|
|
|
|
Responses to Prediction and the Middle East Peace Process
Steven Weber, UC Berkeley
The Mershon Center
February 1997
The election of Benjamin Netanyahu as Prime Minister of Israel threw into flux the Middle East peace process. What in early 1996 seemed to many observers the almost inevitable "working out" of a decades-long conflict that had gradually become unsustainable on all sides, was by late 1996 seen more clearly as part of a contingent unfolding set of events which could drive the region in more than one direction, including backwards toward explicit conflict and even war. This presents unique theoretical, analytic, and policy opportunities.
This project brings together an interdisciplinary group of scholars to develop scenarios for the future of the Middle East peace process. In a theoretically disciplined fashion, we will try to identify the most important and most uncertain driving forces operating in this situation, and use these to generate a set of conditional forecasts for the region. These forecasts will be updated and critiqued at regular intervals -- approximately every 6 months, for a period of 3 years. We intend to report the process on a regular basis in Security Studies.
IR theorists generally seek to understand what are the most critical driving forces behind events that have complex conjunctural causality. But they usually do so in a post-dictive setting, when the outcome is known. In contrast, we intend to go on the record in a predictive fashion -- not only with bald statements about outcomes, but with clear causal hypotheses tied to broader theoretical foundations. We intend to explore whether and how formal methods of several kinds add value to predictions based on "informal" theory-based argument. We hope also that this project will prove valuable to policy makers, by alerting them to possibilities they may not have considered, and to connections between causes and feasible points of intervention that were not so clear to start.
Project Objectives
The return of the Likud Party to power in Israel in May 1996, through a hotly contested election in which Netanyahu won by less than 1% of the vote, brought dramatic shifts in the immediate stance of the Israeli government toward its Arab neighbors and the on-going peace process in the Middle East. It also provoked discussion about some basic theoretical arguments in politics and history. How do theorists (and policy makers) understand the causal significance of this shift? Some perspectives view the individual decision maker as a key cause. Other perspectives stress more strongly particular aspects of the environment within which decision makers operate. Domestic political constraints, economic constraints, or international constraints delimit and condition strongly the set of choices open to any decision maker and government.
For each of these perspectives there is an intriguing methodological challenge. Is it possible to develop predictive arguments of some kind, that go on the record before events have transpired? Or are the events in question subject to such complex causal conditions that theory is useful only for post-diction, or explanation of events that have already occurred. These questions overlap with recent work on counterfactual reasoning in world politics and particularly with important questions that arose in that work about the logical differences -- if there are any -- between forward and backward reasoning in time. It is also relevant to psychological studies of how people actually do think predictively and whether social science is better at that task than "layman reasoning". In practical terms, is prediction and error-correction is an effective means of encouraging learning from history? Finally there is the question of whether formal models add value to predictive arguments in world politics, and if they do so in a way that is similar to or different from what they do in the explanation of past historical events.
Predictive power may or may not be a necessary component of social science theory. But the process of predictive testing can still be a useful way to re-insert a sensible notion of contingency into theoretical arguments in international relations which have tended to take on a deterministic flavor. Parsimony is surely desirable. The field of international relations tends to privilege arguments that reach back into past cases and parse out one, or at most two, independent variables that are then stressed as the major driving forces behind events. But since the field generally favors independent variables that are "structural" or otherwise parametric, the endeavor has a built-in bias toward over-deterministic explanations. This tendency is quite strong in respect to the Arab-Israeli conflict. What seemed almost unthinkable before it happened -- the Sadat trip to Jerusalem, or the Peres-Arafat moves toward peace -- are in retrospect said to be almost over-determined by big structural causes such as the loss of a major war or the collapse of the Soviet Union.
Yet the same theorists who feel comfortable with this perspective on the past would likely bristle at the thought that the future was equally deterministic. We hope that forcing theorists to project their arguments into future - time will also force them to re-evaluate their beliefs about what they think they know about causality in the past. What role does and should theory play in that process -- is it primarily a discipliner and organizer of complexity, or does it widen the scope of causes we consider and alert us to unforeseen possibilities (or some combination). This is an experiment that seeks to incorporate elements of contingency in a manner that will not be paralyzing to social science debate, but will open that debate to be more acceptable toward more complex and multi-variable explanations with a continuing sense of humility about how much we really have (and can have) explained.
