Water conflicts and cooperation up and down African rivers
Esther Delbourg & Eric Strobl
« Why go to war over water? For the price of one week’s fighting, you could build five desalination plants. No loss of life, no international pressure and a reliable supply you don’t have to defend in hostile territory ».
Israeli Defence Forces analyst in Wolf, 1995b
This study investigates the impact of water endowments over cooperation and conflict be-tween African transboundary countries from 1949 to 2007. Using the Basins at Risk database (Wolf, Yoﬀe and Giordano, 2003), we make two contributions to the literature: first, we take account of an aspect largely ignored which is relative water scarcity within transboundary river basins, namely streamflow. Second, we concentrate our study on upstream-downstream country pairs to understand how exogenous power asymmetries over water can aﬀect the mechanisms of water cooperation and/or conflict. Our results show that over the last 50 years, downstream nations have played a decisive role in triggering interaction and cooperation with their upstream counterpart, often allowing geographical asymmetries to be oﬀset by economic leverage over the region. In particular, the availability of water seems to increase cooperation (or conflict) only when the downstream nation is at least as well (worse) oﬀ than in the previous years. Results indicate that dams are a factor of cooperation rather than conflict and that cooperation is cor-related with past interaction, meaning that transboundary basins with a history of cooperation over water are likely to keep cooperating in the near future. On the other hand, past conflicts do not aﬀect likeliness of cooperation today.
Over half of the African population lives along major river basins shared with at least one or two countries, making transboundary water management a key policy issue for water security. Sharing a common river basin implies that any use of the resource will aﬀect, in some way, any possible use by others. As major rivers such as the Congo/Zaire, the Nile or the Niger are shared by more than ten riparian countries, it becomes inevitable for them to interact and find common grounds regarding the management of their resource (Yoﬀe et al., 2004; Uitto and Duda, 2002; Yoﬀe et al., 2003; Kameri-Mbote, 2005; Lindemann, 2005; Turton, 2005; Dinar, 2008, 2009).
A major obstacle to smooth cooperation over water is the fact that countries are not equal in accessing the resource and their ability to use it. Countries are first and foremost constrained by their geographical features, such as relative position within a basin, soil and climatic conditions. Second, economic and political leverage can overturn those geographical advantages, enabling less advantaged countries to enforce their own rules of water allocation. For instance, the 1959 bilateral sharing agreement between downstream Egypt and Sudan over the Nile waters was enforced because of the economic and military force they could exert on upstream nations. As upstream Ethiopia is now growing stronger and gaining international support, a new power balance is settling in the region. In 1992, Zimbabwe avoided serious economic drawbacks when the RDC (former Zaire) accepted to forgo part of its hydroelectricity share to oﬀset power shortages due to important droughts. In this case, Zaire was undoubtedly advantaged by its upstream position within the basin and higher potential for hydroelectricity production. Recent allegations of downstream Mozambique cutting hydroelectric supplies to Zimbabwe over unpaid debts in 2008 is, on the other hand, a case of economic leverage overtaking geographical asymmetries.
All three examples, taken from many cases in Africa, show that the very nature of trans-boundary management is aﬀected by exogenous and endogenous asymmetries between countries. Understanding what drives countries to interact and determines whether the outcome is cooper-ative or conflictive thus requires to disentangle these various mechanisms. This article addresses the determinants of water interactions and their outcome, by building upon an important liter-ature in the disciplines of economics, geography and international relations.
Starting with the popular mindset that water will become a source of conflict in the coming century (Starr, 1991; Gleick, 1993; Lowi, 1993; Homer-Dixon, 1994; Ismail Serageldin, World Bank, 1995; Klare, 2001), the international relations and economic literature agree that the very nature of transboundary water management provides higher incentives for states to enter into a cooperative mode. They also argue that the history of water politics is one of cooperation rather than one of confrontation (Allan, 1997; Wolf, 1998; Yoﬀe et al., 2003 and Turton, 2005). Yet this does not imply that water agreements are fully cooperative or that tensions have not occurred before giving way to eﬃcient collaboration.
To understand such cooperative outcomes, the literature first turned to politics and the absence of relevant and eﬃcient institutions (Kameri-Mbote, 2005; Bhaduri and Babier, 2008; Dinar, 2008; Brochmann, 2012) and how they shaped the very perception of potential economic and political gains from cooperating instead of acting unilaterally (Sadoﬀ and Grey, 2002, 2005; Whittington et al., 2005, 2006; Ambec and Ehlers, 2008). In particular, Sadoﬀ and Grey (2002) argue that identifying the inter-related benefits of cooperation beforehand1 is central to succesful transboundary management schemes2.
