A Game-Theoretic Analysis(1)

Tore Nyhamar

This article discusses the preconditions for settling ethnic conflict through a constitutional compromise: democracy. The focus is on the conditions for transition to democracy amidst intense ethnic strife. What factors facilitate transition to democracy and what factors are obstacles? It is assumed that the attitude of social groups to democracy is determined by their leaders' rational calculations of the prospects of social, economical and political benefits. In other words, social groups have the capacity to formulate collective interests and act strategically to further them, and their leaders choose the alternative path of action with the highest expected benefits among those available. To extend the argument, I will first draw on some recent analysis in the rational choice literature on institutions. Second, I will analyse two very different contexts in which transitions to democracy were attempted, the events in Angola 1974-75 and in Zimbabwe in 1979-80. Rational choice theorists try to discover the meaning of rationality in different contexts, and the study of strategic choices and interaction of the six political elite groups in Angola and Zimbabwe, each with a core ethnic constituency, makes empirical probing and refining of the propositions of rational choice theory possible.

An ethnic community is rooted in perceptions of common ascent, a shared culture or language. Ethnic identity is a somewhat elastic concept, taking both objective physical evidence and subjective group conceptions into account. A minimal scale requirement is that ethnic membership transcends the range of face-to-face relationships (Anderson, 1983, p. 14; Horowitz, 1985, p. 53). A civic or national community, on the other hand, is rooted in geographic space, integrating laws and institutions, citizenship and shared values. The relative importance of the subjective and objective component in ethnicity is disputed. Those stressing the subjective component argue that ethnicity is a matter of choice (Gellner, 1983; Hobsbawn, 1990). Those stressing the objective component argue that people's choice of identity is severely circumscribed at best (Smith, 1986; Hroch, 1985; Connor, 1994). Civic integration, or inter-ethnic nationalism, means that the people's preferences for identities have changed from an ethnic identification to a civic, national or patriotic identification. Many regard civic integration as a precondition for democracy (Gottlieb, 1993, pp. 43-44; McGarry & O'Leary, 1993, p. 16ff). Whatever the foundations of identity, replacing ethnic identities with a state citizen identity, or somehow separating the two, is a tall order for ending ethnic conflict. Normative commitment to democracy may be a necessary condition for its long-term stability, but is unlikely amid ethnic strife in areas with weak democratic traditions.

Comparing Angola and Zimbabwe eliminates changes in identity and in preferences about democracy as causal factors because they are constant across the two cases (Meckstroth, 1975; Frendreis, 1983, p. 262). No significant differences in identities or preferences about democracy emerge among the leaders that could explain why a democratic constitution was agreed upon and implemented in Zimbabwe, while the attempt to establish democratic institutions failed dismally in Angola. Theoretically, leaders may switch to a democratic strategy even though preferences concerning democracy remain the same (Przeworski, 1991). Empirically, only the transition to democracy in Spain has been satisfactorily explained without changes in preferences about democracy (Colomer, 1995, pp. 1-11). This article will demonstrate that a transition to democracy ended the war in Zimbabwe, even though the actors' primary identification remained ethnic and there was no normative commitment to democracy. Moreover, the article draws general lessons from the two cases about how to facilitate transitions to democracy.

The Theoretical Framework

Adam Przeworski hones in on two central features of democracy: The outcome of the political process is uncertain, and the process is determined by "the people." To understand the potential of democracy for conflict resolution, the distinction between agreement about the constitution and agreement about substantive outcomes is crucially important. The make-up of the constitution affects all issues and the stakes are therefore higher in constitutional issues than in substantive issues. Democracy is an institutional compromise, not a compromise over outcomes. On the contrary, the institutional democratic compromise subjects all interests to competition. No group can reverse an undesired outcome ex post facto; democracy is the institutionalisation of uncertainty. On the other hand, an established democracy reduces the stakes of the political game because specific outcomes may always be reversed later.

