Informal Water Tankers, Their Network Structure, and Drivers of Cooperation and Competition: A Case Study in Beirut, Lebanon

Abstract Areas with chronic water shortages rely on informal water tankers for daily water needs. Informal tankers operate outside the reach of the state, hence the quality and price of the services they provide and the pressure their activities inflict on water systems go largely unchecked. Informal tankers use their social relationships to manage water sources and operate like an informal network. Yet, little is known about the drivers that explain their relationships and their impact on the overall water system. We investigate these issues in the empirical case of Beirut (Lebanon) using 20 semi-structured interviews and social network analysis. We find that: cooperative and competitive relationships between tankers give rise to different network structures; the drivers of cooperation and competition are religion and seniority in the market; climate change affects informal water markets by increasing competition for wells and clients. We propose recommendations for practitioners and policy-makers, and future research.


Introduction
Water intermittence hampers access to safe, reliable and affordable water (Choueiri et al. 2022b).As a result, informal water supply systems have become a fact of life in many areas of the world that struggle with chronic water shortages.In these areas, while public provision exists, a combination of poor infrastructure quality and low trust in institutions drives the public's reliance on informal water suppliers, i.e., non-state actors such as informal water tankers and vendors.The informality and illegality of informal water tankers (Bakker et al. 2008) make it extremely difficult to identify them and study their behavior and relationships, hence empirical research on informal water systems is relatively limited (though see Wutich, Beresford, and Carvajal 2016 for an important exception).However, around 30% to 60% of people in the Global South still rely on informal water suppliers (Ahlers et al. 2014).Since their operation takes place beyond the reach of the state, the quality and price of the services they provide and the pressure their activities inflict on water systems go largely unchecked (Constantine et al. 2017).
Given their prevalence and impact on water sources (in terms of quantity, quality and price), and social stability, it is important to understand how tankers operate, and on what basis they select their partners.To understand how informal tankers operate we analyze their cooperative and competitive relationships, which shapes how they manage their informal market.Therefore, our research question is: what are the drivers of cooperative and competitive relationships among informal water tankers?Like all individuals in illegal markets (Raab and Milward 2003), informal tankers rely on social relationships to support and organize their trade (Ahlers et al. 2014;Jepson and Vandewalle 2016;Balazs and Lubell 2014;Wutich, Beresford, and Carvajal 2016).They maintain both cooperative and competitive relationships (Cheng 2014;World Bank 2019).Cooperation and competition have both positive and negative consequences.For example, cooperative relationships allow informal tankers to self-regulate to prevent resource depletion, but can also lead to collusion, allowing tankers to monopolize prices across a vast geographical area (Wutich, Beresford, and Carvajal 2016).In turn, competition can drive down prices, increasing water affordability.However, lower water prices increase demand, which drives more drilling for well-discovery and more pumping from existing wells, potentially depleting water resources.
Our case study focuses on the empirical case of informal water tankers in the city of Beirut (Lebanon).Beirut is taken as an example where the informal water system, composed of water tankers, appeared first in the 1970s.Our network comprises the informal water tankers that are connected through social ties of either cooperation or competition-two types of relationships that structure illegal markets of all kinds (Beckert and Wehinger 2013).
We rely on semi-structured interviews and Social Network Analysis (SNA) to tackle the following two goals: (1) identify and describe the network structure of informal water tankers; (2) identify the drivers of network structures of cooperation and competition.SNA focuses on how patterns of interaction between individuals may affect their behavior (Everton 2012).After confirming that tankers cooperate and compete via qualitative interviews, we create two networks of water tankers: a network composed of cooperative ties, and another composed of competitive ties.We analyze each network in turn to identify characteristics and drivers of their structure using a standard method for inferential analysis of network structures: the Exponential Random Graph Model (ERGM).
Our analysis of qualitative and quantitative data from fieldwork in the country reveals that cooperative and competitive relationships between water tankers give rise to different network structures, which are driven by different social processes: trust and reciprocity in the cooperative network, and adversarial competition against new entrants in the competitive network.The implications of the study have relevance beyond the specific case of Beirut, for all socio-political contexts where informal water markets flourish on the background of lacking public investment in water infrastructure and poor management of water resources.
We put forward two recommendations for practitioners and policymakers to deal with informal water markets: improving the quality of public infrastructure, and investing in data collection on tankers' operations and water sources to prevent groundwater depletion and curb violence.The case study of Beirut provides a wealth of knowledge into informal water markets and tankers.Further research in other contexts is needed to enable comparative analysis and devise appropriate strategies of incorporation, regulation or sanction of informal water tankers.