For this reason, this project is not a forecasting "tournament" (analogous to Axelrod's Prisoner's Dilemma tournament) where advocates of different theoretical perspectives are asked essentially to generate differential predictions on a single dependent variable, in the hope of later finding "winners and losers". Instead, we build on the understanding that a degree of consensus is possible (although not necessary to our objectives). Different theoretical traditions do not and need not share the same dependent variables -- just as they do not share the same independent variables. For example, structural-realists might generate arguments about broad outcomes of the peace process and the future shape of the region, while having less to say about Israeli foreign policy in particular. A perspective stressing religion as a driving force might say more about the evolving preferences and policies of some major actors (not all of whom may be states). The payoff for the project comes in seeing how and when these different perspectives succeed in identifying key driving forces that contribute to outcomes of interest, and how and when dependent variables can be combined in complementary ways. We focus on the peace process, while recognizing that this is a function of strategic interaction between a set of major actors within and outside the region. The project seeks to answer a question like "does state power over-ride religious preferences and cultural predilections?" in a more subtle way than IR theorists typically do, by allowing proponents of different perspective to fully develop their arguments using their preferred independent and dependent variables -- and cooperate with each other in evaluating evidence as it comes in.
The Middle East is a particularly interesting and important place to carry out this project. Present in the region are strong elements of structural, parametric variables with which IR theory is relatively comfortable: power, superpower influence, even nuclear weapons. But there are also strong elements of many variables that IR theory is less comfortable handling: ideology, religion, hatred, and the powerful influence of sub-state actors capable of violence. Indeed, the notion of "statehood" is not a given in the Middle East -- which means that a priori assumptions about the efficacy of the "Westphalian" solution may be deeply problematic. Perhaps as a result, the Middle East is also a region where there seem to be clear and strong elements of historical and local contingency. Policy makers and theorists alike seem to share the feeling (which may or may not be justified) that "small" or "chance" events can have a very large impact on the process or the outcome: the assassination of a political leader; razor-thin election victories; back-to-back terrorist bombings which change the public mood; accidental crashes of helicopters; and so on. From a psychological and decision making perspective, contingency can be used in peculiar ways that complicate theoretical advancement. It seems that analysts and policy makers often use "contingency" to cover up the fact that we mispredict events; pay too much attention to some things and not enough to others; tend to over-compensate for that bias in retrospect; and at the same time often forget what we thought at the time we made the prediction, and why we thought it. Clearly it is important to counteract these tendencies if we are to improve our understanding of our own theoretically driven thought processes, and of events in the world around us. This is a major justification for creating a written record of the process of scenario construction and revision.
Prediction, whether explicit or not, is always an important part of policy making. Major actors (including but not limited to the US) are engaged in significant efforts to influence the peace process, but often do so without a clearly articulated and consistent understanding of what variables matter and how a particular intervention will operate. At a minimum, this project will alert policy makers to the implicit arguments with which they are acting and illustrate in a constructive way some of the key uncertainties. By providing alternative views and some unexpected possibilities, the project should challenge the "official future" that decision makers frequently carry around with them and use as the basis for making and evaluating policy. Generating early indicators of important causal processes is a particularly useful means of alerting policy makers to unfamiliar forces and trends that may be operating, and forcing these issues up the attention threshold earlier than otherwise might have been the case -- possibly avoiding detrimental surprises.
II. The Method
The project brings together an interdisciplinary group of scholars, whose task will be to develop a set of scenarios for the possible future(s) of the Middle East peace process. The group includes general IR theorists, Middle East experts, decision-making specialists, modelers, and others. (Attached is a list of participants in the first meeting). We plan to expand the group as necessary to tap expertise in areas that come up as important causal forces that we have not yet identified. We also plan to invite policy makers to future meetings where they would be asked to evaluate and critique the scenarios, as well as help us to see what the group of scholars has produced that may surprise them or change their views of what is important and/or possible.
The "scenario methodology" that underlies this project has three main objectives. The first is to draw out fundamental theoretically-driven assumptions about the major driving forces that impinge on events. The second is to develop and analyze interaction possibilities among those driving forces. The third is to organize that complexity into a small or moderate number of possible future scenarios.
