumption [household dataset]. For the variables ‘achali’ and ‘autoali’ –

expenditures and autoconsumption of nutritional products – the univariate analysis (MVA)

shows around 270 extreme high values on each of the variables. It is decided not to delete

any values in this stage, since there might be connections with other variables.

• Measures of healthcare consumption [household dataset]. For the measures on values of

expenditures on health we aggregate into variable ‘tothealth’. By evaluating of z-scores we

see that – for each variable – approximately the top 30 values are considered extreme. The

boxplot shows two high values on the variable ‘tothealth’. These are removed from the dataset

together with the values on ‘value of expenditures on medical analysis’ (J32) and ‘value of

expenditures on medicines’ (J33). Now the boxplot for ‘tothealth’ does not show discontinuity.

Figure C-2: Boxplot for indicators of expenditures on health services

• Access to own resources [household dataset]. All indicators; ‘surface of owned land’, ‘number

of large cattle owned’ and ‘number of small cattle owned’ show significant numbers of extreme

(high) values. However, no values are deleted since no evidence is found that extremes

appear in a non-random manner or are related.

• Hhsize – number of household members [household dataset]. Ranges from 1 to 55. According

to the univariate statistics (MVA) households larger than 13 members are extremes. When

double-checking with the individual dataset, there is indeed a household having 55 members.

Therefore, no values are deleted.

Page 146: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

124

• B5 – age of the individu [individu dataset]. The age ranges from 0 to 99 years. According to

the World Bank (2005) the average life expectancy is 48.5 years and the average age 17

years (World Concern Website). The univariate statistics (MVA) individuals older than 58

years are outliers (2876 (5.3%) of the values). The threshold for the z-score is exceeded for

individuals above 94 years (62 values). The boxplot does not show discontinuity, therefore, it

is concluded to delete values above 94 years.

• Agemonth – age of the child [infant dataset]. For this controlling variable the non-declared

cases are marked with the number 99. Clearly, these values are considered as outliers –

supported with z-score above 4 – and are deleted from the dataset, so instead it has become

a missing value. No other extreme values were detected, since the maximum age was 59

months.

• I5a – weight of the child [infant dataset]. The analysis shows that there are 22 extreme high

values (above 25.5 kg). No values are deleted.

• I5b – length of the child [infant dataset]. The analysis shows that there are 27 extreme low

values (below 38 cm). No values are deleted.

Step 2 – Bivariate detection (only related continuous variables)

• Scatterplot of household size over total expenditures on health [household dataset]. Most

outliers are detected as household size is relatively small, with high expenditures on health.

However, no values are deleted.

0 10 20 30 40 50

Household size

0

100000

200000

300000

Tota

l exp

endi

ture

s on

hea

lth

Figure C-3: Scatterplot ‘household size’ * ‘total value of expenditures on health services’

• Matrix scatterplot of household size over total value of autoconsumption and value of

expenditures on food [household dataset]. Remarkable outliers are when household size is

Page 147: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

125

relatively small, with high values of food consumption, and when household size is relatively

large with low values of food consumption. However, no values are deleted.

0 10 20 30 40 50

Household size

0,00

2500000,00

5000000,00

7500000,00

Valu

e of

exp

endi

ture

s on

nut

ritio

nal p

rodu

cts

0 10 20 30 40 50

Household size

0,00

2000000,00

4000000,00

6000000,00

Valu

e of

aut

ocon

sum

ptio

n of

nut

ritio

nal p

rodu

cts

Figure C-4 left: Scatterplot ‘household size’ * ‘value of expenditures on nutritional products’

Figure C-4 right: Scatterplot ‘household size’ * ‘value of autoconsumption of nutritional products’

• Scatterplot of household size over total revenue [household dataset]. Remarkable outliers are

when household size is relatively small, with high values total income, and when household

size is relatively large with low values of total income. However, no values are deleted.

0 10 20 30 40 50

Household size

0,00

10000000,00

20000000,00

Tota

l rev

enue

Figure C-5: Scatterplot ‘household size’ * ‘total income’

Page 148: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

126

• Scatterplot of length of the child over weight of the child [infant dataset]. The graph does not

show clear extremes, therefore, no values are deleted.