Informal Water Systems
Informal water systems are widespread geographically as they can cross Global South and North borders (Ranganathan and Balazs 2015;Jepson and Vandewalle 2016), and in some cases can be the dominant system (Ahlers et al. 2014).Informal water systems can take different forms and shapes for example this can include water tankers delivering water to households, people directly pumping water from wells or from existing central piped infrastructure, or using water bottles for drinking needs.These informal systems tend to operate in complete autonomy (Stein, Ernstson, and Barron 2011) with "little or no state regulations" (London et al. 2021).In some areas, they are accepted as new modes of water provision (Ahlers et al. 2014;Wutich, Beresford, and Carvajal 2016;Walter, Kooy, and Prabaharyaka 2017).Tankers operate at multiple scales, pursue their own interests, and enable water access to those experiencing frequent water shortages (Ahlers et al. 2014;Peloso and Morinville 2014).
The literature on informal water systems has so far focused on theoretical definitions and empirical analysis of the socio-economic implications of informal water sources.Theoretically, studies have defined informality by analyzing its relation and interdependence with formal systems (Roy 2005;Ahlers et al. 2014;Schwartz et al. 2015;Liddle, Mager, and Nel 2016).Empirically, studies have quantified the socio-economic impacts of informality, by assessing added cost of informal sources and water access disparities (Pattanayak et al. 2005;Christian-Smith et al. 2013;Nastiti et al. 2017;Amit and Sasidharan 2019;Choueiri et al. 2022b), and their energy and environmental impacts (Choueiri et al. 2022a).Other contributors have analyzed psychological distresses of end-users that stem from water insecurity (Jepson, Budds, et al. 2017;Jepson, Wutich, et al. 2017;Young et al. 2019) and investigated health hazard impacts from water quality and waterborne diseases from unmonitored informal water sources (El-Fadel et al. 2003;Constantine et al. 2017).However, few studies have looked at the relationships between informal water tankers (though see Wutich, Beresford, and Carvajal 2016).

Relationships in Informal Water Governance Settings
In the context of informal water markets, we can think of water extraction and use as posing the same collective action problems of common pool resources (McGinnis and Ostrom 2014).Common pool resources are non-rival (i.e., shared) and non-excludable.This means that no user of the resource can prevent other users from accessing it and take as much as they want of it.According to Hardin (1968), this inevitably leads to resource overuse and eventually depletion: every user will overharvest the resource, knowing that even if they do not, some other user will.The only solution, according to Hardin, is 'mutually agreed coercion'-in order to survive, users have to accept enforcement of property rights by either allowing government to own and manage the resource, or privatize the resource.This point was challenged and dismantled by Ostrom (1999) who argued that if resource users can communicate, it is rational for them to cooperate and regulate resource extraction to prevent depletion and collectively benefit from extraction in the long term.Government ownership and privatization are not needed; cooperation can solve the collective action problem.
Research on natural resource management of common pool resources show that this is, in fact, the case: cooperation between users prevents resource depletion and ensures resource sustainability in the long term (Ostrom 1999;Schlager 2002;Agrawal 2014;Slough et al. 2021).Given the positive impact of cooperation on resource management, the literature on water governance focuses on the relationships between actors to understand what drives their successful cooperation (Eberhard et al. 2017).Analyzing social relationships is the realm of social network analysis.Thus, scholarly work in this field investigates the drivers of cooperation in social networks between resource users (Hileman, Bastos, and Lubell 2018).
Informal water tankers (such as water vendors) also have cooperative relationships.They have a shared interest in preventing water resource depletion to ensure the viability of their business.According to the theory (Ostrom 1999), if tankers are able to communicate they should be able to prevent resource contamination or depletion.Informal water tankers, however, also compete with one another (Cheng 2014;Wutich, Beresford, and Carvajal 2016).Cooperation and competition can impact resources in different ways, including control of resource use and extraction (Easter, Rosegrant, and Dinar 1999, 393-410), service quality, and service price (Solo 1999).Cooperation can improve water access (Wutich, Beresford, and Carvajal 2016;Amirova, Petrick, and Djanibekov 2019), help informal water suppliers self-regulate and control the quality of delivered water (Wutich, Beresford, and Carvajal 2016).However, cooperation can also lead to collusion on prices, dampening affordability and forcing end-users to spend an increasing amount of their income on water (Wutich, Beresford, and Carvajal 2016).
Competition also has consequences: on the one hand, competition drives down water prices, improving water affordability (Easter, Rosegrant, and Dinar 1999, 393-410;Collignon and V ezina 2000); on the other hand, however, competition can alter existing patterns of control of water access (e.g., wells), leading to conflict between tankers.Further, competition can disrupt the cooperative equilibrium and lead to resource overuse and, therefore, depletion.
In the following section, we borrow from the literature on social networks in illegal markets and the literature on resource management to formulate our hypotheses concerning the drivers of cooperative and competitive relationships between informal water tankers in the city of Beirut.