Scenarios then act as impressionistic pictures that combine in different combinations causal variables at different values. Scenarios are not, and are not intended to be "point predictions" of specific events. Rather, scenarios make the following kinds of claims: if the political economy of Israeli labor needs turns out to be a critical driving force, then we should start to see more of "X" kinds of things happening, and less of "Y" kinds of things. The creative part of the exercise comes in agreeing on what "X" and "Y" are, as well as on working out the means by which to evaluate the predictive "results" as future history becomes data.
The scenario methodology has 7 steps: identify driving forces, specify predetermined elements, identify critical uncertainties, develop scenarios with clear "plot lines", extract early indicators for each scenario, consider the implications of each scenario, and develop "wild cards" that are not integral to major possibilities but could change the situation dramatically if they were to happen.
Driving forces are the causal elements that surround a problem, event, or decision. While some driving forces are likely to derive from standard independent variables in major social science theories, others are not -- for example, the importance of religion in the middle east region. It is common in developing explanations for past events to try to limit candidate causes or driving forces to just a few, or perhaps even two. In generating scenarios for the future, we will put on the table a longer list to start, and then whittle down by re-organizing and grouping driving forces that belong together, as well as dropping (but only on a contingent basis) forces that participants agree will be less important causes.
Predetermined elements are things that are relatively certain, parameters that can safely be assumed for the scope and span of the scenario exercise. The point is to separate out what is certain or very close to it, from what people simply think or believe, without falling prey to well psychological tendencies to treat routine events, "causes" of "effects", background actors, sins of omission, and "structural" (not human agency) causes as immutable. There are no easy experiments and control situations in world politics, but that does not mean that there do not exist things that are relatively certain. Four possible examples of the kinds of things that can be pre-determined within reasonable bounds in the Middle East are:
Critical uncertainties develop in seeking out what is most uncertain and most important to a particular decision or to understanding a set of events. Critical uncertainties may derive from a deeper understanding of driving forces, or may be closely related to pre-determined elements particularly where those elements create problems that must somehow be resolved. Scenarios place the critical uncertainties out in front, ahead of the plot lines or connecting principles that pull the story together. Standard social science theory "testing" treats as mutable the "independent variables" suggested by connecting principles that we already know well. In scenaric thinking, plot lines have to work with the critical uncertainties rather than the other way around.
A scenario plot line is a story about how things happen. It describes how driving forces might plausibly behave as they interact with predetermined elements and different combinations of critical uncertainties. Plots have their own logic (or sometimes more than one logic) that drive the story forward and suggest the directions in which uncertainties resolve. The logics may be familiar elaborations of standard IR theories. For example, balance of power theory emphasizes the way in which a strong driving force (states' desire for independence and autonomy) interacts with predetermined elements (power configurations) and critical uncertainties (who will ally with whom) in an international system to produce outcomes. Or new logics may emerge. The advantage of the scenario method is that equifinality, multifinality, and complex conjunctural causation are no longer stubborn inconveniences that need to be minimized or simply ignored. Instead they are treated as natural and fundamental aspects of reality. Unlike the ideal scientific notion of an explanation of the past, there will be several scenarios that result, as a way of capturing the probabilistic nature of the arguments at play.
Early indicators link scenarios to "observable implications" that help a data collector or a participant evaluate, as history unfolds, which scenario (or which parts of scenarios) may be coming to pass. Developing early indicators is an exercise in "process-tracing", extrapolated into the future. Participants will ask, if a particular set of driving forces were to become most important and lead to a given scenario, what would be some of the early indications that events were indeed unfolding along that particular path and not along another? The strategy is a modified version of the simple idea of increasing the number of observables that differentiate between one set of explanations and another. By doing so in future time, we reduce post-hoc determinism and force ourselves to confront in a creative way historical contingency.
Implications of scenarios are aimed explicitly at decision-making and action. One of the valuable consequences of thinking about historical contingency in a disciplined way, is that it forces people who are going to make decisions to ask what they would do if they found themselves in a world different from the one they expect. They might also recognize the extent to which their own actions could be an important pivot or determinant of what kinds of futures were likely to evolve. Considering at once the behavioral implications of more than one scenario helps to clarify the stakes, risks, and uncertainties connected with any single course of action that an individual or a state might choose.