A transition from authoritarian rule to democratic rule has taken place when the threshold beyond which no one can intervene to reverse the outcomes of the political process has been crossed (Przeworski, 1991, p. 14). Przeworski argues the evidence available to him that a necessary precondition for transition is that the "right," the former undemocratic power holders, are able to win the first election because they will not relinquish control of outcomes unless they are reassured immediately that they can compete effectively for power under democracy. Two social groups are especially important during transitions to democracy because they often possess the capability to control the political process ex post facto: those with arms and those who control productive resources (Przeworski, 1988, p. 72; Przeworski, 1991, p. 51). Armed ethnic groups have the ability to reverse policy outcomes; their disarmament is therefore necessary for the transition to democracy to be complete.

Second, the transition to democracy is facilitated if economic and political privileges are separable. In Eastern Europe, the old nomenklatura class supposedly had access to property only by virtue of their position within the power apparatus. In Spain and Latin America, the former power holders could abandon political control while retaining their economic position. The transition was easier in the latter case because the stakes were lower--the play was only for political, not economic survival (Przeworski, 1988, p. 75).

The central feature of the democratic process is that it invariably produces periodic winners and losers. The question is why the losers should comply with the outcome of the democratic process. There are three possible answers. First, compliance may be spontaneous, motivated by self-interest. Complying with the outcome, even when it is a defeat, and acting within the institutional framework, is better for the relevant political forces than trying to subvert democracy. Second, compliance may be enforced. Democracy holds because actors that would be better off not accepting the outcome are punished by an exogenous third party. Third, compliance may be morally motivated. People accept outcomes detrimental to their interests because they are morally committed to democratic institutions. Spontaneous compliance makes democracy more likely, since one should not expect too much morally inspired commitment to democracy in situations of ethnic violence. If the actors are self-interested and act strategically, spontaneous compliance requires the outcome of the political process to constitute a Nash equilibrium. In equilibrium, the individual actors' strategies are a best reply to each other, yielding a stable outcome because no one will regret his choice of strategy.

The expected life of institutions is longer than the expected effects of individual policies. Institutions thus create iterated games between political elites. More precisely, preferences over institutions can be considered to be preferences over streams of outcomes or policies because the institutions create repeated political situations (Tsebelis, 1990, p. 104). However, everything need not be unknown about the effects of the constitutional arrangements introduced in a transition situation. Indeed, knowledge of the kinds of outcomes different constitutions produce makes it possible for actors to have preferences over institutions. Institutions can not be neutral. On the other hand, the transition to democratic institutions is a one-shot game between the leaders of the armed groups. The transition phase includes the cease-fire, disarmament, and the first elections. Only when the parties have been disarmed and have elected their representatives, can the parties expect to play a repeated political game.

To lay bare the essential features of the choice situation facing the leaders of armed ethnic groups, I will assume that they only have the choice between a Democratic strategy and a Military strategy. The choice of one strategy by each actor leads to a jointly determined outcome that entails a payoff for each actor. In the case of mutual Democratic strategy, each player receives democratic compromise DC. In the case of mutual Military strategy, each player receives the civil war outcome CW. If one actor pursues the Democratic strategy and the other the Military strategy, the actor choosing the Democratic strategy receives the sucker's payoff S, and the actor choosing the Military strategy the temptation payoff T. Three different orders of preferences on democratic compromise, civil war, sucker and temptation emerge among the actors in the Angola and Rhodesia conflicts: T > DC > CW > S (preferences or payoffs like in Prisoner's Dilemma game); T > CW > DC > S (payoffs like in the Deadlock game); and DC > T > CW > S (payoffs as in the Assurance game). To get an initial impression of how these preference orders affect choices, let us consider each of the single-shot two-by-two game they create. In the one-shot Prisoner's Dilemma and Deadlock games, the Military strategy is dominant (better regardless of what the other actors does), and CW is the only equilibrium. In the Assurance Game, the players do not have any dominant strategy, and, consequently, there are two equilibria: mutual Military strategy leading to the CW outcome and mutual Democratic strategy leading to the DC outcome.