Hypotheses
When individuals are involved in illegal activities, they tend to cooperate based on their level of trust (Berman 2011;Tilly 2005).This is because illegal networks have to operate in the shadow of the law.Therefore, every exchange becomes difficult and potentially dangerous (Raab and Milward 2003).Informal water systems are similar to illegal activities since they are organized by individuals that operate outside legal frameworks.For this reason, cooperative relationships of informal water tankers are few and based on mutual trust.Hence, cooperative networks are formed primarily through "strong ties" (Granovetter 1973) between like-minded individuals that have strong partnerships based on repeated interaction.Strong partnerships lead to the formation of triangular network structures (a phenomenon whereby "a friend of a friend is a friend") (Henry and Vollan 2014).Thus, we expect tankers in the cooperative network to have few strong partnerships, leading the network structure to be "closed," fragmented into small "islands" of cooperation that are interconnected by individuals that act as "brokers" who bridge across subgroups (Everton 2012).
Although competition in informal markets is typically scarce given their inherent lack of transparency and information on prices (Beckert and Wehinger 2013), informal tankers do compete for clients.Competition is driven not by trust but by demand: increased demand for water can lead to increased competition between informal water tankers.We expect the network of competitive relationships between informal water tankers in Beirut to be "open" (Bakken, Moeller, and Sandberg 2018), meaning that the network structure will feature little triangulation and more unreciprocated ties of nominations.These considerations lead us to formulate our first hypothesis, which concerns the structure of the networks under investigation: Hypothesis 1: The structure of the cooperative network of water tankers will be decentralized with tankers forming few strong partnerships; in contrast, the structure of the competitive network will be centralized around key individuals.
As mentioned, informal water suppliers may tend to form trust-based ties with those that share similar attributes or beliefs; this attitude is referred to in the SNA literature as "homophily" (Henry, Lubell, and McCoy 2011).We focus on religious homophily mainly because of the religious diversity of water tankers and their historical evolution as this is an important social characteristic in the context of Lebanon (more details are provided in the next section).Ethnic homophily means that those that belong to the same ethnic group and have cultural, religious and linguistic similarities have a tendency to befriend each other or work together.This behavior has been identified in children (Leszczensky and Pink 2019), street gang members (Grund and Densley 2015), and fishermen (Barnes-Mauthe et al. 2013).Homophily creates an "us and them" attitude (Borgatti and Foster 2003), where suppliers of similar religious identity cooperate more, while those of different religions compete.These considerations inform our second hypothesis: Hypothesis 2: Informal water tankers will tend to form cooperative ties with others that share the same religious affiliation, however religious affiliation won't impact the formation of competitive ties.

Case Study Details
Piped water supply for Beirut comes from two main areas, shown in Maps 1 and 2 in the online appendix.The main source is a northern water treatment plant in Dbayeh, which supplies around 77% of the city's water use.The remaining 23% are supplied from a cluster of 11 wells in the southern Naameh area.Water from these locations is usually pumped to reservoirs in Beirut for storage before final delivery.The reservoirs cannot supply water on a continuous basis.To cope with water intermittence, most households purchase additional water from informal water tankers.

Characteristics of Informal Tankers
Tankers tend to be unregistered and operate in complete independence and autonomy from the government (Personal communication Howayek, 2019).They are usually managed by private individuals that own one or several tankers (sometime up to 20 tankers), and vary in capacity (from 1,000 to 30,000 liters).Figure A2 in the online appendix displays the trucks used by informal tankers.Tankers mainly operate during water intermittence periods, usually in summer and dry seasons and can serve multiple customers per day, reaching up to 70 trips per day.
Both registered and unregistered wells supply water to these tankers.The location of tanker businesses is usually determined by the wells.In terms of water quality, the north-eastern periphery of the city is water rich, because of the presence of multiple rivers (the Kalb, Mot and Beirut rivers), that replenish groundwater (Shaban 2020).Howerver, wells within Beirut, and those on the southern periphery of the capital, are of lower quality because of seawater intrusion from years of over-extraction (Alameddine, Tarhini, and El-Fadel 2018).Nevertheless, some tanker businesses locate in those areas and use UV and/or RO filtration systems to treat their water (Personal communication Tanker Owner AS, 2019), as shown in Figure A3 in the online appendix.

Historical Changes: Ethnic and Religious Diversity, Cooperation and Competition
Cultural traditions and identity are part and parcel of water resources management (Akpabio 2011).Lebanon witnessed a 15-year long civil war between Muslims and Christians: religion is an important demographic trait shaping social relationships (Faour 2007).Religion also influences the operations and relationships between informal tankers.The oldest informal tankers are based in the water-rich, northeastern periphery of the capital (see Map 1 in the online appendix), (Personal communication Tanker Owner CHR, 2019).In the northeastern region, the informal water market has historically been led by Christian owners, with a minor presence of Muslim owners (Map 2 in the online appendix).The oldest water tankers had a tendency to work together and cooperate over water sources (Personal communication Tanker Owner SK, 2019).However, the demographics of the tanker operations are changing.Several droughts in the last years (Sarant 2021), have reduced piped water deliveries and increased demand for water from tankers, creating more opportunities for informal water tanker businesses (Maloy 2017).The newer informal tanker entries have been mainly Muslim owners (Personal communication Tanker Owner SK, 2019), so the ratio of Christian to Muslim informal water tankers has been decreasing in recent years.