25,0 50,0 75,0 100,0

Length in cm of the child

0,0

10,0

20,0

Wei

ght i

n kg

of t

he c

hild

Figure C-6: Scatterplot ‘length of the child’ * ‘weight of the child’

• Scatterplot of population density over reservoir density [household dataset]. The graph shows

one clear extreme, however, no values are deleted since both data are supported by literature.

0,00000000 0,01000000 0,02000000 0,03000000

Density [res/km2] province level

0,00

100,00

200,00

300,00

400,00

Popu

latio

n De

nsity

[#/k

m2]

Figure C-7: Scatterplot ‘reservoir density’ * ‘population density’

Page 149: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

127

Missing data analysis

Any statistical data analysis should start with an examination of the missing data processes. Missing

data may cause interpretation issues since the available sample size is reduced or the statistical

results are biased (when not missing at random). The main concern of missing data analysis is to

identify the patterns and relationships underlying the missing data, in order to maintain as close as

possible the original distribution of values when any remedy is applied. The extent to which missing

data occur is the second concern.

First step in the analysis is to determine the type of the missing data, i.e. (completely) missing at

random or not missing at random. Dependent on the randomness, a remedy or imputation method can

be chosen. Second step is to determine whether the extent of missing data is low enough to not affect

the results, even if it operates in a non-random matter (Hair et al., 1998). Generally, missing data

under 10% for an individual case can generally be ignored, except when the missing data occur in a

specific non-random fashion. Also, the number of cases with no missing data must be sufficient for the

selected analysis technique if replacement values will not be substituted (imputed) for the missing data

(Hair et al., 1998).

Third step is choice of imputation method, if necessary. There are two options, either the missing data

process is classified as Missing Completely At Random (MCAR), or the process is non-random or

Missing At Random (MAR). Each of both options requires a different approach towards the imputation

of missing data. In case the missing data process is classified as MCAR the following remedies can be

applied (Hair et al., 1998):

• Complete case approach (listwise). This remedy includes only those observations with

complete data. Disadvantages are that this method reduces generalizability in case of any

non-random missing data, and it reduces the sample size.

• All-available approach (pairwise). This remedy includes only valid data and does not actually

replace the missing data, but instead imputes the distribution characteristics (means and

standard deviation) or relationships (correlations) from every valid value.

• Using replacement values. These are estimated values based on relationships among

variables in the sample.

In case the missing data process is classified as MAR or non-random only one remedy is available.

This set of procedures explicitly incorporates the missing data into the analysis, either through a

process specifically designed for missing data estimation, or as an integral portion of the standard

multivariate analysis. The first involves maximum likelihood estimation techniques that attempt to

model the processes possible, e.g. the iterative EM approach. The second involves the inclusion of

missing data directly into the analysis, defining observations with missing data as a select subset of

the sample. This approach is most applicable for dealing with missing values on the independent

Page 150: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

128

variables of a dependent relationship. All observations having missing data are coded with a dummy

variable (cases with missing values have value one, other cases have value zero), then the missing

values are imputed by the mean substitution method. Finally, the relationship is estimated by normal

means. The dummy variable represents the difference for the dependent variable between those

observations with missing data and those observations with valid data. The dummy variable coefficient

assesses the statistical significance of this difference (Hair et al., 1998).

• For none of the selected indicators for poverty in the household dataset the percentage of

missing values exceeds 1% (let alone 10%). For simplicity we exclude cases listwise.

• For none of the selected indicators for poverty at individual dataset the percentage of missing

values exceeds 1% (let alone 10%). For simplicity we exclude cases listwise.

• Famine [infant dataset] – Selected indicators are ‘malnutrition’, ‘wasted’, ‘stunted’ and

‘underwei’. The missing values on ‘stunted’ and ‘underwei’ seem to be non-random. But the

cross-tab does not actually proof non-randomness. Since we use famine as an aggregate

measure we should investigate appropriate remedies. For simplicity we exclude cases

listwise.

