STRATEGICALLY HOPEFUL ACTION

By strategically hopeful action we mean a certain kind of calculated risk-taking in the face of imponderably complex circumstances, the aim of which is to replace a destructive status quo with a new and better dispensation. The paradoxical overtones of “strategically hopeful” are intentional: The phrase blends aspects of political action that are typically kept separate. The action must be strategic, because attempts to resolve chronic conflicts without hardheaded calculation would be merely wishful thinking and, in some circumstances, dangerously reckless. The action must be hopeful because it requires a willingness to bolster historical adversaries and take personal risks for a better future when key determinants of that future are, at best, imponderable. Displaying hope adds a performative ingredient at a time when it is essential. In periods of rapid political and institutional flux (such as South Africa in the early 1990s) it is impossible for any leader to foresee, much less control, the wider flow of events. Yet a skilled leader can estimate the likely response of an adversary to an unexpected opening and act to improve its chance of being constructive. Among other skills this ability requires empathy, the capacity to imagine oneself in another's place and comprehend how the conflict looks from his or her perspective.

We illuminate strategically hopeful action by drawing from the indefinitely iterated prisoner's dilemma featured in the work of Robert Axelrod. In contrast to Axelrod's unitary rational actors, however, we highlight the strategic dilemmas confronting individual leaders facing chronic conflict while retaining the support of refractory constituencies. Bitter communal conflicts resemble prisoner's dilemmas (PDs) because most people on both sides would gain, compared to the miserable status quo, were they able to cooperate. Yet, without mutual trust, no party has good reasons to cooperate; thus potential mutual gains go unrealized. This holds both for single-round PDs and for iterated PDs where the stakes are high and the number of rounds are known beforehand.

Actual conflicts are often of uncertain duration, generating more interesting possibilities. If A knows she will face B again in an indefinitely iterated interaction, then it might be in A's interest to cooperate, hoping B will reciprocate; and it may be in B's interest to reciprocate because if he defects, A will retaliate in the following round. This is the essence of “tit for tat” (TFT): cooperate on the first round, then mimic the other player (Axelrod Reference Axelrod1984, 13). However, A takes a risk by cooperating first. If instead of reciprocating, B exploits A's cooperation by defecting, A then defects on the next round, B defects again in turn, and the result is a self-reinforcing, mutually costly cycle. In contrast to a one-shot prisoner's dilemma, where the dominant strategy is always to defect, the indefinitely iterated PD offers genuine choices.

Axelrod found that TFT won in computer simulations of iterated dilemmas, beating a wide range of alternative strategies. Ironically, although TFT offers the highest average payoff, in any given round a TFT player never scores better than the other player, and it always scores lower than an opponent who responds to cooperation by defecting. Yet players employing “nasty” strategies score badly against other “nasty” players: Each drags the other down. TFT won “not by beating the other player, but by eliciting behavior from the other player which allowed both to do well” (Axelrod Reference Axelrod1984, 112).

Real-world communal conflicts rarely permit a fresh beginning, but exhibit instead an oft-repeated history of mutual violence and distrust. Axelrod recognized that with simple TFT, “once a feud gets started, it can continue indefinitely” (Reference Axelrod1984, 138). In later work Axelrod investigated strategies for restoring cooperation when it has broken down. These include “generous” TFT (occasionally cooperating in the face of an opponent's defection) and “contrite” TFT (cooperating in response to defection that was in response to one's own previous defection).Footnote ¹ Such restorative strategies can work if they are employed about 10% of the time, not more; otherwise they will be exploited (Axelrod Reference Axelrod1997, 33–9). They are useful here because they allow us to pinpoint the potential contributions of leaders to cooperative outcomes, and they highlight the risks leaders must be willing to take. Given that generous TFT will be exploited if used too frequently, leaders who employ it assume a significant risk—because they must commit 100% to the move once chosen.

Moreover, in long-standing conflicts such as those of South Africa, Northern Ireland, and Israel-Palestine, it is misleading to speak of “restoring” cooperation (as in Axelrod's model) because cooperation might never have existed. Strategically hopeful leaders must instead construct intercommunal cooperation for the first time and create an enduring constituency for it. And even if most people on both sides prefer peace to war, there typically remains a determined minority on each side that prefers war to any negotiated settlement. Such hard-liners must be marginalized if they cannot be convinced or coopted. This is among the most difficult and essential tasks of a strategically hopeful leader. If one cannot be confident that hard-liners (in the adversary's camp as well as one's own) can no longer be effective spoilers, then one cannot know what strategic game is being played. For among hard-liners it is a zero-sum conflict, not a prisoner's dilemma.

Axelrod did not address the role of leaders; his actors could be whole communities or even distinct species “cooperating” over the span of evolutionary time (Axelrod Reference Axelrod1984, 88–105). His narrative of the tacit “Live and Let Live” ethic (which emerged across trench lines mutually to limit casualties in World War I) shows that leaderless cooperation can sometimes reduce the death toll (73–87), but it did not end the war. One can imagine a comparable ethic occasionally emerging in neighborhoods of East Jerusalem, without resolving the larger conflict.

Individual leaders can initiate new, tentatively cooperative approaches more readily than can whole communities, but in the process they accept a different degree of risk than the community does. In addition to political failure and repudiation, such leaders risk assassination by extremists on both sides. The community can take a longer view: If one leader fails, it can elevate another; if a cooperative initiative fails, the community can pivot back to a warlike approach. Communities in long-term conflict thus mirror Axelrod's indefinitely iterated PD. Yet individual leaders may have only one major chance at intercommunal cooperation; if it fails, they could be finished. Their predicaments more closely resemble single-round PDs: The safe money bets the other way. To press forward with a cooperative opening, despite the odds, is a mark of strategically hopeful leadership. In effect, such leaders internalize the costs of solving collective action problems that will otherwise continue plaguing their constituencies.

Game theorists typically assume utility-maximizing actors with fixed preferences who are indifferent to one another (Morrow Reference Morrow1994, 19–20, 34). Axelrod made more illuminating psychological observations. He reported that in real-world approximations of the iterated PD, “the very experience of sustained mutual cooperation altered the payoffs of the players, making mutual cooperation even more valued than it was before.” Thus the successful experience of mutual cooperation can itself predispose people to view interactions in variable-sum rather than zero-sum terms. But the reverse also occurred: Failed attempts at cooperation evoked “a powerful ethic of revenge” and reinforced tendencies to view conflicts as zero-sum, even when successful cooperation would have produced mutual gains (Axelrod Reference Axelrod1984, 85, 110–12).