Data Collection
Qualitative and network data were collected via 20 in-person semi-structured interviews with informal water tankers 1 in and around Beirut over a two-month period, from August 2019 to September 2019.The target population was defined as any person that owns or manages at least one informal water tanker.Interviews took place at offices of water tankers (except for two conducted in caf es).The interviews were in Arabic and all answers were collected as written notes by the first author.Interviewees were recruited with a convenience sampling process, based on their ease of accessibility, willingness to participate, and geographic location (Etikan, Musa, and Alkassim 2016).At the end of every interview, a new list of informal truck owners was generated.The researcher would then contact by phone the new list to take appointments.This sampling process was repeated until contact list closure was reached, and the direct contact list was exhausted, i.e., no new truck contact was answering for new appointments (Alexander, Bodin, and Barnes 2018;Hanneman and Riddle 2005).An estimate of total number of trucks operating in the region was not available; thus it was not possible to gauge the size of the overall network.In total, we interviewed 20 truck owners and managers from a total list of 68 identified informal water tanker businesses, for an overall response rate of 29%.
The interview guide was divided in two.Initial questions focused on the characteristics of informal water truck owners, and factors that have affected and potentially changed their businesses.Interviewees were asked to describe their business, in particular focus on their perception of how the business has changed over time, and their opinion on what has contributed to this change.They were also asked to describe their relationship with the government, and the formal water establishment.Basic attribute data were collected for each interviewee including business size (i.e., number of clients, and number of trucks), geographic focus (i.e., business location, service location and well location), and water quality (i.e., number of wells used and their water quality), and religious affiliation. 2Religious background (Christian or Muslim) was inferred from the names of the owners and not directly asked, given the sensitivity of this topic in the region. 3 The second part of the interviews focused on collecting network data: number and type of ties among tankers (cooperative or competitive 4 ), and the factors that contribute to tie formation.A cooperative relationship was defined as having mainly worked amicably together, communicating or helping each other over clients, maintenance of trucks, wells, and/or prices.Competition was defined as having had an argument, conflict or fight over multiple issues including clients, prices, and/or wells.
Network sampling started with one main business owner.There was no limit on the number of nominated individuals, as it was important to collect as many names as possible since network size was unknown.This approach tends to capture strong ties (based on the number of ties) which helped identify key individuals (Alexander, Bodin, and Barnes 2018).As a result of data collection, the networks have different sizes.Thirty-seven percent of tankers appear in both networks, while several others appear in only one of the networks.This prevents us from analyzing this network as a single multi-layer or multiplex network, where the same individual has different types of relationships.Rather, we analyze the two networks separately-first descriptively and then inferentially-and compare their structures based on our hypotheses.

Qualitative Analysis Based on Interviews
The 20 qualitative interviews are based on a semi-structured interview guide conducted in Arabic.Hence, we started by transcribing field notes and interview answers to English.Then, to identify interview themes, we followed a four-step process based on Ryan and Bernard (2013).We started by proofreading the written material and underlining key phrases.We used two main techniques to identify themes: repetition and missing data.Repetition and thorough re-reading interview answers helps identify occurring and reoccurring topics and regularities mentioned by the interviewees (Bogdan, Taylor, and Taylor 1975).By reading the text repeatedly, we were also able to identify some of the salient themes that are not explicitly expressed by the interviewees (Bogdan, Taylor, and Taylor 1975), mainly because of the sensitivity of the data, including episodes of violence and religious affiliation.Finally, we followed a "pile and sort" method (Weller and Romney 1988): after initial marking and highlighted different sections of the text, we cut and sorted some quotes that seemed important and arranged them into piles of similar topics (Lincoln and Guba 1985).

Social Network Analysis for Informal Water Systems
Social network analysis (SNA) allows to understand network structure and characterize (cooperative and competitive) social relationships (Fischer and Ingold 2020).It allows the researcher to identify, e.g., the most connected individuals in a network (Stein, Ernstson, and Barron 2011;Bodin and Crona 2008).It also helps understanding the role that individual characteristics (or attributes) play in tie formation (McPherson, Smith-Lovin, and Cook 2001), especially looking at how having similar characteristics leads to the formation of cooperative ties, which increase trust among individuals (Henry, Lubell, and McCoy 2011).
SNA has been used to analyze different types of networks including (formal) public and private institutions, self-organized network, networks composed of a mix of stakeholders, policy networks (Fischer and Ingold 2020) and "dark" networks (i.e., networks that are covert and illegal (Raab and Milward 2003))."Most illegal markets are shaped by a network structure" (Beckert and Wehinger 2013).Unlike dark networks' activities that happen completely in a secret and hidden way, informal networks fall between what is legitimate and illegitimate (Roy 2005), where some informal activities happen in the open and are considered accepted by the state, and other activities are sanctioned.Hence, understanding social relations of informal individuals and their impact on water sources is challenging because of the nature of their informality, i.e., hidden characteristics and lack of regulation (Bakker et al. 2008).