Recent models even started to account for hydro-geographic variables and how they would aﬀect hydropower and irrigation potentials. The Nile Economic Optimization Model (NEOM) computed by Whittington, Wu and Sadoﬀ (2005) and Wu and Whittington (2006) determines the annual pattern of water use that will maximize the sum of economic benefits from irrigated agriculture and hydropower. Their results suggest that Egypt would gain tremendous value in outsourcing hydroelectricity to upstream Ethiopia where production potential is higher given its position in the basin. This would, in turn, increase Ethiopia’s share of the Nile waters with positive externalities for domestic, agricultural and industrial consumption of water. Neverthe-less such a recommendation, based on hydro-geographical and cost-benefit analysis, was until recently incompatible with Egyptian economic and political considerations.
The previous results directly stem from the upstream-downstream asymmetry between Egypt and Ethiopia, giving an economically weaker country (Ethiopia) an undeniable advantage over a more economically powerful one (Egypt). Upstream-downstream configurations are particularly challenging because of the unidirectional (Rogers, 1997; Dombrowski, 2007) and/or reciprocal externalities (Barrett, 1994) they induce. Negative externalities in an upstream/downstream configuration can derive from upstream storage, through the use of dams or water pollution. Positive externalities can include upstream wastewater treatment or provision of retention area. Although Turton (2005) explains how availability of water has dictated economic development and induced positive externalities within the Orange basin in the last 50 years, no study has proved it empirically.
Indeed, these externalities heavily rely on the issues that countries are willing to address together. For instance, it is initially more complicated to get countries to act upon water alloca-tion issues than in the area of environmental protection, an area which easily obtains consensus (OECD/CSAO, 2009). The Basins at Risk project (Wolf et al., 2003), which we use in this study, finds that treaties over water quantity, quality, joint management and hydropower tend to be highly cooperative while conflictive relations tend to center around quantity and infras-tructure concerns. Yet they still conclude that no single indicator can clearly explain conflict or cooperation (among climatic, water stress, government type, dependence variables), arguing that willingness and ability to cooperate are relative to a historical context where diﬀerent types of leverages are at work3.
The complexity of isolating causes of cooperation drove researchers towards another set of explanations, the first pertaining to the types of country pairs and river configurations likelier to interact and cooperate. Song and Whittington (2004) show that transboundary rivers that cross riparian countries with countervailing powers, population and economic hegemon are more likely to cooperate. They also find that river types appear more important in determining cooperative treaties than do country pair types, suggesting that cooperation is easier to achieve when there is little history of it4. Based on Shlomi Dinar’s finding (2009) of an inverted U-shape relationship between water scarcity and cooperation, Dinar, Dinar and Kurukulasuriya (2011) study three cooperation variables for the whole of each basin (interaction, number of treaties signed and share of water-allocation issues) and show that more developed states use economic incentives to stir cooperation when they need it.
Lastly, a growing literature has also sought to relate climate change to conflicts, mainly by exploring the rivalry induced by the degradation of available resources, forcing populations to migrate internally or to cross borders (Homer-Dixon, 1994; Hauge and Ellingsen, 1998; de Soysa, 2002; Miguel et al., 2009; Dinar et al., 2014).
Although the literature has addressed geographical and economic asymmetries, surprisingly little attention has been given to the decisive exogenous features of geography, climate and water availability together, arguably constituting an important gap in the literature. First of all, the upstream-downstream feature has never been specifically exploited when analyzing water events. Second, a most relevant variable that is directly aﬀected by upstream/downstream relative po-sitions – the quantity of water running through countries, namely streamflow – has not been studied within the scope of water agreements.
Our paper contributes to the literature on the determinants of water cooperation by empiri-cally investigating how streamflow and evapotranspiration – a measure of plant transpiration and a main climatic parameter – can trigger water cooperation or conflict between African nations that specifically display an upstream-downstream relationship. Using the Basins At Risk data (Wolf et al., 2003) which lists and grades all the water events according to their cooperative or conflictive nature, we build our dataset to determine relative position of countries within basins and use it to understand how economic asymmetries are at play when interacting with geographical features. Our first innovation in this approach is to explicitly take account of actual relative water scarcity within a basin, as measured by historical river flow between nations. Streamflow is computed on an annual basis by Blanc and Strobl (2013) from 1949 to 2007 and contains infor-mation on land cover type, soil characteristics, daily precipitation and network coverage. Our second contribution is the computation of upstream/downstream relations coupling the Hydro1K (USGS), a dataset containing topographically derived data sets of streams and drainage basins with the Pfaﬀstetter numbering system (1989) for watershed identification which describes the regional anatomy and enables us to identify the direction of streams. Combined with geograph-ical, spatial and climatic data, this water event database oﬀers a resource for a qualitative and quantitative exploration of African water issues.
