Gaming the system in the Caspian Sea: Can game theory solve a decades-old dispute?
Kathleen Cantner, AGI
Copyright Shutterstock.com/Orkhan Aslanov
Water-rights disputes are never easy. Whether they are over pumping from aquifers shared among adjacent landowners, allocation of resources at the municipal, county or state level, or division of a river or lake shared between neighboring countries, there is an inevitable push and pull among stakeholders over who gets what and how much. In the end, resolutions tend to be compromises — often meted out by governing authorities — which, while not ideal from any one party’s standpoint, appeal to the desire of the group for stability over strife.
Not surprisingly, however, disputes become more complex as the size and perceived importance of the resource — as well as the number of parties involved — scale up, making resolution harder to achieve. Throw in historical tensions and questionable precedents, imbalances of power among the parties, the lack of an outside body to broker a deal, and an international stage replete with regional and global political interests, and resolution becomes significantly more difficult still. And to all that, add confusion over what sort of water body is even being considered — thus muddying if and what international standards might apply — and the fact that the discussion is less about water itself and more about ownership of large oil and natural gas reserves, and you have the makings of a longstanding and stubborn conflict.
Welcome to the Caspian Sea.
Few if any international water-rights disputes have proven as unmanageable as that over the Caspian, a massive body that stretches over hundreds of thousands of square kilometers in Central Asia. The modern conflict dates to the breakup of the former Soviet Union in December 1991, which abruptly increased the number of littoral countries bordering the sea from two — Iran and the Soviet Union — to five, each with its own economic and geopolitical interests. Newly independent and suddenly without Moscow’s economic backing, Azerbaijan, Kazakhstan and Turkmenistan joined Iran and Russia at the table, dismissing prior agreements between the latter two nations that had long governed the Caspian and asserting their own claims to its resources.
A number of proposals to divvy up control of both the Caspian surface and seabed have been floated by the five countries since negotiations began in 1993. None, however, has been unanimously accepted, despite frequent and ongoing talks. This lack of agreement has left the sea’s legal status and governance in question for two decades and has set the stage for occasional unilateral moves — resource grabs and exploration and export deals struck with outside companies, for example — as well as various bilateral treaties not recognized by all of the other players.
The conflict is a unique multinational mash-up of economic, political, energy resource and environmental concerns that has attracted the attention of researchers outside the usual crowd of policy analysts and political scientists. Looking to advance the discussion past the current stalemate, some researchers have taken novel approaches to modeling the conflict based on the methods of game theory and related ideas not conventionally applied to such disputes.
Such methods have long been used by economists and sociologists to evaluate, for example, different bargaining strategies or voting patterns based on the likely motivations of voters. But the methods have only found limited use in analyses of water- and energy-rights disputes — most often being applied to questions of how to fairly apportion groundwater or river flows. It’s too early to tell if insights gained from such theoretical modeling will help resolve the Caspian conflict, but at the very least, the researchers suggest, the work will help illuminate each country’s underlying motivations and clarify potential outcomes of a very muddled situation.
What Is Game Theory?
“Game theory is a mathematical study of competition and cooperation” — the study of strategic decision-making, says Kaveh Madani, a civil engineer and founder of the Hydro-Environmental and Energy Systems Analysis group at the University of Central Florida, who has been modeling the Caspian Sea dispute since 2008. With game theory, researchers try to see how the interactions of individual behaviors result in the overall behavior of the system of which they are part, he says. The idea is to include the distinct viewpoints that individual stakeholders may have regarding financial, physical or other considerations, to “respect individual opinions, individuals’ [ideas of] utility,” Madani says. That differs from the traditional top-down systems-engineering approaches that model systems as monolithic problems, without accounting for individual preferences or for which solutions can be found depending on prescribed optimal outcomes.
An example, Madani says, is to consider the decisions the U.S. government might make to improve the economy versus the decisions individuals might make. The government is interested in maximizing the gross domestic product (GDP), but at the individual level, people don’t care about that as much, he says. Instead, individuals tend to be more interested in having a job that pays well and a nice place to live, he notes. So, although modeling GDP maximization strategies from the government’s top-down perspective might be informative, it’s less realistic than modeling outcomes when the actions of individuals are taken into account.