To have any chance of bridging long-standing divisions, strategically hopeful leaders must first judge for themselves that the conflict is potentially positive-sum. They have to believe, and persuade others to believe, that all parties lose if the conflict continues unabated (even if they lose unequally), and that all stand to gain if the conflict can be resolved. Yet they must recognize that accumulated bitterness and past failures at cooperation have led many to perceive the conflict as zero-sum—resolvable only by victory for one side and defeat for the other. Game theoretic models mask this judgment problem when utilities over outcomes are assigned to actors. The payoff schedule of a PD indicates that both actors are better off if they cooperate; the problem is that neither can trust the other. Strategically hopeful leaders confront the trust problem, but they also face another, equally difficult challenge: convincing skeptics on both sides that any potential outcome will leave both parties better off. The mutual cooperation outcome in the PD interaction, if never realized, may disappear altogether from participants’ consciousness, making it appear a game of pure conflict. Thus strategically hopeful leaders must not only play the game strategically; they must also persuade others about which strategic game is being played.

Strategically hopeful leaders must also know how to circumvent preconditions. Parties locked in chronic conflict typically insist on mutually incompatible preconditions for negotiations. Preconditions highlight and often magnify the prisoner's dilemma because each side insists on securing concessions before negotiations that could only be obtained with great difficulty through negotiations; in the unlikely event that its preconditions were met, it would have no further incentive to negotiate. Waiving one's own preconditions looks like waving the white flag. Strategically hopeful leaders must somehow sidestep their constituency's preconditions without acceding to their adversary's, and they risk failing if “the enemy” refuses to meet them halfway. Strict preconditions—such as demands to end violence or decommission—also enable hard-liners to kill negotiations by deliberately violating those preconditions.

Strategically hopeful leaders recognize their bridge-building attempts will fail unless preferences on both sides change. If leaders instead seek to maximize the satisfaction of prevailing preferences—which in practice means the preferences of the constituency to whom they are immediately accountable—then peace efforts are doomed. In times of regime transition, when stakes are high, institutions are in rapid flux, and the outcome is uncertain, political preferences become interdependent to a significant degree: The aims of each side become importantly conditioned by signals from the other side. It is here that strategically hopeful leadership can have its greatest impact. A cooperative opening that, against the odds, achieves early if limited success can change each side's perception of the other side's intentions. Changes in what is perceived to be possible alter in turn what is possible, to a degree unforeseeable to many before the process began. The assumption before 1994 that white South Africans would never relinquish their monopoly on political power turned out, given the right circumstances, to be unfounded.

The dynamic of interdependent preferences cuts both ways. On one hand it encourages hope for a new opening. However risky and difficult the first step, if it induces a constructive response, this can alter the preferences and perceptions of others, generate a surplus of good will, and make subsequent moves possible that previously were not. However, the same effect works in reverse: Any serious misstep risks modifying preferences in negative directions, making future cooperation harder than before. Strategically hopeful leaders recognize that, if they fail, their failure constricts the options of their successors.

Strategically hopeful leaders implicitly see power as variable-sum. Political power is frequently, perhaps even normally, gained at another's expense: One candidate wins an election, another loses. More fatefully, in the cases that concern us here, the success of reformist leaders depends on facing down their own radical flanks, whose members will see reformists’ gain of power as their own loss and vice versa. To perceive power as variable-sum does not mean that literally everyone gains or that gains are equally shared, only that gains can exceed losses or losses exceed gains. Even during the horrific meltdown of the former Yugoslavia in the 1990s (where, it is often remarked, “everyone lost,”Footnote ² and whose path South Africa might have followed), every atrocity produced its own relative winners and losers. A variable-sum understanding of power requires looking past the more obvious political gains and losses and taking into consideration the wider community's collective power over its own future, which can be enhanced or diminished by leaders’ success or failure in risk-taking efforts at cooperation.

Strategically hopeful leaders must recognize that their own power to secure a resolution acceptable to their constituency depends on preserving and, where necessary, reinforcing the power of the “enemy” leader across the table. They must resist the temptation to weaken an antagonistic rival and divide the opposition; otherwise their rivals will be unable to persuade their own constituencies to support an agreement. Here one's own power stands or falls with the power of one's counterpart on the other side. This is easy to recognize in principle, but acting on it is difficult.

We turn next to two decisions, one taken by Nelson Mandela in 1985 and the other by F. W. de Klerk in 1992, that exemplify strategically hopeful leadership. Neither leader possessed comprehensive, impartial understanding of the events in which they participated. Both were limited in what they could see and even more limited in what they could control. Nor were they saints, untainted by ordinary political ambition. Yet their political ambitions led them, at decisive moments, to risk career and perhaps life on making cooperative moves across the divide when the usual political incentives, and the advice of trusted colleagues, prescribed the opposite. Different leaders might well have chosen differently, with very different consequences for the country.

Some rational choice analyses of leadership (Frohlich, Oppenheimer, and Young Reference Frohlich, Oppenheimer and Young1971) treat leaders as political entrepreneurs who solve collective action problems in exchange for personal “profit” (appropriating revenues, distributing patronage, continuing in office). That analysis describes many politicians, but fails to illuminate strategically hopeful efforts to resolve long-standing bitter conflicts when all previous attempts have failed. Even success in those efforts may ultimately spell the end of a political career, as it did for F. W. de Klerk and David Trimble, and failure will end a career even sooner. Strategically hopeful leaders recognize the human cost of continued battle. Their ambition is perhaps to be remembered as one of those who resolved it.

TWO LEADERS, TWO GAMBLES

In 1985 Nelson Mandela (then in Pollsmoor Prison) decided to initiate secret talks with representatives of the National Party government, without the knowledge or approval of the African National Congress (ANC) executive committee—neither its leaders in exile nor those imprisoned at Pollsmoor with whom he was in regular communication. Mandela did not propose formal negotiations with the government; he invited “talks about talks.” Yet even this step was exceedingly risky, both for the ANC and for Mandela personally. The ANC's policy was that negotiations could begin only after the government satisfied various preconditions, none of which had been met: revoking the legal ban on the ANC and other anti-apartheid groups, releasing all political prisoners, and allowing open political opposition. On the contrary, the government's repression of anti-apartheid activity was increasing (Prime Minister P. W. Botha would declare a new state of emergency in 1986). The government's own preconditions excluded negotiations unless and until the ANC permanently renounced violence and dismantled its military wing, Umkhonto we Sizwe.