Our results show that the likeliness of interaction over transboundary waters and the coop-erative outcome of such water agreements are greatly the doing of downstream nations. Their relatively weaker position provides incentives to interact when they need to and agreements turn out to be more cooperative when they are relatively better oﬀ in terms of streamflow than the previous year. In particular, the availability of water seems to increase cooperation when there is little asymmetry between upstream and downstream availability of water. Furthermore, wa-ter interaction and cooperation are likelier to occur when the downstream nation is relatively wealthier, suggesting that they use their relative economic strength to initiate interaction and influence the outcome of the agreement. We also find that the building of upstream dams has been a factor of cooperation rather than conflict and that the history of cooperation between two nations does trigger further interaction, providing hope for many African nations and basins for which transboundary water management is fairly recent.
We begin with a brief review of the determinants of water cooperation according to the recent literature. We then discuss the complex nature of our diﬀerent data base and out variables before explaining our identification strategy and displaying our results.
Data and Summary Statistics
Region and Unit of Analysis
This study will focus on 16 river basins in the African continent, namely: the Awash, the Congo/Zaire, the Gambia, the Incomati, the Juba-Shibeli, the Kunene, the Lake Chad, the Limpopo, the Niger, the Nile, the Okavango, the Orange, the Ruvuma, the Senegal, the Volta and the Zambezi. The basins are represented in Figure 1.1 and Appendix 1.6.2 provides further economic and demographic details.
We choose our unit of study to be country pairs to specifically observe the upstream vs down-stream unilateral relationship and enquire into bilateral mechanisms of cooperation and conflict. Our database thus comprises every possible country pair for each basin of study, regardless of whether countries share a common border or not.
Figure 1.1: International River Basins of Africa – Copyright Transboundary Freshwater Dispute Database, 2000
The Basins At Risk scale
The Transboundary Freshwater Dispute Databse (TFDD, Wolf et al., 2003) and the BAR scale, taken from the Basins At Risk project, covers historical incidents of international water cooper-ation and conflict from 1950 to 2008. The events are ranked by intensity from -6 to +6 using precise definitions of conflict and cooperation as defined in Appendix A. Formal water events can occur between two or more countries and the dataset attributes the same BAR scale to each of the countries involved in a specific water-related event (see Appendix C for an example of a query for the Nile basin). As such, water treaties can include countries that do not share a river basin but we limit ourselves to country interactions within river basins and not between them. Studying such interactions goes beyond the scope of this paper because countries are no longer constrained by their relative positions within a specific basin.
The diﬃculty of assessing the intensity of water events lies in the very complexity of defining conflict and cooperation. In the database, water conflicts can be understood as a series of tensions or specific non-cooperative acts perpetrated by one of the countries regarding the others. On the other hand, cooperation will designate a series of agreements and events fostering bilateral or multilateral action for the sake of the common resource. A main issue is that the popular approach that conflict ends when cooperation begins is limited as both of them can co-exist at the same time (Allan, 2012). Also, conflict and cooperation are not to be understood locally in time but rather widespread over several periods which are diﬃcult to identify. The 1959 bilateral agreement between Egypt and Sudan over the Nile waters was still being informally enforced until recently even though this conflictive event was never oﬃcially renewed and does not appear again in the database. The dataset does not take into account such a time eﬀect, as it lists agreements or events in the year they took place and will consider years to follow as empty of such events, almost as if the interaction had ceased existing. This aspect will play a role in interpreting our results.
Also, we might be tempted to acknowledge the absence of conflict rather than the existence of cooperation. This was described by Johan Galtung (1969) as « negative peace », namely the absence of violence without further constructive collaboration. In fact, when nothing is happen-ing on a formal basis – that is the absence of cooperation or conflict for a given year or basin – it does not imply that countries are not interacting. It may mean that countries are currently satisfied with the status quo, working on future regulations or following previous treaties; it can also mean that certain forms of cooperation do not require formal agreements.