Under the headings of game theory and the related methods of bargaining, social choice (voting) and multicriteria decision-making, numerous scenarios and techniques can be applied depending on the characteristics of the problem to be analyzed. At their most straightforward, these techniques are primarily “descriptive” — using known information to allow predictions of likely outcomes of a dispute or other competitive situation based on the expected motivations and behaviors of stakeholders — and typically consider parties that are acting purely in their own interest (or noncooperatively). In addition, though, the techniques can also be “prescriptive” or “normative,” seeking, respectively, to advise either a single party in a competition, or informing all parties of “socially optimal” outcomes that benefit the group as a whole and result from cooperation.
One of the most common game theoretic scenarios for framing many water-rights disputes — involving groundwater pumping, for example — is called the “prisoner’s dilemma.” In the archetypal case, two prisoners are held without being allowed to communicate with one other for a crime for which the police do not have enough independent evidence to convict either one. Each prisoner can either confess and offer evidence against the other, for which the confessor is promised a reduced sentence, or he or she can remain loyal to his or her partner in crime and not confess (in which case the outcome is unknown to the prisoner).
Thus, there are four distinct outcomes: one confesses while the other does not (and vice versa), both confess or both remain silent. In the first two cases, the confessor is immediately released for cooperating while the silent prisoner is given a long sentence. If both confess, both go to prison but for a shorter length of time. Lastly, unbeknownst to the prisoners, if they both remain silent, each is released after only a brief time in jail due to the lack of evidence needed for a conviction.
Given the incentive of a reduced sentence, a lack of trust between the two, and the fact that people tend to make decisions based on what’s best for themselves (noncooperatively), the most likely outcome is that both prisoners confess. Were they clued into the lack of evidence against them and willing to cooperate with each other (remain silent), however, their collective outcome would be optimized.
In a groundwater-sharing dispute that plays out like the prisoner’s dilemma, the most likely outcome — again, given that people are primarily self-interested and may not trust each other — may be that each party ignores the socially optimal solution in favor of withdrawing the maximum amount of water he or she can. If all parties engage in this behavior, the result may be the worst possible option, exhaustion of the groundwater resource.
“Cooperative game theory solutions, or methods, essentially try to tell you what is fair and what is efficient when you want to share the benefits of cooperation,” Madani says. The idea, he says, is to move discussions over disputed resources or territorial claims beyond the zero-sum game where “my loss is your gain, and I never want to lose to you.” This is easier said than done because everybody comes to the table with their own narratives of a problem and with the direction they want negotiations to go, he says. “But is there any possibility to develop a narrative that is fair, that doesn’t take any side, and is impartial, realistic and acceptable? That’s what we are trying to do.”
Of course, Madani notes, the more stakeholders or participants are involved, the more complicated the situation becomes.
Clamoring Over the Caspian
The presence of petroleum in the Caspian region has been recognized and exploited for centuries, though most of it was pumped from onshore deposits rather than from beneath the Caspian. Historically, the value of the sea itself was as an important fishery, famously producing more than 90 percent of the global supply of sought-after sturgeon caviar, for example.
Prior to 1991, water rights on the Caspian were governed by a pair of treaties between the Soviet Union and Iran signed in the first half of the 20th century. Beyond 18.5 kilometers (10 nautical miles) from the shoreline, which demarcated the boundary of each country’s territorial waters, the treaties provided for joint ownership of fishing and transit rights to the Caspian. “Essentially, because oil and gas were not an issue of concern, relations were really good,” Madani says.
Following the dissolution of the Soviet Union, however, increased hydrocarbon exploration and production starting in the early to mid-1990s revealed tremendous amounts of both oil and gas. Proven reserves — many of them offshore in the Caspian Sea — are thought to account for roughly 4 percent of the world’s total, according to a 2006 Congressional Research Service report based on data from the U.S. Energy Information Administration. And the potential additional resources are even greater.
“Now things are different,” Madani says. On one hand, “we have three new independent countries with poor economies, and oil and gas revenue makes a big difference for them.” On the other hand are Russia and Iran, which are less dependent on the Caspian Sea for energy but hold most of the power in the region.
Ownership of the Caspian is “sort of a puzzle box that got opened when the Soviet Union collapsed,” and the outcome means a lot to a lot of people, says Marc Kilgour, an applied mathematician at Canada’s Wilfrid Laurier University in Waterloo, Ontario, who, like Madani has looked at possible resolutions to the dispute using social choice and game theory methods.
“It’s really the multinational nature that makes the Caspian interesting,” Kilgour says. “There are relatively few cases in the world where you have a multinational lake or sea [surrounded by more than two countries], except for some really big seas like the Mediterranean.”