Mandela recognized that military overthrow of apartheid was “a distant if not impossible dream” (Mandela Reference Mandela1994, 457). Yet the idea of armed struggle, even on a limited scale, was enormously important to the ANC rank and file and could not easily be relinquished. By 1985 the Soviet Union had ended the support it once provided to the ANC and the South African Communist Party (Jung, Lust-Okar, and Shapiro Reference Jung, Lust-Okar, Shapiro and Shapiro2011, 94). This would have increased the ANC's motivation to negotiate with the government, but also heightened the risk in doing so, because eagerness to talk might telegraph the ANC's diminished military capability and worsen its bargaining position. Under the circumstances, for Mandela to invite talks could betray ANC weakness in an increasingly high-stakes struggle. The government might accept Mandela's invitation with the ulterior aim of trapping him and dividing ANC leadership at a crucial moment.Footnote ³

Nevertheless Mandela decided to go forward and to keep it secret from his ANC colleagues until he was committed:

Once his initiative was underway, Mandela informed his ANC colleagues. Their responses were sharply divided, but most were guardedly willing to permit Mandela to continue an action they would not have approved in advance (Lodge Reference Lodge2006, 158–60; Mandela Reference Mandela1994, 466–7; Read Reference Read2010). Mandela's strategic observations could be summarized as follows:

• • The South African conflict was potentially variable-sum. Although many on both sides perceived it as zero-sum, in fact both stood to gain from a political settlement and both would lose terribly in an escalating racial war.

• • Nevertheless, if present trends continued, the lose-lose outcome (“dark night of oppression, violence, and war”) would occur, because “both sides regarded discussions as a sign of weakness and betrayal.” Mandela was describing a classic prisoner's dilemma.

• • A leader must resolve this impasse, “move out ahead of the flock,” hoping this will enable people to perceive, and act on, the positive-sum possibilities. But this is risky because the leader cannot control the process he or she sets in motion; the other side may indeed suspect weakness and escalate its demands.

• • In taking this step Mandela internalized much of the risk, thereby diminishing it for other anti-apartheid activists. He realized that his initiative might backfire, finishing him as a top ANC leader. Indeed he discerned bargaining leverage with the government in the fact that, if “matters went awry,” his colleagues could limit the damage by renouncing the initiative of an irrelevant old man. Mandela might “move out ahead of his flock,” but he knew the “flock” was free to denounce him and refuse to follow.

Mandela's 1985 decision to “talk to the enemy” was one vital step in replacing apartheid with racially inclusive democracy (Lodge Reference Lodge2006, 165–6). Equally crucial and risky was the decision taken in 1992 by President F. W. de Klerk, the man who had released Mandela from prison two years earlier. Unlike Mandela's 1985 move, which occurred far from public view, De Klerk's gamble took place in the glare of publicity.

De Klerk belonged to P. W. Botha's cabinet during the 1980s, but had not participated in, and initially was not informed of, the government's talks with the still-imprisoned Mandela. Before he became president de Klerk appeared to be a typical conservative Afrikaner politician, displaying little reformist behavior. However, he became privately convinced that apartheid had failed and that only fundamental reform “could pull South Africa back from the edge of the chasm on which we were teetering” (de Klerk Reference de Klerk1998, 89). Botha had admitted as early as 1979 that South Africans must “adapt or die,” but was unwilling or unable to follow through—and even if Botha had committed to reform, there were then no ANC leaders with whom he could have negotiated. In contrast, from the outset of de Klerk's presidency in 1989 (by which time he would certainly have known of the Mandela talks), he judged Mandela to be someone with whom “it would be possible for us to do business” (158). In February 1990 de Klerk surprised South Africa and the world by lifting the ban on the ANC, the South African Communist Party, and the Pan Africanist Congress (PAC) unconditionally. He announced plans to release all political prisoners, including Mandela, and to begin negotiations toward democracy (de Klerk Reference de Klerk1998, 229–32; Jung and Shapiro Reference Jung, Shapiro and Shapiro1996, 194–5).

By 1992, however, de Klerk's reform effort was greatly endangered. Political violence had escalated, especially between ANC supporters and followers of Mangosuthu Buthelezi's ethnically Zulu Inkatha Freedom Party (IFP) in Natal. Negotiations on a new constitutional settlement had collapsed. There was an apparently unbridgeable gulf between the ANC's demand for full majority rule and the government's insistence on permanent, constitutionally guaranteed veto rights for all minorities (the white minority included). Instead of trusting the government to lead the reform process, as de Klerk hoped, the ANC had stepped up its campaign of mass action. Though it had announced a “suspension” of the armed struggle in 1990, it had not dismantled its military wing. The economy was imploding. Personal relations between de Klerk and Mandela had soured. Mandela accused de Klerk of complicity in government-instigated violence, a charge the latter vehemently denied; de Klerk in turn suspected Mandela of concocting such charges to maximize his bargaining leverage.Footnote ⁴ The country appeared headed for the abyss.

De Klerk's standing among the white electorate, to whom he owed his presidency and thus any opportunity to engage in reform, was eroding dangerously. In late 1991 and early 1992 the Conservative Party, which was intensely opposed to de Klerk's reforms, began winning by-elections in former National Party strongholds. The election results were widely interpreted as a rejection of the National Party's reform proposals on the part of white voters, who were at that point still the only voters who mattered.Footnote ⁵ Recalled de Klerk (Reference de Klerk1998, 229–32), “The mandate that I had received in 1989 from the white electorate was visibly slipping away from me and the National Party.” According to conventional wisdom, de Klerk at this point should have moved to the right, shoring up his political base and salvaging his party's governing status—even at the cost of scaling back or postponing his plans to dismantle apartheid. Many of de Klerk's supporters, advisors, and cabinet members urged this course.

Instead, on February 20, 1992, de Klerk surprised the public and many in the National Party leadership by calling a snap referendum among white voters on the question: “Do you support the continuation of the reform process that the state president started on 2 February 1990 and which is aimed at a new constitution through negotiations?” De Klerk typically sought consensus in his party and cabinet, but in this one case he acted unilaterally: “If I had put my decision to the vote the majority of the caucus would have opposed what they then regarded as an over-hasty and risky decision” (de Klerk Reference de Klerk1998, 232). He made clear that if the referendum failed he would resign. He campaigned vigorously for the March referendum, winning an impressive 68.7% Yes vote.

De Klerk later explained his reasoning in calling the referendum: People were expressing their fears and dissatisfactions in the by-election results, but when confronted squarely with the issue of the country's future they would rise to the occasion and embrace the need for change. There were no polling data on the subject. Instead, he was relying on his intuitive sense of what moved his compatriots and why (interview with the second author, Cape Town 2003).

De Klerk survived one risky move only to face another. The referendum had not ratified any particular post-apartheid settlement; it had merely authorized him to continue a process whose ultimate outcome he could not fully control. He would likely have been finished politically if, after winning the referendum, he had failed to secure an agreement with the ANC that was also acceptable to a critical mass of white South Africans. In 1992 the two sides’ constitutional demands were still fundamentally opposed. De Klerk and the National Party had promised during the referendum campaign that they would never give up constitutionally guaranteed power sharing, whereas the ANC rejected any such provision as apartheid in another form. Only in late 1993 did de Klerk relinquish entrenched power sharing, settling instead for constitutional guarantees on property rights, civil freedoms, and a two-thirds majority to alter the constitution. The white minority's willingness ultimately to accept in 1994 what it still appeared implacably to reject in 1992—black majority rule over which whites possessed no guaranteed veto rights—illustrates the fluid character of political preferences as events unfolded. (Whether de Klerk knew in 1992 that he was promising white voters more than he could later deliver is an open question.Footnote ⁶ )

De Klerk recognized he could not impose a constitutional settlement unilaterally. He needed the cooperation of Mandela and his supporters in the ANC—this after the National Party government and the ANC had been through decades of bitter conflict in which countless people had been uprooted, imprisoned, tortured, and killed. Unlike his predecessor, de Klerk was willing to plunge into the Rubicon and start swimming (de Klerk Reference de Klerk1998, 103–6). He had no guarantee, however, that anyone on the other bank—the ANC and its radical allies—would help him across if he began to drown.