We use the BAR scale in two diﬀerent ways: first, by observing whether an event has occurred or not in a given year for a specific country pair, with a dependent binary variable. Second, by taking its value (from -6 to +6) when country pairs have interacted at least once, according to Wolf’s terms, over the years 1949 to 2006. We built our database in order for each water event to appear once in the database if it only involved two countries and as many times as there are country-pairs if it involved more than two countries.
Upstream vs downstream relationships
Given the complexity of basin and river configurations, there is no international database avail-able that precisely determines upstream-downstream relationships for country pairs with complex configurations. Figures 1.2 and 1.3 illustrate the diﬃculty of determining the type of hydro-geographic relation that exists between two countries, in the whole of the African continent, and in the specific case of the Congo/Zaire river basin. Indeed, a great number of riparian countries do not feature a clear downstream/upstream relationship: rivers can form a perfect border be-tween both countries or flow from country A to country B, with part of it being fed by a tributary source that comes from country B and flows itself into country A.
A main contribution of this article is that we built our upstream-downstream variable through the comprehensive use of Geographic Information Systems (GIS) and exploitation of the Hy-dro1K database from the U.S. Geographical Survey (USGS)5. Making extensive use of Arcgis, a geographic information system (GIS) for working with maps and geographic information, we overlaid African river maps with basins and countries, thus disaggregating our maps into the smallest units of basins possible as defined by the Pfafstetter numbering system. The Pfafstetter numbering system, as developed by the Brazilian engineer Otto Pfafstetter in 1989, describes the regional anatomy of stream networks using a hierarchical arrangement of decimal digits. Watersheds are distinguished between basins, interbasins and internal basins. Basins are the headwater of rivers and do not receive any inflow from other water areas; interbasins receives flow from upstream watersheds and internal basins are closed. The Pfafstetter system attributes levels of classification from 1 to 5 which help identify the direction of inflows and outflows. We provide complete details of our methodology in Appendix D.
Level 1 classification is the highest one and is attributed to major river basins, level 5 being the lowest and attributed to minor rivers or streams. At each level, the four largest basins are identified and assigned Pfafstetter digits 2, 3, 6 and 8 in a clockwise direction. The five largest interbasins are assigned digits 1, 3, 5, 7 and 9, clockwise. Internal basins are assigned the number 0. Basins continue to be subdivided and numbered as before at levels 2 to 5. Interbasins continue to be divided into the 4 largest basins at levels 2 to 5 but instead of numbering the resulting basins in a clockwise direction, the basins are numbered from the most downstream basin to the most upstream. We use this method to determine whether each watershed, at the smallest level, is upstream or downstream from its nearest neighbor.
Our second main innovation comes from using a measure of the water resources over which coop-erative/conflictive events occurs and proxying this by a measure of streamflow. More precisely, streamflow is a component of the water runoﬀ and the main mechanism by which water moves from the land to the oceans. It is measured in m3.s−1 and is mainly formed by precipitation runoﬀ in the watershed and the contribution of other tributary rivers or water bodies. As such, streamflow is shared by countries and crosses borders. It has the particularity of changing from day to day, mostly because of natural6 or human-induced mechanisms and increases as it flows further downstream7. It is a mostly relevant and useful indicator for our study as streamflow will determine both the amount of water available to a country but then also its leverage on or its dependency regarding the whole of the basin.
We use the streamflow computed by Blanc and Strobl (2013) where they compute streamflow data per year and per country using the GeoSFM model (Geospatial Stream Flow Model) built by the USGS (U.S. Geographical Survey). It is a semidistributed physically based hydrologi-cal model, with particular relevance for Africa’s hydrology (Asante et al., 2007a; Asante et al., 2007b). It stimulates the dynamics of runoﬀ processes using spatial information on river basin and network coverage, land cover type, soil characteristics and daily precipitation and evapo-transpiration data. Blanc and Strobl then use the HYDRO1k data to delineate basins and river network. and drainage basins derived from the USGS30 arc-second digital elevation model (DEM) of the world (GTOPO30).
It allows scale modeling and analyses of African rivers.
Evapotranspiration combines normal evaporation and plant transpiration from land surface to atmosphere and is measured in millimetres (mm) per unit of time. It represents an evaporative demand of the air within a basin. Because plants will transpire more as soils are liquid, such a measure is important in order to manage catchments for water supply and irrigation. Contrary to streamflow which flows from country A to country B, evapotranspiration is a purely local variable. It is also a comprehensive weather indicator as it is a combination of solar radiation and temperature.
We use the data computed by Blanc and Strobl (2013)8 where they follow the computation in Hargreaves and Samani (1985):
ET = 0.0023(T avg + 17.8)(T max − T min)0.5Ra (1.1)
where Tavg, Tmax and Tmin are mean, maximum and minimum temperature, respectively and Ra is the extraterrestrial radiation calculated following Allen et al. (1998) 9.