In fact, part of the problem in the negotiations has been that it’s unclear whether the Caspian Sea, despite its name, should be treated as a sea, or as a lake. As an enclosed inland water body, it might seem to qualify as the latter. But, it is by far the single largest such water body in the world — dwarfing the next largest (by surface area), Lake Superior, which is 4.5 times smaller — and has characteristics (such as high salinity) that are more akin to other seas.
More than simply a semantic consideration, the distinction might establish which international precedents to follow. As a lake, the Caspian could be treated under a so-called “condominium regime” in which all five countries would jointly administer the entirety of the lake, and resources and revenues would be distributed equally. On the other hand, as a sea, the Caspian could be divided, according to the United Nations Convention on the Law of the Sea, along lines of equidistance extending from international borders, thereby granting each country exclusive economic and governance rights to its adjacent maritime sector.
Since negotiations among the five littoral countries began in 1993, five main strategies for either shared governance or division of the Caspian have been proposed, with the condominium regime and division based on the UN Convention as the initial two. The other three proposals include an equal areal division under which each country would have exclusive control over 20 percent of the sea surface and seabed; division based on Soviet-era maps that delineated seaward boundaries of each of the Soviet republics surrounding the Caspian; and a mixed approach providing for shared governance of the sea surface but division of the seafloor based on international law, thus allowing shared access to fishing resources but exclusive control over energy sources.
Modeling the Negotiations
The negotiations made for an interesting problem to model, Kilgour says, in part because the possible agreements the countries could make were laid out pretty well, and because there is an agreed-upon mandate that the governance option eventually chosen must be unanimously accepted by all five countries. But each option has different implications for each country that aren’t aligned with the others, Kilgour notes, so each country has had more- and less-favored options, largely based on potential economic or political benefits resulting from access to oil and gas reserves.
For example, Iran has long advocated either the condominium or equal division approach due to its relatively small share of the coastline, whereas Azerbaijan, Kazakhstan and Turkmenistan — which have relatively long Caspian coastlines and are heavily dependent on Caspian energy for revenue — have preferred sectorial division and administration of their own shares. Russia, which has abundant sources of fossil fuel elsewhere and relatively small reserves of oil and gas off its Caspian shores, likely has the least to gain or lose economically. Thus, it has seemingly shown the most flexibility in negotiations, perhaps preferring to maintain regional influence through deal-making rather than focus on direct economic gain from the sea.
The number of players involved and the fact that each has different preferences based on their individual rationality combines to make for complex negotiations. Add in external powers like the U.S., China, Turkey and the European Union, which also have interests and influence in the region, and it all makes for a very complicated problem to model.
Thus, to start out with, the models that Kilgour, Madani and their colleagues employed to study the Caspian Sea conflict were fairly straightforward. “Early on, we were just trying to look at standard ways of [how] this kind of cake division problem” might be resolved, Kilgour says. “We tried to be as descriptive as possible [in our modeling]. That is, we actually tried to look at the situation on the ground, or under the ground.”
The researchers assumed each country had its own interests in mind (as in the noncooperative case of the prisoner’s dilemma scenario), and a particular order of preference of the proposed plans, which the researchers assigned based on the public statements made by each over the course of past negotiations. The goal was to predict the likely outcome of the negotiations if the countries were to bargain either over which of the five governance and division methods to use — with each “falling back” to successively less-preferred options as needed until a mutually preferable option was found — or to essentially vote on a consensus, or “socially optimal,” resolution.
Although there were consensus options that emerged as the most popular (or, more aptly perhaps, the least unpopular) — division based on Soviet-era maps, for example — the researchers recognized that, because none of the options inspired unanimous agreement, none were likely to be viable.
Still, there were important findings to come out of the early descriptive models, Madani says. For one thing, it’s really important to consider the utilities [the monetary value of the oil and gas resources] of these countries from the division methods,” he says, because this reveals relative differences in preferences between different scenarios for each country, rather than simply an order of preference. This point was reinforced through computer modeling using the Caspian Sea Negotiation Support System (NSS; see sidebar), a tool Madani and his colleagues developed to determine optimal nautical boundaries based on different division scenarios.