Creating Cooperative and Competitive Networks
We used the interviews to construct two networks using the link-trace method (Frank and Snijders 1994), which is commonly used to study hard-to-reach populations such as drug users or criminal organizations.We use three pieces of data collected via the semistructured interviews to construct the networks: (1) list of names of water tankers that are connected to each other, (2) their characteristics (religious affiliation, seniority in the market, etc.), and (3) type of relationship (whether cooperative or competitive).We first develop a database of ties (in SNA this is referred to as an edgelist) with all the names of tankers, their characteristics and relationships.Then, depending on whether the relationship is cooperative or competitive, we divide this dataset in two (a cooperative and a competitive edgelist).We build each network by combining the edge list and node list, allowing us to specify two networks-one cooperative, one competitive-composed of their respective edges and nodes.The two networks comprise the 20 tankers who were interviewed plus all the tankers they named as their cooperators or competitors.This results in a total 68 tankers, of which 25 appear in both networks.

Descriptive Analysis
We analyze our cooperative and competitive networks at three levels: macro, meso and micro, using the igraph package in the R statistical environment (Csardi and Nepusz 2006).This three-tier approach (Wasserman and Faust 1994;Borgatti, Everett, and Johnson 2018;Ebrahimiazarkharan et al. 2020) helps develop a holistic understanding of the networks (Bodin, Mancilla Garc ıa, and Robins 2020).The macro-level analysis identifies overall network characteristics; meso-level analysis detects clustering and subgroups; micro-level analysis identifies key individuals.We focus on six components: network size, density (the number of realized ties compared to the number of possible ties), homophily (the extent to which individual with similar attributes share a tie), reciprocity, cliques (groups of connected individuals), and centrality measures, which identify the most central nodes in the network.The two networks are similar in terms of their density and their average degree (average number of ties per tanker).However, they differ in their levels of centralization (closeness), reciprocity and homophily: the cooperative network features more reciprocal ties and more homophilous links, and is more decentralized.Table A3 in Appendix I includes an extended explanation of the descriptive analysis.The reader should bear in mind, however, that the two networks are not directly comparable because of the discrepancy in tanker numbers: the network of cooperation features 39 nodes (tankers), while the competitive network features 54 tankers.There are 68 unique tankers between the two networks.Of these, 25 tankers appear in both networks.The size of the population of informal water tankers of the city of Beirut is unknown.As explained in more detail in the next section, we carry out our analyses on the assumption that the properties of the networks we observe scale with the properties of the full network, had it been observed.

Quantitative Analysis: Exponential Random Graph Models
Exponential Random Graph Models (ERGMs) are statistical models that explain patterns of ties in a social network leading to network structure (Lusher, Kremer, and Robins 2014).They model the configuration probability of an observed network, compared to all other network configurations that have similar structural characteristics to the observed network, such as the same number of nodes and density (Lusher, Kremer, and Robins 2014).
Similarly to other groups involved in illegal operations, Beirut's informal water tankers are a hard-to-reach population.This means that it is impossible to know the exact size of the network and network data has to be sampled via the link-trace method, which starts with data collection from a convenience sample of individuals and then relies on snowball sampling to reconstruct the whole network (Krivitsky and Morris 2017).As mentioned, the 20 interviews yielded a list of 68 tankers, including the 20 interviewees.Although the sample of tankers is biased by the recollection and information provided by the first seed tanker, it comprises tankers that are heterogenous in terms of ethnicity, religion, and seniority in the market as well as resources (e.g., trucks, see appendix).Hence, we consider that our sample of tankers is representative in the sense that the average number of relations in the inferred network that a tanker with a particular value of a given attribute (e.g., religion) has with other actors with a particular value of a given attribute (in this case, the same attribute) is close to that of the population of tankers.In other words, we consider that the sampled data has a degree distribution similar to that of the population and similar mixing properties (Krivitsky, Handcock, and Morris 2011).Furthermore, to account for missing data in the analysis, we indicate non-response in the ERGM by including a matrix of structural zeros for all the tankers who were not interviewed, who therefore have no outgoing ties to other tankers, but only incoming ties from the tankers who named them.This allows the ERGM to take into account missing data while still being able to develop inference (Handcock and Gile 2010).
To test our hypotheses, we analyze three network properties: centrality, homophily and activity.We build our ERGMs using the R statnet, sna and ergm packages (Handcock et al. 2008;Krivitsky and Morris 2017).Table A4 in Appendix I explains the model terms in detail.