Both Mandela and de Klerk recognized the interdependent character of the negotiated transition: Neither could succeed without the cooperation of the other. Both viewed the situation as variable-sum (they and their respective constituencies stood to gain or lose together, but each also hoped to gain more or lose less than the other). Both knew there were hard-liners on both sides who saw the conflict as irreducibly zero-sum, anticipating a violent showdown. De Klerk observed of Mandela: “We realized that we both bore the ultimate responsibility for ensuring that there would be a negotiated settlement and we were both committed to carrying out this responsibility” (de Klerk Reference de Klerk1998, 169). Mandela correspondingly recognized de Klerk's “genuine and indispensable contribution to the peace process,” remarking that “[t]o make peace with an enemy one must work with that enemy, and that enemy becomes one's partner” (Mandela Reference Mandela1994, 533). Both believed that the failure to cooperate would produce horrifying losses for all—“a dark night of oppression, violence, and war” (Mandela Reference Mandela1994, 457), “a prolonged struggle so bitter and destructive that there would be little left for anyone to inherit” (de Klerk Reference de Klerk1998, 121).

Yet both also recognized that there was conflict over the terms on which these common interests would be met. De Klerk and Mandela acted “in the full knowledge that we were opponents with divergent goals” (de Klerk Reference de Klerk1998, 169; see also Mandela Reference Mandela1994, 503). In addition to conflicts among the constituencies they represented and between the deeply divergent constitutional proposals, there was little personal trust between Mandela and de Klerk during the negotiations process. Yet both were willing to take significant personal risks to bridge the racial divide, hoping but not knowing that the other would reciprocate.

The South African transition was far from bloodless. Between February 1990, when Mandela was released from prison, and April 1994, when the ANC won South Africa's first democratic election, more than 14,000 South Africans died in political violence (Sisk Reference Sisk2008, 92–8). By some conventional measures the struggle over apartheid between 1985 to 1994 crossed the threshold of civil war; the early 1990s conflict between ANC supporters and Inkatha in Natal certainly did (Sambanis Reference Sambanis2004). Political violence motivated Mandela and de Klerk to persist in negotiations because both feared matters would become much worse if they failed to cooperate. Other leaders in their position might well have acted differently with radically different consequences.

LEADERSHIP, STRATEGIC INTERACTION, AND THE PRISONER'S DILEMMA

Mandela and de Klerk realized that resolution of the conflict depended on an interdependent decision: Neither could impose a unilateral solution, but each had instead to offer something that could become acceptable both to the adversary and to his own constituency. The tragedy of interdependent decision is that outcomes that are worse for all may result because leaders cannot agree on the terms of cooperation.

There is an enormous literature, scholarly and popular, on leadership but little of it addresses the problem that concerns us here: how leaders democratically accountable to one constituency in a long-standing communal conflict can initiate cooperation across the divide and then persuade their own constituency to start crossing a bridge that is still under construction. Analyses of democratic leadership that presuppose effectively functioning institutions widely perceived as legitimate, such as the large literature on presidential leadership, do not reach our case.

Rational choice theories of leadership often feature what William Riker calls “heresthetics” or “the art of political manipulation”—in which leaders strategically deploy words and actions to structure decisions to their advantage, to get their way without having to persuade anyone to modify their preferences (Riker Reference Riker and Finifter1983; Reference Riker1986; see also Dewan and Myatt Reference Dewan and Myatt2012, 432). For example, by manipulating the order in which a set of alternatives is voted on, one can win a decision—without altering preferences—that would have been lost had the voting order been different. Both Mandela and de Klerk were, among other things, skilled “political manipulators” in Riker's sense: At key moments both took risks that their colleagues would have opposed had they been consulted beforehand, but were guardedly willing to support afterward. Yet Mandela's and de Klerk's short-term “agenda manipulation” was directed to longer term ends that could only be realized if other key elites and substantial numbers of their constituents were persuaded, through both word and act, to modify their political preferences.

The notion that leaders merely manipulate fixed preferences provoked an initially useful, but now excessive counter-reaction in the enormous literature on “transforming” leadership pioneered by James MacGregor Burns (Burns Reference Burns1978; Reference Burns2003; Conger Reference Conger and Goethals2004). In contrast to merely “transactional” leaders, “transformative” leaders are said to create new possibilities by modifying the character and preferences of followers. Transformative leadership theories identify an important element missing from rational choice accounts, but too often treat transformation as a quality inhering in a charismatic leader or emerging from that leader's interactions with committed followers. At least equally important for understanding acts of communal bridge-building are leaders’ strategic interactions with rival leaders and with their constituencies—for whom an adversary's charisma might actually be a threatening liability. In the South African transition, Mandela's transformative capacities depended on de Klerk's responses and vice versa and on the fragile support of skeptical constituencies—as both leaders fully understood. For these reasons we find theories of strategic interaction, even when they exclude or underplay leadership, more useful to the problem at hand than much of the leadership literature. However, it needs to be supplemented by greater attention to the way leadership shapes the outcomes of strategic interactions. Axelrod's iterated prisoner's dilemma has already been discussed; here we examine Thomas Schelling and Josep Colomer.

Despite its vintage, Schelling's The Strategy of Conflict remains among the most fruitful examinations of the type of decision problem considered here. Schelling makes a fundamental distinction between zero-sum (or “constant sum”) conflicts, where “more for one participant inexorably means less for another,” and variable-sum, or “mixed motive” conflicts, in which “there are common as well as conflicting interests among the participants” and “mutual dependence as well as opposition.” Strategy in a variable-sum game must take account not only of “the division of gains and losses between two claimants” but also “the possibility that particular outcomes are worse (better) for both claimants than certain other outcomes. . . . There is a common interest in reaching outcomes that are mutually advantageous.” Neither participant can fully control the outcome; instead, “the ability of one participant to gain his ends is dependent to an important degree on the choices or decisions that the other participant will make” (Schelling Reference Schelling1980, 5). Cooperation might fail because whoever first indicates a willingness to cooperate signals potential weakness, inviting exploitation by the other side, or because one or both sides are unable credibly to commit to positions from which they will not later be tempted or pressured to move. For negotiations to succeed, expectations on both sides “must somehow converge on a single point at which each expects the other not to expect to be expected to retreat” (35, 70).