Considering the monetary value of the resources got researchers thinking about modeling the negotiations from a cooperative standpoint in their ensuing efforts. What if the countries moved past the basic division schemes that they’ve debated for two decades and worked together to find a solution that would benefit each country more than the current status quo stalemate? The status quo, Madani says, has been based on the old treaties between Iran and the Soviet Union, essentially limiting exploitation of Caspian energy resources by each country to those located close to their own shores. But coming to a unanimous governance agreement would open up the entire Caspian to exploration and development, vastly increasing the potential economic windfall for all five nations.
Enter cooperative game theory. Using the monetary value of each country’s resources under the status quo as a starting point (estimated with the Caspian Sea NSS), as well as the estimated value of the total oil and gas resource in the Caspian (based on both proven and potential reserves), the researchers used game theoretic strategies — going by names like the Nash-Harsanyi, Nucleolus and Shapley methods, for example — to treat the distribution of wealth as a mathematical problem that could be solved with distinct “optimal” solutions. They also used standard methods for dividing funds in bankruptcy situations — where various parties are owed amounts from a single pot that, in sum, surpass the total value of the pot — such as apportioning amounts from the total to each party proportionally based on the stated claim of each.
The results of these studies showed that each country indeed stood to gain substantially from such cooperative approaches. So far, the solutions to each of the game theory strategies employed, as well as some of the bankruptcy methods, have converged around what amounts to an almost equal sharing of the total economic value of the oil and gas resources as the socially optimal and potentially stable resolution. Such a resolution would not be far off from the condominium approach that imagines equally shared revenues and that has been on the table since the early 1990s.
What the results suggest is that if the countries are willing to cooperate, “the most efficient and fair solution, when you consider individuals, is the condominium regime,” Madani says. “This is exciting” and a “good sign,” he says, because it shows there is some overlap between the individual rationalities of the countries and the ideal rationale of the group as a whole.
The trick, perhaps, is getting the countries to recognize and accept this group rationale. Even considering the so-called optimal solution, in which everyone is profiting more, there may still be winners and losers to an extent. Imagine, Madani suggests, two companies — a smaller one that makes $1 million a year in profit on its own and a bigger one that makes $2 million on its own — that figure they can make $10 million a year if they work together. Should that $10 million then be split evenly between both companies? It’s still more profit for each, but the gain is greater for the smaller company. So maybe the companies “never cooperate simply because they cannot agree over how to share the benefits,” he says.
From their game theory calculations, Madani and his colleagues found that the total value of Azerbaijan’s and Russia’s shares of oil and gas would increase more than 100-fold over their status quo values, while the others would see increases between 10- and 20-fold. (This difference has to do in part with the fact that Iran, Kazakhstan and Turkmenistan have far larger amounts of near-shore oil and gas reserves in the Caspian that were included in their status quo values than do Azerbaijan and Russia. In reality, the disparity would perhaps not be quite so large because, for example, Azerbaijan is already allowing active development of reserves farther offshore than was accounted for in the calculations.) “The concern is about how to fairly and efficiently share the gains of cooperation,” Madani says. “Many times, we know we can cooperate and gain more, but the problem is that we cannot agree on how to share the benefits.”
There are other complicating factors as well. In order to best model the negotiations, other considerations — such as environmental resources like fish stocks in different parts of the Caspian (the value of which, however, pale in comparison to the energy resources; see sidebar, below), or transportation and shipping concerns — should be included. And adding in political leverage is also important, both Kilgour and Madani say, even though it’s a difficult task given the inherent subjectivity in estimating such power. Kilgour notes that researchers have so far only considered “local leverage” — basing each country’s political weight on factors such as economic independence, military strength and financial support from other countries, among other factors — “and not geopolitical implications of the resolution.” Preliminary results that include political weighting factors have shown variously that division based either on the UN Convention or on a condominium approach may be the likely outcome.
Unfortunately, it’s not possible to model the full complexity of the system, Madani notes. “No model, no study can include all those parameters or variables,” he says, which is one reason it’s important to study the case using a variety of scenarios and methods. Overall, with their modeling efforts, he says they’ve “tried to essentially focus on the governance issue and to find out what is fair and what is feasible in practice.”
Realistically, if there is resolution to be found among the Caspian countries, both Madani and Kilgour suggest it will probably come down largely to politics and who has the leverage to push it past the deadlock, rather than purely economic issues. It’s a “classic prisoner’s dilemma … problem that we have in natural resources,” Madani says, and in the end, the countries will probably “end up with a resolution that is not socially optimal, but it’s something that is doable” given the politics.