Interviews Themes
Qualitative interviews were grouped into two themes.The first theme concerns drivers of cooperation and competition.It shows that religion and seniority are important: well-established tankers tend to cooperate with tankers of the same religion and who have been longer in the market; they compete with tankers of opposite religion.The second theme concerns the sources and the effects of increased competition: after the dry years of 2013-2015, competition leads to reduced prices but also an increase of violent behavior among the tankers.Below, we provide some quotes from the interviews that illustrate the tankers' perceptions concerning their cooperative and competitive relationships.
Theme 1: The drivers of cooperation and competition.
"We [the Christians] were the first ones to distribute water.Today anyone can own a water tanker.In West Beirut [Muslim Area], there were almost no trucks at all.The competition with the Muslims has increased a lot.They [Muslim Tankers] do business in a different and more aggressive way, they are ready to reduce the cost of the service and reduce their profit margin."-SK(Christian, operating for 58 years)."With the older trucks the relation is a cooperative one.However, with the new companies the relation is a competitive one."-BCH,(Christian, operating for 21 years).
Theme 2: The sources and effects of competition.On prices: "Since 2014, what I see is that the competition has increased.Also, the prices have dropped a lot.The truck used to be for a 100$ or 150$ (2014-2016) today it is around 50$."-AL, (Christian, operating for 5 years)."Therewas a boom in the trucking businesses after 2014, which was a very dry winter.There was so much work that the business did not stop during that year except for the month of March."-PR,(Christian, operating for 31 years)."The increase of competition has decreased my work by 70%"-AS, (Muslim, operating for 20 years).

On social stability:
"They [informal water tankers] help each other out by kicking out a new tanker that is coming to the neighborhood"-PR, (Christian, operating for 31 years)."New water tankers have a different behavior [ … ] in some cases the drivers are beaten up"-BCH, (Christian, operating for 21 years).

Descriptive Analysis Results
Figure 1 shows both the cooperative and competitive networks of informal water tankers.Each colored point represents an informal water tanker, and the lines represent their connection (also referred to as edges).nodes are color-coded according to religious affiliation.Edges are also color-coded: dark grey edges represent ties between Christian tankers; light grey edges represent ties between Muslim tankers; and black edges represent ties between tankers of different religious affiliation.Figure 1 suggests that the cooperative network is much more homophilous than the competitive network (more connections with individuals of similar religious affiliation); in the competitive network black edges are much more numerous.Figure 1 was generated using the ggnet2 function from the GGally package in R (Schloerke et al. 2018).
Table 1 shows that the cooperative network we were able to reconstruct is dominated by Christian tankers, who have been in the market on average 44 years.By contrast, the competitive network features both Christian and Muslim tankers.The presence of tankers who are "new entrants" in the competitive network brings down the average years of operation to 24 years.Note that only 25 tankers are found in both networks: the rest are tankers who have been nominated but did not respond to invitations to interview.Therefore, the two networks are not directly comparable.Still, an overview of descriptive network statistics allows us to grasp some of the structural differences that we will probe in the statistical network analysis.For example, the measure of closeness centralization (a measure of how close nodes are to each other, on average) suggests that nodes in the competitive network are "closer" to each other than in the cooperative network; that is because many interviewees named the same tankers as competitors.In contrast, closeness centralization is much lower in the cooperative network, because four of the Muslim tankers in that network form their own disconnected component (see Figure 1).Even after removing those four tankers, the closeness centralization statistic in the cooperative network is 0.29: still lower than the competitive network, suggesting that there is more variation in tankers' nominations of their cooperators than in their nominations of competitors.Refer to Appendix I for an extended description of the network at three levels of analysis: macro, meso and micro.