In one-shot games, the preferences of the actors determine the outcome. In the Assurance Game, the DC outcome is the socially optimal Nash-equilibrium both in the one-shot and the iterated version of the game. In iterated Prisoner's Dilemma and Assurance Game, on the other hand, the Democratic strategy is more likely to occur the greater the difference in payoffs between the DC and CW outcomes. The power of the backwards induction argument prevented understanding that the repeated Prisoner's Dilemma has different equilibria than one-shot Prisoner's Dilemma: Since the last round is known in advance, both players will choose the Military strategy as there is no future influence. Given this common knowledge, both players will choose the Military strategy in the penultimate round, and the possibility of democratic compromise will unravel in the same way until the first round.(2)

If the Prisoner's Dilemma game is played infinitely, or a large unknown number of times, the Democratic strategy may be chosen under complete information. First, if you know that your opponent will choose 'Always military strategy' as her strategy in the repeated game (for whatever reason, narrow-mindedness, paranoia, belligerence or believing in backwards induction), your best reply is 'Always military strategy,' and it thus forms a Nash-equilibrium with itself. Moreover, if your opponent plays 'Always democratic strategy,' your best reply is 'Always military strategy,' and the implication is that 'Always democratic strategy' does not form a Nash-equilibrium with itself. In fact, knowledge that the opponent will use any strategy that is not dependent on her own, makes 'Always military strategy' the preferred choice (Tsebelis, 1990, p. 75). However, if the players let their choice of strategy be contingent upon the other player's choice in the previous round(s), the Democratic strategy may be chosen. The best known contingent strategy is Tit-for-Tat, who chooses the Democratic strategy in the first round and in all later rounds of the game matches the opponent's choice in the previous round. Tit-for-Tat immediately punishes the Military strategy by choosing the Military strategy in the next round (Morrow, 1994, p. 260ff). To conclude, in one-shot games and repeated games without contingent strategies, the actors' preferences determine the outcome. With the exception of the Deadlock game, in iterated games with contingent strategies, the magnitude of the payoffs determine the likelihood of adopting the different strategies. To study the effects of iteration further, we need a discount parameter. The discount parameter expresses that unlike present payoffs, future payoffs are uncertain, and that even certain future payoffs may be of less worth for an actor than present ones.

The Democratic strategy is more likely to be supported as a Nash-equilibrium as: the value the players put on future payoffs increases, the reward from cheating decreases (T decreases), the cost of armed conflict increases (CW increases), the democratic compromise yields higher returns (DC increases), and the cost of attempting a transition to democracy increases (S increases). But contingent strategies have four limitations in inducing co-operation in repeated Prisoner's Dilemmas (Morrow, 1994, p. 267). First, 'Always military strategy' is always a sub-game perfect equilibrium for any discount parameter.(3) Second, reciprocal punishments may not be credible because the players are better off not playing the Military strategy. Third, in addition to 'Always military strategy,' there is a large number of other outcomes that satisfies the demands of an equilibrium solution, and no obvious way to choose between them. Finally, the players can not know which equilibrium they are playing. Without a common conjecture of what contingent strategy they are playing, the players may fail to co-ordinate on punishments that enforce co-operation.

Ethnic Conflict in Angola and in Zimbabwe

There are three main ethnic groups in Angola: the Mbundu, the Bakongo, and the Ovimbundu. After the Portuguese arrival in 1483, the three ethnic groups were contained in different political systems until the Portuguese finally established control over the whole territory as late as in 1915. In turn, all three ethnic groups traded members of the other groups to the Portuguese. At least two million slaves reached the New World from Angola, and it has been estimated that a similar number died in transit. To put these numbers into perspective, Angola's total population in 1983 was estimated to 8.3 million. The separate political systems of the three groups during more than four centuries, the extensive trading of slaves and the Portuguese policy of playing the groups off against one another, had the predictable result of creating deep ethnic conflict in Angola.

On the other hand, the three Angolan liberation movements, the MPLA (Movimento Popular de Libertação de Angola), the FNLA (Frente Nacional de Libertação de Angola, and the UNITA (União Nacional para a Independência Total de Angola), all had Angolan nationalist perspectives: All three vehemently denounced tribalism and regionalism, and none relied exclusively on the support of one ethnic group(4) Nevertheless, the conflict among the three movements increasingly assumed a de facto ethno-regional character during 1975 as competition fuelled the efforts to consolidate their core base of support. The MPLA had its main base of support among the 1.5 million Mbundu concentrated in the Cuanza valley and in Luanda. The FNLA mobilised the Bakongos in the north-west of Angola and the refugees in Zaire, about 1 million. The UNITA had its traditional support among the 2.5 million Ovimbundus in the south. The three main groups make up about 75% of Angola's population (Ovimbundu 37%, Mbundu 23% and Bakongo 14%). A large number of smaller tribes make up the remaining 25%.