Finally, participants will develop a list of wild cards. These would be conceivable (if low probability) events or actions that might undermine or modify radically the presumed course of events. Some wild cards could constitute extreme values on a familiar independent variable; others might be entirely outside of the realm of standard social science arguments. In either case, doing this prospectively could change our views on what variables should be a part of theory, or what an "extreme" value really is -- since it avoids the possibility of post-hoc certainty. It would also be revealing if we were to miss entirely a wild card - type cause, or if wild cards happened but were "damped out" in their effects by other kinds of causes.
III. The Process
This is the first of a series of "research notes", which the project will produce after each meeting. We will publish these as an on-going record of the process, its findings, and its self-critique. The idea here is to provide scholars with a documented history that includes not only what succeeds, but also modified arguments, falsified hypotheses, failed predictions, revised understandings of the importance of driving forces, and most importantly a sense of what and how we learn in our efforts.
The group held its first meeting at the Mershon Center of Ohio State University in February 1997. Developing a joint understanding of how we would configure the dependent variable, exactly what it was we would try to predict, was the first problem. We began with the idea that we would use Israeli foreign policy as a window into this substantial transformation in international conflict and cooperation, with the outcome of the Middle East Peace Process as the primary dependent variable. But this strategy seemed flawed: while Israeli policy appears in early 1997 to be an important pivot, it is possible that this is simply a reflection of the particular time at which we begin the project. We chose to drop this idea and to focus, for the sake of tractability, on the Israeli-Palestinian settlement as the main dependent variable for this phase of the project.
From a methodological standpoint, this does not mean ignoring the complexity of other elements of the peace process. It simply is a decision to organize that complexity around this particular focal point. Three underlying assumptions (if they are correct) nearly guarantee some movement in this aspect of the peace process over the relevant time frame: the status quo is unstable or at least very uncomfortable for the major actors; the set of choices available are unpleasant and will require trade-offs -- there is no obvious grand bargain at hand; and preferences remain fluid and subject to change as the process unfolds.
If the future will indeed be different from the past and the present, is there a consensus on what it will look like -- an "official future" around which most expectations do indeed converge? It seemed so, at least when we began. While there was some disagreement over timing, there emerged a broad consensus on what would be a stable outcome, and was indeed the most likely (even nearly inevitable) outcome. Compromise between moderate factions on both sides would lead to an exchange of territory and a "two state solution". A Palestinian state would come into being with restricted sovereignty. It would gain membership in the UN and other international organizations, but would be limited in its foreign policy autonomy and security/military capabilities. Radicals on both sides would continue to oppose this arrangement and would test the political will to sustain it by violent action. Particularly as most Israeli settlements would remain in place, there would be plenty of opportunity for such testing.
This consensus was remarkable in its simplicity and in the level of implicit confidence that most participants felt in the prediction. Yet as we explored in greater depth this scenario, we found that it was actually more highly contingent and dependent on a wide variety of conditions, What major forces would have to be in place for the official future to arrive? A first cut at this question led to a list of conditions that were probably necessary but almost certainly not sufficient:
It was not necessary to go further at this point. Our confidence estimates in the likelihood of the official future had been reduced substantially, which opened up the discussion to the level of contingency in this process and how to handle it analytically. We wondered, as a next step, if it would be possible to organize arguments by varying some of these conditions and predicting their effect on the process. For example, assume a major shift in the balance of power relations in the region (Syria is able to bring Iraq back in to the picture as a major actor; or there is a political collapse in Egypt). If Palestinian leverage were to be increased by these kinds of events, Arafat might not be constrained to accept the "constrained statehood" deal and might press for more. What impact would this have on the process? We found that we could not agree on this point, not even on the ‘sign’ of the effect. It is possible that the process could collapse into confrontation and end up in a more violent version of the status quo. Alternatively, Israel might settle quickly. Or a clash of hardened bargaining positions (on both sides) could lead to sustained violence, economic decline in the occupied and PNA controlled territories, social anomie in one or both nations; and even the threat in the medium term of civil war.
Clearly, there were a range of possible Israeli-Palestinian relationships that could emerge in this process -- from full confrontation, to low-level violence, to co-existence with separation, and even to federation. We decided at this stage to add two additional "outcomes" as possible futures.