Either party can jeopardize the possibility of an agreement by weakening the other too much. Mandela illustrates Schelling's point when he writes that despite the frictions “I never sought to undermine Mr. de Klerk” because “the weaker he was, the weaker the negotiations process” (Mandela Reference Mandela1994, 533). Mandela did on occasion push ANC demands close to the limits of what de Klerk could accept, but never forgot that de Klerk had his own constituencies to satisfy.

Schelling's variable-sum model aptly describes the South African racial power contest as perceived by both Mandela and de Klerk. Both agreed that through cooperation each side stood to gain—at least compared to the likely outcome without cooperation. Both recognized the interdependent character of the decision: If either refused to sign or agreed to a pact he could not persuade his constituency to support, then both constituencies would lose, as would the leaders themselves, politically and personally. Each recognized that the other also recognized these things.

Yet Mandela and de Klerk also understood something Schelling does not discuss: Whether a conflict is perceived as zero-sum or variable-sum may itself be at stake in what leaders do and say. Schelling takes as given that a conflict is either zero-sum or variable-sum, which in turn is a function of the respective, independently derived preferences on each side. This initial classification is decisive because “the intellectual processes of choosing a strategy in pure conflict and choosing a strategy of coordination are of wholly different sorts” (Schelling Reference Schelling1980, 4, 96). But Schelling does not explain how participants decide whether they believe the conflict to be zero-sum or variable-sum. Mandela's inviting “talks about talks” with the government in 1985 was risky because, though he perceived the conflict as potentially variable-sum, he did not know whether his colleagues or the National Party government would agree. For Mandela to frame the conflict as involving potential shared gains was a hopeful act.

Even if leaders on both sides view a conflict as potentially variable-sum, cooperation might fail because their respective constituencies—which retain the power to block agreements and replace their leaders—might instead prefer a fight to the finish. Mandela noted that many anti-apartheid activists in the early 1990s demanded “a victory on the battlefield, not the negotiating table,” and he noticed signs at rallies reading, “MANDELA, GIVE US GUNS” and “VICTORY THROUGH BATTLE NOT TALK” (Mandela Reference Mandela1994, 506, 526). De Klerk was equally aware of the white right who anticipated a violent showdown (1998, 316–19). Both faced a twin challenge: first to manage a difficult bargaining process with adversaries and then to persuade a critical mass of the rank and file on both sides to support an agreement.

Unlike more formalized rational choice approaches, Schelling does not stipulate invariant preference functions. Yet his bargainers maximize their share of a fixed supply of gains from cooperation. He ignores the ways in which actors with interdependent preferences can expand (or shrink) the available surplus depending on their behavior in successive rounds of bargaining. This is the performative dimension of cooperation: When successful, it creates a new reality (Austin Reference Austin1975).

Schelling also downplays the moral dimension of leadership: Negotiations between warring underworld gangs, or kidnappers and victims, do not differ in strategic logic from any other bargaining situation in which conflict and cooperation intermix (Schelling Reference Schelling1980, 12, 43–4). Schelling's undemanding leadership assumptions reach their limits in explaining anything as difficult as the South African settlement. What Mandela asked white South Africans to give up—their monopoly on political power, in exchange for an uncertain future under majority rule—went well beyond the kind of concessions featured in The Strategy of Conflict. It is unsurprising that many people on all sides saw the conflict as zero-sum and, in particular, that white South Africans anticipated harsh treatment from an empowered black majority they had oppressed for so long. White South Africans still retained decisive military and economic superiority, and some on the right would have gladly pushed de Klerk aside to spearhead a violent contest. Both de Klerk and Mandela kept a close eye on the South African military throughout a transition that the military could have halted had it chosen to do so.Footnote ⁷

Yet in the end white South Africans turned over power to a black-majority government led by Nelson Mandela, who promised them no more—but also no less—than to live as equal citizens in a multiracial democracy. That “South Africa belongs to all who live in it, black or white” was affirmed in the ANC's Freedom Charter of 1955, but few South Africans of any race had personally experienced this kind of political community. One of Mandela's accomplishments was to make this vision believable, both to its advocates after nearly a century of frustration, and to white South Africans who feared that majority rule meant oppressor and oppressed exchanging positions.

Mandela understood this fear and addressed it: “I have fought against white domination, and I have fought against black domination. I have cherished the ideal of a democratic and free society in which all persons live together in harmony and with equal opportunities” (Mandela Reference Mandela1994, 322; Reference Mandela2004, 59–62). Yet it was his actions that made the statement believable, just as de Klerk's turning on the white right and plunging into the Rubicon established his credibility. Taking these risks enabled them to persuade their supporters to rethink the conflict in variable-sum terms. Once they did, they could focus on their enduring common interests beneath the more obvious conflicts.

Whereas Schelling does not specifically discuss democratic transitions, Josep M. Colomer's Strategic Transitions: Game Theory and Democratization (Reference Colomer2000) is immediately directed to the problem at hand. Colomer examines possible coalitions among six types of strategic actors: radical democratic opposition, moderate democratic opposition, radical soft-liners (reformists in the regime), moderate soft-liners (less committed reformists), moderate hard-liners, and radical hard-liners (Reference Colomer2000, 49). His model emerges from democratic transitions in the former Soviet bloc, but with some modification can also illuminate South Africa, Northern Ireland, and other cases.

Colomer seeks to explain why confrontation between a regime and its democratic opposition sometimes produces stable compromises and at other times triggers “frontal conflict” whereby both sides risk “becoming an absolute loser” (Reference Colomer2000, 1–3). Like Schelling, Colomer presupposes variable-sum interdependent decision: Neither the regime nor its democratic opposition is powerful enough unilaterally to determine outcomes, and failure to cooperate leads to consequences that are worse for both. Following Axelrod, Colomer takes the prisoner's dilemma as a starting point (57–61) and seeks to explain how “cooperation can develop even among antagonists” (51).Footnote ⁸ However, Colomer argues that critical transitions more closely resemble single-shot PDs because “actors’ opportunities to repeat interactions in a process of regime change may be remote” (51).

To explain how cooperation might emerge from a single-shot PD, Colomer endows actors with the capacity for foresight. Mutually destructive outcomes result from “short-term-looking decisions. . .made in ignorance of other actors’ choices.” Yet rational actors should be able “to foresee not only the immediate consequences of their choices but also the consequences of the other actors’ reactions, their further counter-reactions, and so on” (Colomer Reference Colomer2000, 2). Colomer's prescient actors do not make simultaneous, blind choices as in the classic PD. Instead they “enter into open, dynamic interactions in which they make choices precisely in the expectation of other actors’ reaction, as is typically the case in political exchanges” (52). Rejecting the assumption that moves once taken cannot be reversed, Colomer introduces retractable moves; for example, “if the hard-liners reject an intermediate formula proposed by the opposition, the latter can resume mobilizations and protests” (57–8). Colomer's model thus builds flexibility into single-shot PDs; mutual defection is not inevitable.