Table C-8: Univariate statistics famine

Table C-9: Crosstabulation famine

Low height for age (stunted) * Low weight for age (underwei)

Low weight for age

1 2 Total

1 2615 1153 3768 Low height

for age 2 943 2847 3790

Total 3558 4000 7558

• Famine [infant dataset] – Selected indicators are ‘length’ and ‘weight’. None of the indicators

has over 10% missing data. For simplicity we exclude cases listwise.

N Missing

Count Percent

wasted 7645 565 6,9

stunted 7558 652 7,9

underwei 7558 652 7,9

malnutrition 7130 1080 13,2

Page 151: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

129

Table C-10: Univariate statistics BMI

C.2 Assessing underlying assumptions

Normality

Normality refers to the shape of the data distribution for an individual continuous variable and its

correspondence to the normal distribution. If the variation from the normal distribution is sufficiently

large, all resulting statistical tests are invalid, because normality is required to use the F and t statistics

(Hair et al., 1998).

The severity of non-normality is based on two dimensions; the shape of the offending distribution and

the sample size. The shape of any distribution can be described by measures of kurtosis and

skewness. Kurtosis refers to the peakedness or flatness of the distribution compared with the normal

distribution. Skewness is used to describe the balance of the distribution (lack of symmetry). The

effect of sample size with regard to the normality of the distribution is that in larger samples the

detrimental effects of non-normality are reduced. Non-normality can have serious effects in small

samples (less than 50 cases), but the impact effectively diminishes when sample sizes reach over 200

cases (Hair et al., 1998).

The simplest diagnostic test for normality is a visual check of the histogram that compares the

observed data values with a distribution approximating the normal distribution. A more reliable

approach is the normal probability plot, which compares the cumulative distribution of actual data

values with the cumulative distribution of a normal distribution. The normal distribution forms a straight

diagonal line. Statistical tests for normality are e.g. Shapiro-Wilks test and a modification of the

Kolmogorov-Smirnov test, that calculate the level of significance for the differences from a normal

distribution (useless in samples with less than 30 cases, or over 1000 cases) (Hair et al., 1998).

The test of normality provides the Kolmogorov-Smirnov statistic. This assesses the normality of the

distribution of scores. A non-significant result (significance value of more than .05) indicates normality.

In case the significance value is .000 it is suggesting violation of the assumption of normality. This is

quite common in larger samples (Pallant, 2001).

Missing

N Count Percent

weight 7813 397 4,8

length 7744 466 5,7

underwei 7757 453 5,5

Page 152: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

130

Homoscedasticity

Homoscedasticity refers to the assumption that dependent variables exhibit equal levels of variance

across the range of predictor variables. Homoscedasticity is desirable because the variance of the

dependent variable being explained in the dependence relationship should not be concentrated in only

a limited range of the independent values (Hair et al., 1998). Most cases of heteroscedasticity are a

result of non-normality in one or more variables. Thus, remedying normality may not be needed due to

sample size, but may be needed to equalize the variance.

Violation of the assumption of equal variances between pairs of variables (homoscedasticity) can be

detected by either residual plots or simple statistical tests. The most common application of graphical

tests is based on the dispersion of the dependent variable across the values of either the continuous

independent variables. Visual examination departures from an equal dispersion shown by shapes as

cones (small dispersion at one side of the graph, large dispersion at the opposite side), or diamonds

(a large number of points at the centre of the distribution). The statistical test for equal variance

dispersion assesses the equality of variances within groups formed by categorical variables. SPSS

provides the Levene test for homogeneity of variance, which measures the equality of variances for a

single pair of variables (Ho, 1996). If the Levene statistic is significant at the .05 significance level or

better, we reject the null-hypothesis that there are equal variances. If more than one continuous

variable is being tested the Box’s M test is applicable (Hair et al., 1998).