Quantitative Analysis Results
The model results for the cooperative and competitive network in Table 2 lend support to our expectations.Centralization in the cooperative network is significant and positive, which confirms our expectations in hypothesis 1: many individuals have a similar number of cooperative ties, showing that the network is decentralized.At the same time, reciprocity and triangulation are positive and strongly significant, indicating that tankers form trust-based relationships with their partners.The coefficients of ERGMs should be interpreted in log-odds, where each coefficient indicates whether the corresponding parameter estimate influences tie formation in the network, all else being equal.The density parameter indicates the probability that any tie in the network exists.The tie probability at the intercept (i.e., the density parameter) for the cooperative network is exp(À4.20)/(1þ exp(À4.20))¼ 0.01, which is lower than the actual density of the network, which is 0.05.These discrepancies arise from the fact that the ERGM model includes the matrix of offsets for the nodes that cannot send any ties because they have not been interviewed.The conditional log-odds of two tankers having reciprocal ties in the cooperative network is, reading from Table 1, À4.20 þ 2.31 ¼ À1.89.The probability-calculated using the formula above-is 0.13.By contrast, the probability of two tankers having reciprocal ties in the competitive network is only 0.02: the log-odds of the density term plus the reciprocity term amount to À3.62, which gives exp(À3.62)/(1þ exp(À3.62))¼ 0.02.
While the positive coefficient on indegree centralization indicates that the cooperative network is decentralized, the negative coefficient of indegree centralization in the competitive network indicate that the network is centralized (Levy 2016), meaning that there are tankers who receive more nominations than others.The lack of significance of the coefficient for reciprocity may be a result of non-response which is higher in this network than in the cooperative network.
Moreover, the competitive network shows no signs of religious homophily, confirming hypothesis 2. The coefficient for activity is similarly not significant, suggesting that neither Christian nor Muslim tankers named more competitors than their counterparts of the other religion.
Religious homophily is strongly significant, which means that individuals form a tie with others of the same religion.As for activity, the coefficient is not significant, indicating that Christian and Muslim tankers have similar patterns of cooperative ties, with neither group being significantly more active (i.e., possessing more ties) than the other.
The fit of the models to data is shown in Figures A4 and A5 in the appendix.

Discussion
This study used qualitative interviews and Social Network Analysis (SNA) to identify and characterize the network of informal water tankers in Beirut (Lebanon) and analyze drivers of their cooperative and competitive relationships.We have three main findings.Firstly, cooperation and competition correspond to different network structures: the cooperative network is fragmented and decentralized; the competitive network is centralized, suggesting new entrants disrupted the preexisting competitive equilibrium.Secondly, religion and seniority contribute to the formation of cooperative ties: long-established Christian tankers tend to cooperate with each other and to compete predominantly with Muslim tankers, who tend to be new entrants.Thirdly, climate change affects informal water markets: the drought of 2013-2015 caused a sharp rise in the demand for water, prompting Muslim tankers-then moderately active in the market-to quickly expand their operations to capture that demand.The effect of increased competition is twofold: while reducing prices for end-users, competition increases overall water extraction and social instability by leading to episodes of aggression and violence between tankers competing for wells and clients.
The combined reading of the descriptive and the statistical analysis of the network data shows that seniority in the market is another powerful driver of network structure.Cooperation is higher among tankers who have been operating longer.Here we see the interaction between seniority and religion: Christian tankers have typically been active in the market for longer than their Muslim counterparts.In fact, tankers in the cooperative network we observed (mostly Christian) have been active in the market for 44 years on average, while tankers in the competitive network show an average of 24 years in operation.
The dry years of 2013-2015 brought change into the informal water market of Beirut.The recurrent droughts have increased household water intermittence, and the demand for the services of informal water tankers.This has led to an increase of competition.Twelve of the twenty interviewees mentioned that in the previous few years competition had increased, mainly after the dry years.Many of the tankers who expanded their operations immediately before and after the drought were Muslim.Thus, Christian tankers tended to nominate Muslim tankers as their competition.
Informal water sources cost "4 to 30 times" more than municipal water (Wutich, Beresford, and Carvajal 2016;Choueiri et al. 2022b).Competition might benefit endusers by increasing water access and affordability: as mentioned by 10 interviewees, competition is breaking the old informal water market, resulting in reduced water prices for households.From a resource perspective, competition can lead to the "tragedy of a commons" situation related to groundwater exploitation (Hardin 1968;Ostrom 1999).The expansion of the informal market may push water tankers to continue pumping and distributing groundwater, which may exacerbate existing seawater intrusion (Alameddine, Tarhini, and El-Fadel 2018).Moreover, competition might lead to an increase in violent behavior among water tankers.Three interviewees mentioned an increase of violence between tankers, a conduct that they believe is provoked by the new market entrants.This is common in illegal markets that feature violence as a way of enforcing rules and punish defection or scare away competitors (Raab and Milward 2003;Beckert and Wehinger 2013).