Colomer's introduction of foresight—the principal innovation of his model—simultaneously reveals a limitation: There is no specific role for leadership. If foresight was equally distributed among human beings and operative in all strategic interactions, PDs would always be resolved and civil wars would seldom occur. However, most people most of the time operate with limited and inflexible cognitive capacities (Kahneman Reference Kahneman2011, 19–108; 259–376). Colomer's actors represent strategic positions (such as “moderate democratic opposition”), not individual leaders, even when he illustrates those positions with proper names like Gorbachev and Havel (Reference Colomer2000, 41). If the classic PD underestimates the cooperative possibilities, Colomer's revision, although insightful, downplays the difficulty of exercising foresight in periods of rapid and extraordinarily complex change.

Nor does the model provide for the different degree of risk run by individual leaders as compared to collectivities during such periods. Colomer assumes, for example, that actors’ strategies will be weighted by a preference for peace over war: “A transition by agreement is less risky for the actors involved than a civil war” (Colomer Reference Colomer2000, 2). This is true as a general proposition (otherwise negotiated solutions would be impossible). However, for individual leaders it is often much riskier, politically and personally, to initiate a cooperative opening where there has been a long history of conflict and distrust than to do nothing. What would ultimately be risk minimizing for the group, if it can be achieved, may be risk augmenting for the leader who makes the first move. The foresight required under such circumstances is more complex than the general recognition that both groups lose if they fail to cooperate.

ZERO-SUM AND VARIABLE-SUM POWER

The prisoner's dilemma is a parable of powerlessness. The prisoners lack power to realize their common interests because they can neither alter the rules nor trust one another to cooperate under those rules. With respect to utility, the PD is potentially variable-sum: Mutual cooperation yields the highest aggregate utility and mutual defection the lowest. However, with respect to power, the prisoners’ interaction appears zero-sum, because the only aspect of the interaction lying within each prisoner's control is deciding whether to cooperate or defect. Cooperation leaves one powerless and vulnerable. Defection promises each prisoner a small degree of power, though of course at the other's expense.

An iterated PD creates a wider range of options: If each actor must take future retaliation or cooperation by the other into account, defection is no longer obviously a dominant strategy. An invitation to mutual cooperation is sometimes successful, though it offers no guarantee. This expanded range of strategies suggests an alteration of the power relations among the actors. This latter possibility has not received much discussion, perhaps because the concept of power is not explicitly employed in formal game theory, though assumptions about power are implicit in the structure of the game.Footnote ⁹ If under the classic one-shot PD, each actor's power comes at the expense of the other, and aggregate power is fixed at a low level, then it should follow that, under an iterated PD where reciprocity and trust can potentially emerge, the power of each actor does not necessarily come at the other's expense, and the power available to both actors is potentially variable, not fixed. This hypothesis requires that we examine the concept of power itself.

The meaning of power has long been contested (Lukes Reference Lukes2005, 14–38, 60–107). Hobbes defined it as one's “present means to obtain some future apparent good” (Reference Hobbes and Shapiro2010, 93). This definition of power is not inherently zero-sum; in principle we might secure our own good by cooperating with others in realizing theirs. Yet where interests conflict or where shared interests go unrealized because trust is absent (as in the state of nature), power for Hobbes becomes zero-sum in practice. Thus elsewhere Hobbes provides the classic zero-sum description of power: “Because the power of one man resisteth and hindereth the effects of the power of another: power simply is no more, but the excess of the power of one above that of another. For equal powers opposed, destroy one another, and such their opposition is called contention” (Hobbes Reference Hobbes1928, 26).

For Hobbes, one's power is demonstrated through victory in a head-to-head contest. This premise is shared by the realist school of international relations theory, which perceives an anarchic realm in which war-winning capacity counts above all (Gilpin Reference Gilpin1981, 94; Mearsheimer Reference Mearsheimer2001, 2). Yet Hobbes's idea is also echoed in the way many political scientists define power even under the operation of peaceful democratic rules. According to Nelson Polsby—speaking here as a “pluralist” in the “three faces of power debate”—in studying power we should focus on who wins and who loses in a “direct conflict between actors” because this is the best measure of their respective “capacities to affect outcomes” (quoted in Lukes Reference Lukes2005, 18). The “radicals” in this faces of power debate pointed out that conflict and domination were not always readily visible. Yet they did not question the assumption that power always comes at another's expense; indeed they drove that premise further (see Lukes Reference Lukes2005 for that debate, including his recent revision of his original view [63–5]).

Conflict and inequality are always relevant to understanding political power. The more problematic assumption is that A's capacity to “affect outcomes” is a direct function of B's lack of capacity to affect those same outcomes. This doubtless describes many power relations, but not the negotiations between Mandela and de Klerk. If those interactions were conflictual and “subtractive” in some respects, they were cooperative and “additive” in others. Conflict and cooperation fused in the same complex relation; both shaped the outcome. To comprehend as power only the conflict and to ignore the cooperation—as though cooperation were passive, something other than power—is to misunderstand the power relationship itself.

The assumption that power is inherently zero-sum has not gone unchallenged. Power has alternatively been defined as the cooperative capacity to achieve collectively shared aims (Talcott Parsons) or the general capacity of human beings to “act together” (Hannah Arendt). Yet Parsons’ and Arendt's arguments, though insightful, are flawed by their treating power as almost wholly cooperative, thus neglecting the element of conflict that preoccupies other theorists of power; for a discussion and critique see Lukes (Reference Lukes2005, 30–5) and Read (Reference Read2012). Some challenges to the zero-sum view do incorporate both conflict and cooperation, but fail to develop alternative variable-sum descriptions in much detail (for example, Baldwin Reference Baldwin, Carlsnaes, Risse and Simmons2002; Giddens Reference Giddens1984; Wrong Reference Wrong1995; for a more fully developed variable-sum argument incorporating both cooperation and conflict, see Haugaard Reference Haugaard2012). The zero-sum view of power, despite its one-sidedness, has produced rich descriptions of political life (for example, Gaventa Reference Gaventa1980). The variable-sum alternative view, to be fully persuasive, must pass this same test.