Population, GDP, Dams, Bilateral Trade and Lagged Values
We control for country-specific features that are likely to play a role in whether countries interact over water management and whether the outcome is cooperative or conflictive. We introduce three variables: population, gdp per capita from the Penn World Table (PWT) 7.010 and number of dams per country.
Countries that build dams over the years are more likely to be involved in water events although this information cannot help predict the intensity of such events. Investments in dams will show concern for water management and water access as they aﬀect the shared resource as a whole. It fosters discussion over water issues and increases the likelihood of interaction. On the other hand, an increase in the number of dams also increases the likeliness of non-compatible projects upstream vs downstream. Controlling for the number of dams enables to understand interaction ahead of time. These issues will be further discussed along with the results. The number of dams per country and per year is provided by Strobl and Strobl (2010) where they use the FAO’s African Dams Database, a georeferenced database of large dams11.
Because we assume that the number of dams will aﬀect water events, we also have to assume that formal water agreements will arise with those changes. Taking the lagged values of both those variables will enable us to incorporate their eﬀects over time. Indeed, building a dam upstream or downstream requires public concern for water-related issues and so perhaps a higher and compelled commitment to transboundary management. Building a dam will also aﬀect the shared basin and foster cooperation or tensions.
We follow the same reasoning for real gdp. As a control variable for economic development, it can be highly correlated with unobserved factors influencing the intensity of water events. African economies rely heavily on farming and irrigation, and are thus easily aﬀected by lack of water, harsh climatic conditions and poor water infrastructures. Enhanced climatic conditions may steer local economy and increase real gdp per capita, making the country either less or more cooperative according to the benefits it can reap out of a formal water agreement. Taking the lag value of real gdp per capita enables us to get round this contemporary correlation.
We will also be using a lagged historical variable representing the lagged sum of events a country pair has been involved in since 1949. This is helpful in analyzing the frequency of a country pair in water events, moreover as water events between two countries generally follow each other closely in time, thus supposedly increasing the likeliness of having an event occur as the years go by.
Our dataset comprises 33 countries that display an upstream-downstream relationship over a span of 58 years. Table 1.1 shows the main summary statistics of our variables of interest. « Et » stands for evapotranspiration. The letters « U » and « D » at the end of the variables stand for « upstream » and « downstream ». Note that upstream streamflow has a slightly lower average value than downstream streamflow. As mentioned earlier, the size of rivers and their streamflow increase naturally as it moves downstream given the construction of the streamflow variable.
Table of contents :
1 Water conflicts and cooperation up and down African rivers
1.2 Data and Summary Statistics
1.2.1 Region and Unit of Analysis
1.2.2 The Basins At Risk scale
1.2.3 Upstream vs downstream relationships
1.2.6 Population, GDP, Dams, Bilateral Trade and Lagged Values
1.2.7 Summary Statistics
1.3.1 Modeling interaction
1.3.2 Modeling the outcome of interaction: cooperation or conflict
1.3.3 Extension to the baseline setting
1.4 Regression results
1.4.1 Results for interaction
1.4.2 Results for the outcome of interaction: conflict or cooperation
1.6.1 BARscale as described in Wolf’s database
1.6.2 Description of the African basins studied in this paper
1.6.3 Extract from a query for the Nile basin in the Barscale database
1.6.4 The Pfafstetter numbering system
2 The Globalization of Virtual Water Flows: Explaining Trade Patterns of a Scarce Resource
2.2 Explaining Virtual Water Flows: Water Endowments versus Water Productivity
2.3.1 Bilateral trade flows
2.3.2 Relative water endowments and footprints
2.3.3 Gravity and control data
2.3.4 Summary Statistics
2.4 Econometric Methodology
2.6 Sectoral Analysis
2.8.1 World Customs Organization – HS Nomenclature 2012
3 Food production and cursed water resources: challenging trade diversification mechanisms
3.2 Virtual Water flows and concentrated exports
3.3 Export concentration, agricultural dependence and water resources
3.3.2 The Theil and Water Theil index
3.3.3 Descriptive statistics
3.4 Regression results
3.4.1 Theil, Water Theil and water-dependence
3.4.2 Food export diversification and water efficiency
3.6.1 World Customs Organization – HS Nomenclature 2012
3.6.2 Additional regression results
3.6.3 The Oil Seeds and Sugars sector in Ethiopia