In Rhodesia, the most important ethnic division was between Whites and Blacks. After the Unilateral Declaration of Independence in 1965, Rhodesian society was organised according to ethnicity. In 1980, almost 60% of the income was earned by the 4% White minority. The Land Tenure Act of 1969 formally divided the territory equally between the 4% Whites and the 96% Blacks; both groups got 46.6% of the land, while the remaining 6.8% was national land. In practice, the differences were even larger because most of the land allotted to the Blacks was held in a form of customary tenure (41.3% of the 46.6%) and areas open to freehold occupation by Blacks was small (3.8% of the 46.6%). In comparison, 40% of the White's 46.6% of the land had no restrictions, the remaining 6.5% were parks and wildlife reserves. Moreover, the land reserved for Whites included more than three-fourths of the highlands, and roughly two-fifths of the middleveld, the areas most suited for agriculture in Zimbabwe. The reason for reserving land for Blacks at all was to create the cheapest labour possible by forcing the families of workers into providing for themselves through subsistence farming, creating wages so low that they only provided for the workers. The White policy forced a restructuring of Black society and disruption of agricultural techniques. Economic and political inequalities and social upheavals created widespread resentment. Estimates of casualties in the war vary, but the most conservative indicate nearly 20,000 war-related deaths over the whole 1974-80 period. Losses rose sharply to over 1,000 a month at the peak of the fighting in 1979. Thus, ethnic relations between Blacks and Whites in Zimbabwe in 1980 were certainly bad.(5)

The Black population of Zimbabwe is dominated by two tribes, the Shona (75%) and the Ndebele (19%). The Ndebele is an ethnic category that grew out of a military state created in 1830-1840, encompassing peoples of different origins. In fact, Shonas living in conquered areas included into Ndebele society after 1840 constitute one of the largest components of the Ndebele. Ethnic identifications in both groups were fairly low, there was no history of enmity between them, and they even shared the history of the Ndebele-Shona Chimurenga resistance of 1896-97 against the British. Moreover, as in Angola, the two guerrilla movements appealed to Zimbabwean nationalism, and neither recruited exclusively from one ethnic group. Nevertheless, the ZANU (Zimbabwe African National Union) was associated with the Shonas and the ZAPU (Zimbabwe African People's Union) with the Ndebele.(6)

In conclusion, the most serious conflict among the six groups was the conflict between Whites and Blacks in Zimbabwe. The tripartite conflict among the Bakongos, Mbundus and Ovimbundus in Angola also ran very deep, but was nevertheless more benign than the former. The mildest ethnic conflict was between the Shonas and the Ndebele in Zimbabwe.

The Dependent Variable: The Fate of Democracy

On January 15, 1975, the three Angolan liberation movements signed the Alvor Accord setting the date of November 11, 1975, for Angolan independence. On this date the Portuguese would hand over power to a coalition government composed of representatives of the three liberation movements. Until November 11, the Alvor Accord provided for a transitional government where the liberation movements shared power with the Portuguese administration. The transitional government was to create an Angolan Army from the military forces of the three liberation movements. It was also to draft a provisional constitution, draw up an electoral law and register voters and candidates for the Constituent Assembly. Elections were to be held before the end of October, 1975. The Angolan transitional government thus was intended to perform two tasks simultaneously: draft the constitution and run the country.

The transitional regime, established in Luanda on March 28, 1975, soon fell apart. Military clashes between the three organisations became increasingly frequent during the spring of 1975, and from July on the situation deteriorated into full-fledged civil war. The Lisbon Government made a last attempt to restore its authority in July 1975 by sending in troop reinforcements, but the attempt proved feeble. In August, the Portuguese formally annulled the Alvor Accord and dissolved the defunct transitional government. Nothing was put in its place. At the date of independence, November 11, all liberation movements declared their own national governments, while the MPLA and the FNLA were fighting a battle outside the capital Luanda. The Portuguese withdrew without having surrendered power to anyone in particular.(7)