The political interactions between Mandela and de Klerk, and between the constituencies they led, suggest that a variable-sum theory of power can persuasively describe significant dimensions of political life—without neglecting inequality, domination, the potential for violence, and the persistence of conflict. Both Mandela and de Klerk understood the struggle in variable-sum, not zero-sum terms. When Mandela reminded South Africans that “I have fought against white domination, and I have fought against black domination,” he was challenging a zero-sum picture of power: that liberation for either community entailed domination for the other. Mandela's hopeful view—one shared with de Klerk, despite their divergent constitutional visions—was that black South Africans could be liberated without dominating whites (or any other group of this multiracial society).Footnote ¹⁰

Yet Mandela and de Klerk also realized that many South Africans on all sides saw one group's gain as the other's loss; such zero-sum perceptions threatened to produce negative-sum outcomes. Both were aware of the conflicting aims of their respective constituencies and of their personal conflicts as rival leaders. Yet both also recognized that each had to keep the other strong enough to retain the support of their constituencies for any negotiated settlement, and for this reason each could not push an advantage too far. They can be imagined as actors in a prisoner's dilemma who, by taking risks to generate trust, have found a way out of the prison. In the classic prisoner's dilemma the powerlessness of each ensures the defection of both. Mandela and de Klerk overcame this dilemma because their interactions generated new power on both sides.

The variable-sum view of power sketched out so far does not “refute” the common assumption that political power is gained at another's expense, for two reasons. First, where power is cooperatively generated, there remain important respects in which power gains for some entail power losses for others. In a stable democracy rival parties compete for a fixed number of powerful offices even as they cooperate to maintain the power of the democratic system itself. In South Africa, the reversal-of-domination scenario feared by white South Africans did not occur, in part because de Klerk and Mandela cooperated to prevent it. In this sense the variable-sum model best captures events. Nevertheless white South Africans had to relinquish significant political power for black South Africans to gain new political power; in this sense the zero-sum view retains its attraction.

Furthermore, the zero-sum hypothesis cannot be dismissed because it remains a self-fulfilling prophesy. If one behaves as though power is radically zero-sum, one can indeed make it zero-sum. Had either Mandela or de Klerk been replaced by leaders who understood the contest in zero-sum terms and acted accordingly, an escalation of racial war would have been the likely outcome. A zero-sum strategy, once decided on, will “prove itself correct” because the adversary responds in kind. A strategically hopeful opening, by contrast, might fail. There is no “best strategy” but instead a morally charged choice among contingent strategies.

UNIONISTS AND NATIONALISTS, ISRAELIS AND PALESTINIANS

The South African constitutional resolution was not replicated in Northern Ireland or Israel. Those conflicts differ in their histories, the ideologies invoked, the role of outside actors, and the design of settlements realized or proposed. Yet there are significant parallels.

First, all three are (or were) high-intensity, self-reinforcing conflicts with no end in sight. Second, none of the parties to these conflicts could win a decisive victory and impose a unilateral solution, but reformers in the government and moderates in the opposition could potentially negotiate settlements that would command enough support to survive. Finally, all three conflicts can be understood either as zero-sum or variable-sum, depending on how one frames them and whether one regards the preferences of the principal actors as fixed or dynamic. There was nothing about the South African conflict that made it inherently more amenable to resolution than the other two. Apartheid was obviously incompatible with nonracial democracy. Northern Irish nationalism contradicts Northern Irish unionism. Israelis and Palestinians advance irreconcilable historical and religious claims to exclusive possession of the same land. If human beings were mere replicas of the ideologies they espouse, then all these conflicts would be inescapably zero-sum. In fact, people who live through chronic conflicts often suffer greatly, and most appear to place more value on physical and economic security and personal liberty than on pursuing ideological stances to their ultimate conclusion.Footnote ¹¹

Protracted, self-reinforcing conflicts will continue unless a leader of at least one of the parties stakes his or her career on altering the dynamic. Even so, the odds of failure are high; there are many more ways for negotiations to fail than to succeed. And leaders who successfully bridge the abyss against the odds may receive few political rewards for doing so, as Northern Ireland demonstrates.

The 1998 Good Friday (or Belfast) Agreement should be considered a qualified success because it ended most of the political violence,Footnote ¹² and all significant players remain committed to employing only peaceful means in pursuit of political goals. (Peace efforts stalled whenever parties demanded preconditions and advanced when these were waived [Mitchell Reference Mitchell1999, 22–38]). The fundamental question—whether Northern Ireland shall remain with the United Kingdom or join the Republic of Ireland—remains unresolved, and Protestant and Catholic communities remain highly politically and culturally segregated (McGarry and O'Leary Reference McGarry, O'Leary and Taylor2009, 65–9; McGlynn, Tonge, and McAuley Reference McGlynn, Tonge and McAuley2012). However, all major parties, including those most committed to a united Ireland, have accepted “the principle of consent”: that “Northern Ireland should remain in the UK as long as a majority of Northern Ireland's citizens support this status” and that any unification with Ireland requires majority support of Northern Ireland's people in a referendum (McGarry and O'Leary Reference McGarry, O'Leary and Taylor2009, 56). The agreement was facilitated by external actors—Britain, the Republic of Ireland, the United States—but would have been impossible without risk-embracing political leaders from both sides of Northern Ireland's communal divide. Yet those leaders occasionally lost their nerve or dragged their feet in ways that weakened the example they set.

In Northern Ireland the political incentives for communal reconciliation have always been slim. Unionist firebrand Ian Paisley expressed the once-prevailing view in both camps: “A traitor and a bridge are very much alike, for they both go over to the other side” (Powell Reference Powell2008, 54–5). In the decades preceding the 1998 agreement, Unionist leaders such as Terence O'Neill and Brian Faulkner who reached across communal lines found their political careers abruptly cut short. For Nationalists and RepublicansFootnote ¹³ (those who seek to unite Northern Ireland to the Republic of Ireland) the political rewards for compromise have been equally meager. The most significant Nationalist bridge-builder has long been John Hume of the Social Democratic and Labour Party (SDLP), who since the 1960s has advocated what ultimately became central principles of the 1998 agreement: commitment to exclusively peaceful measures, the legitimacy of both Northern Ireland political traditions, and the principle of consent (McLoughlin Reference McLoughlin2010). Yet Britain's violent repression in 1972 energized the Irish Republican Army (IRA) at the expense of Hume's peaceful approach. Republican leaders who agreed, even provisionally, to a partitioned Ireland invited the fate of Michael Collins in 1922. Gerry Adams, the Sinn Féin leader who eventually turned the republican movement toward politics and (gradually) away from violence, knew that in doing so he risked assassination (Powell Reference Powell2008, 100, 147–8).