Linearity

Because correlations represent only the linear association between variables, non-linear effects will

not be represented in the correlation value. This omission results in an underestimation of the actual

strength of the relationship. The most common way to assess linearity is to examine scatterplots

(straight line depicts linearity) of the variables, and to identify any non-linear patterns in the data. An

alternative approach is to run a simple regression analysis and examine the residuals. These reflect

the unexplained portion of the dependent variable, thus, any non-linear portion of the relationship (Hair

et al., 1998).

Linearity can easily be examined by residual plots. For non-linear relationships, corrective action to

accommodate the curvilinear effects of one or more independent variables can be taken to increase

both the predictive accuracy of the model and the validity of the estimated coefficients (Ho, 1996).

Page 153: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

131

Test for normality

• All continuous data of the infant dataset (age, length, weight) are not normally distributed

according to the Kolmogorov-Smirnov test, i.e. the significance of the test-statistic is zero.

However, the visual examination for the histogram and normal Q-Q plot reveals a tendency

towards normality for the indicators ‘length of the child’ and ‘weight of the child’.

• All data of the household dataset are not normally distributed according to the Kolmogorov-

Smirnov test, i.e. the significance of the test-statistic is zero. However, the visual examination

for the histogram and normal Q-Q plot reveals that some indicators (e.g. ‘household has

problems satisfying nutritional needs’, all indicators on ‘time to reach resource/service’) look

like they are normally distributed, but none of the histograms indicates normal distribution.

Therefore, it is assumed none of the indicators are normally distributed; hence non-parametric

tests should be applied.

• All data of the individual dataset are not normally distributed; all significance values of the

Kolmogorov-Smirnov test are zero and neither the visual examination reveals tendency

towards normality.

• Indicators ‘reservoir density’ and ‘population density’ are not distributed; all significance values

of the Kolmogorov-Smirnov test are zero and neither the visual examination reveals tendency

towards normality.

Test for homoscedasticity

We apply one-way ANOVA with ‘number of the province’ as factor.

• For all variables of the infant dataset the test of homogeneity of variances has a significance

value of zero, except for the variable ‘age of the child’. For all other variables we reject the

hypothesis that there exists homogeneity; hence there exists heterogeneity.

• For all variables of the household dataset the test of homogeneity of variances has a

significance value of zero. For all variables we reject the hypothesis that there exists

homogeneity; hence there exists heterogeneity.

• For all variables of the individual dataset the test of homogeneity of variances has a

significance value of zero. For all variables we reject the hypothesis that there exists

homogeneity; hence there exists heterogeneity.

• Indicators ‘reservoir density’ and ‘population density’ are assessed to be heterogeneous.

Page 154: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in
Page 155: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

133

Appendix D. Explorative Correlation Analysis

In this appendix gives the results of the explorative correlation analysis, the outcomes of which are

drawn up in Chapter 6. This appendix is not meant for exhaustive reading, moreover as detailed

background to the performed analysis. Therefore, it is given digitally on the enclosed CD-ROM.

Page 156: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in
Page 157: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

135

Appendix E. Explorative Regression Analysis

In this appendix gives the results of the explorative regression analysis, the outcomes of which are

drawn up in Chapter 7. This appendix is not meant for exhaustive reading, moreover as detailed

background to the performed analysis. Therefore, it is given digitally on the enclosed CD-ROM.

Page 158: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in
Page 159: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

137

Appendix F. Validation Tables

In this appendix gives the results of the technical validation of the correlation analysis, the outcomes of

which are drawn up in Chapter 9. This appendix is not meant for exhaustive reading, moreover as

detailed background to the performed analysis. Therefore, it is given digitally on the enclosed CD-

ROM.

Page 160: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in
Page 161: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in

139

Appendix G. Interpretation Tables

In this appendix gives the results of the correlation and regression analysis for the rural and urban

case, the outcomes of which are drawn up in Section 8.2: Interpretation of the rural versus urban

environment. This appendix is not meant for exhaustive reading, moreover as detailed background to

the performed analysis. Therefore, it is given digitally on the enclosed CD-ROM.

Page 162: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in
Page 163: Water Storage and Poverty faculteit...Water Storage and Poverty Impact analysis of small reservoirs on the well-being of rural households in Burkina Faso Master thesis submitted in