Policy Implications and Conclusions
In many areas around the world, relying on informal water markets is often a necessity rather than a choice (Wutich, Beresford, and Carvajal 2016).Hence, it becomes important to understand the drivers behind informal water systems and their impact on overall water resources.The literature on informality initially adopted a negative connotation of informal water systems, referring to them as activities that only serve the urban poor (Hart 1970;AlSayyad and Roy 2004, 7-30).More recent scholarship recognizes that informal water tankers have the capacity to adapt to changing conditions and allow communities to be more resilient to water shortages (Revell 2010).
In the context of informal water markets, water resources can be considered common pool resources for informal tankers.Tankers have a common interest in preserving water resources to sustain their trade, which should prompt the formation of cooperative, close-knit ties between them.This is certainly what happened in our case study: the city of Beirut, where informal water provision has been coexisting alongside formal public provision for many years already.However, this being a market, it is subject to the laws of supply and demand.A rapid increase in demand prompted the emergence of new market entrants who, in trying to gain market quotas for themselves, disrupted the cooperative equilibrium.Competition drives down prices, but triggers social instability and potentially a race for water that may have adverse impacts on the sustainability of water resources in the area.
Our hypotheses and findings focused on two elements: the network structure of informal water tankers and individual characteristics of the tankers.In terms of network structure, we find that the cooperative network is decentralized and fragmented, with little triangulation.Thus, cooperation is structured around rather small groups of tankers who share a small number of trusted relationships.In contrast, the competitive network is more centralized, with some dominant "new entrants" and adversarial relationships between them and the tankers whose business they disrupted.As for individual characteristics, we find that cooperative ties follow religious lines: Christian and Muslim tankers tend to form cooperative relationships with tankers of their same religion.However, religious affiliation does not appear to drive the establishment of competitive ties and tankers compete with other tankers of either religious group.Still, qualitative evidence shows that most of our interviewees see their main competitors as being the newest (Muslim) entrants in to the market.
One of the main reasons for the emergence of informal water markets is the scarce efficiency of water infrastructure in supplying residents.Today, leakage in the Lebanese water grid is 45% (Shaban 2020).Recent efforts by the water utility to reduce leakages, i.e., installing household water meters to monitor losses (Personal communication Howayek, 2019) is a step to improve grid efficiency.However, in Lebanon, public institutions have a history of mismanagement (El-Fadel et al. 2003), and investment in public infrastructure is rarely sufficient.Moreover, we also know that formal and informal water systems work in tandem (Jepson, Budds, et al. 2017;Ahlers et al. 2014).Some scholars have suggested to formalize informal tankers through management frameworks, registration fees, licenses and taxes (Constantine et al. 2017;Ahlers et al. 2014).Our findings would help water institutions better implement these recommendations with the knowledge of the drivers of social relationships between tankers.Namely, we propose a two-pronged strategy.Firstly, water institutions and practitioners should bear in mind that cooperative and competitive relationships exist simultaneously; we split them in two different networks for analytical clarity and convenience, but really the two networks overlap: they are a single network.The results of the analysis suggest that if public authorities did investigate the state of the informal water market, they would stumble upon key central actors quite quickly and could use them to reconstruct the structure of the broader network.Although collecting data on water tankers is difficult given their informal nature, this study alone was able to identify a total of 68 tankers, locate them, their service area and their wells.By mapping the tankers' location and the location of their wells using remote sensing tools, we would be able to monitor how much groundwater they are pumping, which will allow us in return to have a better assessment of the state of Lebanese groundwater.Pending much needed public investment in the grid (whose results will take time to materialize), the water institution could work with the informal water tankers to expand its own service area while the public grid is being upgraded.
Secondly, whichever form of mainstreaming of informal water tankers into "official" water service provision needs to bring together tankers from both main religious groups to prevent existing tensions and competition for water resources from escalating.It is also important that water tankers with businesses of different sizes are involved.Otherwise, this approach might result in a monopoly by larger water tanker companies that already control the market, as they might be the only ones able to engage and negotiate with public institutions.Also they might be the only ones able to afford formalization's added costs from registration fees, licenses and taxes (Ahlers et al. 2014).Smaller tankers would run out of clients, potentially turning to other criminal activities.
Around 30% to 60% of the urban population in the global south relies on informal water systems (Ahlers et al. 2014).Moreover, some communities in higher income areas also rely on informal sources including bottles and water trucks (London et al. 2021;Balazs and Ray, 2014;Jepson and Vandewalle 2016).Hence, there is an opportunity for future studies to analyze the social network of informal water tankers in other locations and identify and compare drivers of informality across borders to better understand their impact on water systems and better inform policy makers' strategies to co-opt, sanction or regulate informal water systems.

Notes
1.As per Internal Research Board agreement all tanker names are coded for anonymity purposes.IRB # 1292073-1.2. There are generally 17 recongnized sub-religious sects in Lebanon (e.g.Catholics, Orthodox, Sunni, Chia etc … ) (Prados 2006).For simplicity of analysis religion is considered as two general types as Christian and Muslim.This also mirrors the tankers' perceptions of religious affiliation.3. We abide by the Internal Research Board agreement of leaving sensitive data out of the questionnaire including not asking interviewees about their religious background.IRB # 1292073-1.4. Tankers may have simultaneous cooperative and competitive relationships, or relationships might change over time.However, we recoded the dominant type of relationship at the time of the interview.
paper.The authors are grateful to the four reviewers for their careful and genuine engagement with this research; their comments have greatly improved the paper.

Figure 1 .
Figure 1.Cooperative and competitive networks of informal water tankers and their religious diversity.

Table 2 .
Exponential random graph models results for the cooperative network.