Yet by 2007 Paisley and his Democratic Unionist Party (DUP) were willing to cross a bridge (the 1998 agreement), whose builders they had denounced as traitors, and accept a power-sharing arrangement with Sinn Féin and Gerry Adams, who in entering government tacitly accepted a divided Ireland, at least for the present. The 2007 electoral success of the DUP and Sinn Féin came at the expense of the parties and leaders who did the most to make the 1998 agreement possible. In the 1980s Hume, recognizing that lasting peace was impossible without the participation of Sinn Féin, risked his reputation by entering into initially secret talks with Gerry Adams, hoping to persuade Republicans to declare a ceasefire (McLoughlin Reference McLoughlin2010, 153–67). (Hume also sought Sinn Féin's inclusion to strengthen the Nationalist bloc in any power-sharing settlement.) The IRA ceasefire was so long delayed that Hume's eventual success appeared as failure at the time. Hume insisted that Sinn Féin be treated as a legitimate party to the Good Friday settlement—perhaps recognizing that a successful peace agreement would boost Sinn Féin's political fortunes at the expense of Hume's own party, as indeed happened (McGlynn, Tonge, and McAuley Reference McGlynn, Tonge and McAuley2012, 10–14).

Unionist support for the agreement depended critically on the Ulster Unionist Party (UUP) and its leader David Trimble. Trimble's party and political career afterward were eclipsed because of his efforts, in part because the IRA's long delay in disarming left him hanging (Powell Reference Powell2008, 203–5). However, Trimble's own limitations as a leader contributed to his decline. In 1997 he made the risky decision to enter negotiations that could have gutted his support base, and he kept the UUP at the table despite significant opposition within his own party and right-wing charges that he was betraying his own people (Mitchell Reference Mitchell1999, 108–17). Yet during the May 1998 referendum on the agreement, Trimble's resolve wavered. In contrast to F. W. de Klerk, who called the 1992 referendum and personally led the successful campaign for a “yes” vote, Trimble—though he continued to voice support for the agreement—stayed largely on the sidelines during the campaign, apparently in response to the heat he had taken earlier. Supporters of the agreement repeatedly advised Trimble “to become more forthright and to engage in more active campaigning,” and in the final weeks before the vote Trimble did become more active (Hancock Reference Hancock2011, 103, 111). The “yes” vote ultimately succeeded among Protestants, but just barely, and many of those votes were shaky (Hayes and McAllister Reference Hayes and McAllister2001).

Trimble's political future was uncertain in any case, but his episodes of hesitation did nothing to restore his political fortunes and clouded the legacy of his more courageous moments. The same is true of Adams’ unwillingness or inability to persuade the IRA to disarm in the years immediately following the 1998 agreement. The IRA's long-delayed decommissioning in 2005 appeared grudging, not an act of communal reconciliation.Footnote ¹⁴ Trimble and Adams deserve credit for the risks they took. Yet in the end John Hume, who first attempted a bridge to Unionists, then risked his reputation and his party's electoral future to persuade Sinn Féin to cross that bridge, set the strongest example for Northern Ireland's future leaders.

Today the Israeli-Palestinian conflict appears impossibly difficult to resolve, despite its continuing urgency. Both sides view any hint of compromise as signaling weakness, and both insist on preconditions each knows in advance that the other will reject. Repeated failures to secure peace have reinforced a penchant to regard as some primordial antagonism what has in fact resulted from contingent choices and repeated failures of leadership. Here we focus on one particular missed opportunity.

In 1993 Israeli prime minister Yitzhak Rabin decided for the first time to talk to the Palestinian Liberation Organization (PLO) and its chairman Yasser Arafat, who was then the most powerful Palestinian leader and essential to any deal. Arafat was no Nelson Mandela. He headed a corrupt and ineffective organization, and his leadership status among Palestinians was often shaky. He lacked Mandela's strategic judgment. Yet Arafat had taken a historic and politically risky step when he publicly acknowledged the State of Israel in 1988. In the mid-1990s Arafat, whatever his faults, seemed willing and able to secure Palestinian support for an agreement that appeared potentially viable (Jung, Lust-Okar, and Shapiro Reference Jung, Lust-Okar, Shapiro and Shapiro2011, 107–14).

By 1995 both Rabin and Arafat were well positioned to manage hard-liners on their respective flanks and consummate an agreement, the main elements of which had been hammered out in secret negotiations in Oslo and announced in the fall of 1993. Rank-and-file support was strong in both communities for the two-state solution envisaged at Oslo. Rabin was a war hero whose dedication to Israel's security was not in doubt. Arafat committed to policing the West Bank to secure Israel from Palestinian attacks. Both were personally invested in the process and recognized a common interest in preventing terrorism. Had the process continued, they might have reached a provisional agreement delivering benefits that would have replenished their political capital for further negotiations.

Rabin's assassination in November 1995 by an Israeli right-winger opposed to the peace process was a stunning blow. Yet tragedy might have been turned into opportunity had Rabin's successor, Shimon Peres, been willing to take greater political risks. Peres could have called a snap election in the wake of the assassination and won an endorsement from the Israeli public for continuing the peace process, analogous to de Klerk's 1992 referendum. At the time, public opinion on both sides strongly favored a two-state solution, and outrage at Rabin's assassination had all but the most fanatical Israeli right on the defensive.Footnote ¹⁵ Major issues remained unresolved, but this was also true in South Africa in 1992. One round of successful negotiations shifts perceptions of what might be possible in the next round, which in turn changes what is possible.

However, Peres missed the opportunity, tacking instead to the right. He permitted the assassination of Hamas militant Yahva Ayyash in January 1996, reinforcing the cycle of violence and closures on the West Bank and Gaza. Peres responded to attacks from southern Lebanon by bombing Lebanese refugee camps. A wave of suicide bombings in the spring of 1996 hardened the Israeli stance in negotiations. Palestinian radicals thus helped secure the victory of the Israeli right. Peres alienated Israeli supporters of the negotiations and lost the May 1996 election to Benjamin Netanyahu, who was openly hostile to the Oslo accords. Arafat's political support among Palestinians had been decisively weakened by his failure to secure an agreement. When President Clinton summoned Arafat and Israeli prime minister Ehud Barak to Camp David in 2000, Arafat was offered a deal similar to what he would have accepted five years earlier, but it was too late; Palestinian support for the deal had evaporated—as had support among Israelis.Footnote ¹⁶ Arafat was no longer able to secure Palestinian consent to anything Barak could have offered. Mandela and de Klerk both realized, despite their conflicts, that each needed to keep the other strong enough to close the deal. Recognition of this strategic reality has been in chronically short supply in the Israeli-Palestinian conflict.

Rabin's fate demonstrates that political risk-taking can literally be fatal. Other leaders involved in the Israeli-Palestinian conflict have shown little appetite for the kind of strategically hopeful action that made the South African transition possible. Leaders on both sides (which since the 2006 elections in Gaza and the West Bank have included Hamas) have not placed high enough priority on resolving the conflict or taken the risks necessary to advance its prospects for success. Indeed, both sides have countenanced policies likely to worsen it. Yet further entrenching the status quo is not risk-free either. Leaders who refuse to take risks to recast festering conflicts thereby increase the costs and dangers faced by someone else, somewhere else, sometime in the future.