In addition to the two-state solution, we could foresee an outcome we called "status quo plus". There would be only one state (Israel). The PNA would gain some additional local autonomy, but not very much territory. Lacking international status, it would not be a member of international organizations and would depend almost entirely on Israel and Jordan for political and economic relations with the rest of the world. Israel would retain substantial residual security rights and military presence in PNA controlled areas. There would be continuing low-level violence, more frequent and destructive than in the two state solution.
A third outcome would be "violent collapse". Here the peace process would come to a halt and violence would be even more widespread. Israel would attempt to roll back concessions granted to the PNA. Arafat would be challenged, probably with success, by more radical leaders and factions including Hamas. Borders would be closed and fortified with both sides hunkering down into a fortress mentality, trying to protect what they had rather than move forward. The West Bank settlements would remain major flash points. A new intifada would likely emerge.
The next step was to list some of what we thought were the major driving forces that would push the process toward one or another of these three possible outcomes. To start, we listed the following:
Paralleling our earlier effort to agree on the direction of effects, we tried to examine these driving forces to estimate vectors -- in what direction, and with what strength they would push the process. Again, we found substantial disagreement and a high level of contingency. In large part this was because many of the driving forces were subject to complex conjunctural causality -- meaning that individual vectors were often dependent on the state of one or several other variables. This proved very difficult to sort out, which led to a discussion about trying to develop formal computational models that could assist in the process.
In the interim, we made several efforts to systematize this indeterminacy by organizing around 2 or 3 "pivot" variables which would drive the process. This proved extremely difficult to do on an a priori basis. We did come to a tentative agreement on two points. First, that there were at least 3 major pivot variables. These were: political balance between pragmatic and radical factions on each side; stability level of major regimes in the region; power ratio between Israel and the Palestinians. Second, that it would be possible to array our three scenario outcomes in a triangular chart where the key pivot variables could be seen as driving the process along the x or y axis. The chart follows:
We were unable to specify and agree on the direction in which changes in major independent variables would drive the process along this chart. The indeterminacy of theory and the problem of complex conjunctural causation were the major reasons for this -- which means the scenario methodology we had tried to follow could not produce the results we wanted. Given this problem, we decided to go back and approach the analytic process from the other side, by focusing first on driving forces rather than outcomes per se. There would be three steps. First, identify a set of driving forces. Second, trace each "cause" through Israeli and Palestinian decision making, with particular attention to the impact of change in the cause, on the balance between center/pragmatic and radical forces on each side (this becomes a key intervening variable). Finally, trace the process to the dependent variable -- movement toward or away from two-states, status-quo plus, or violent collapse. We agreed to consider 7 driving forces to start, grouped into three categories. (This would be a preliminary not a final list and all participants would be free to suggest additional or replacement driving forces at the next meeting). They are: I. Global Forces A. Global Economy (protection vs. continued liberalization; specialization of markets; price of oil; business cycles as possible important factors) B. US Engagement in the Peace Process -- both priority and strategy of US in trying to shape incentives for the local actors. II. Regional Forces A. Major Regional States -- stability of regimes, and of foreign policy orientation. B. Governing Elite in Israel and PNA -- stability, turnover, radicalization. C. Israel-Syria -- progress, conflict, stalemate III. Domestic Political Forces A and B. Israeli and Palestinian ability to implement major elements of Oslo II and beyond. Major elements -- For Israel: release of prisoners, airport/seaport, connecting road, withdrawal from West Bank. For PNA: close offices in Jerusalem, maintain security, amend charter. Each participant will return to the next meeting with arguments on these seven driving forces. These arguments will include:
Autonomy MInues
CENTER HOLDSTwo States
\ /
\ /
\ /
\ /
EXTREMES DOMINATE
Violent Collapse
This will accomplish several of our preliminary objectives. It will force participants to make explicit major assumptions and in doing so will begin to unpack the theoretical reasoning behind major arguments. It will help to illustrate and to make more precise the areas of agreement and disagreement. How much variance is there among participant’s arguments, and from what does that variance arise? It will generate a data substrate for efforts to develop computational methods of combining predictive estimates, which Phil Tetlock and Don Sylvan will be constructing in the meantime. And it will serve as the primary input for the next meeting, to be held in August, at which we will try to devise in more detail the three scenarios by characterizing more closely what they would look like, and how the process could get from its present state to each outcome.
Email this citation