On October 22, 1979, an agreement on the constitution of the future Zimbabwe was signed at Lancaster House in London. The agreement was a constitutional compromise. The White Rhodesians, 4% of the population, got 20 of the total 100 seats in Parliament, while the Black Zimbabweans got the remaining 80.(8) The parties reached agreement on the transitional government in November and on the cease-fire arrangements late in December, 1979, securing that elections were carried out in March 1980. The elections resulted in an unexpectedly large victory for ZANU and Robert Mugabe. They won 57 of the 80 Black seats (with 63% of the votes cast), ZAPU won 20 seats (with 24% of the votes), while the white hope, UANC (United African Nation Congress), won only 3 seats (with 8% of the votes). The Rhodesia Front won the 20 White seats. Robert Mugabe became Prime Minister as Zimbabwe became independent on April 18, 1980. The new government immediately began to reorganise the armed forces, and the White minority thus soon lost control over their last instrument to overturn the new constitution and to reverse the outcomes of the political process. In Rhodesia, a peaceful transition to majority rule and democracy took place in the middle of a bloody war created by the most severe ethnic conflict, whereas in Angola the attempt at transition ended in war.

Democracy and the Struggle for Political Power

How did the actors rank the payoffs DC (democratic compromise), CW (civil war), S (sucker) and T (temptation)? For all of them, we can assume that the worst outcome was S, facing the Military strategy while pursuing a Democratic strategy yourself. Second, the actors probably preferred T > CW, getting autocratic power for free was better than fighting a war that may lead to autocratic power. To complete the ranking of the payoffs for each player, we need to draw on the ideology and behaviour of the six actors.

In Angola, the FNLA did not advocate democracy but an international class struggle interpreted in terms of racial concepts. Black people were the exploited proletariat, and white people were the exploiting bourgeoisie. The FNLA leadership and membership were almost exclusively of Bakongo background, so in practice 'Black' meant 'Bakongo'. The FNLA made an attempt to seize power in November 1974, but were persuaded by President Kenyatta of Kenya to return to the negotiations leading to the Alvor Accord. In February and March 1975, before the transitional government was established, FNLA forces reinforced with motorised Zairian units, moved into Angola from Zaire to attack the MPLA in Luanda, and broadened its attacks to MPLA forces outside the capital in April. The FNLA clearly preferred the Military strategy, delivering the principal blow in unsettling the possibilities of a compromise on democratic institutions. Since the FNLA counted on a swift victory, it is not possible to determine whether they preferred the mutual democracy outcome to the civil war outcome. Thus, we can not infer from the power struggle alone whether the FNLA had the preferences T > DC > CW > S (Prisoner's Dilemma) or T > CW > DC > S (Deadlock).

The MPLA was primarily committed to traditional international class struggle rather than democracy. The leadership had an urban character, consisting of blacks assimilated into Portuguese culture or persons with partially white parentage, and the movement emphasised Portuguese culture as a factor favouring national integration in Angola. The MPLA leadership was made up of persons that could expect to prosper individually under democratic competition. The MPLA did not carry out a bid for power before the Alvor Accord and abided by the agreement until attacked, suggesting that the organisation was willing to tolerate democratic political competition. On the other hand, no positive initiatives were forthcoming in the transitional government, indicating that the MPLA had democracy only as a second-best outcome, but better than civil war. Thus, the MPLA appears to have had the preferences, T > DC > CW > S (Prisoner's Dilemma).

The UNITA was essentially dominated by the undemocratic but charismatic personality of Jonas Savimbi, who appealed to ethnic and tribal Ovimbundu sentiments. He initially pursued a strategy aiming at political compromise, but willingly switched to the Military strategy when South African assistance tempted him with control of Luanda. Hence, democratic compromise was not Jonas Savimbi's best outcome but it was better than civil war, yielding the preferences T > DC > CW > S (Prisoner's Dilemma).

In Zimbabwe, Ian Smith often stated his opposition to democracy and repeatedly rejected real powersharing solutions from 1974 and onwards. The so-called internal settlement of March 3, 1978, the Rhodesia Front's own attempt to come to terms with the Black majority, gave the White population 28 of the 100 seats in Parliament, a blocking fourth for constitutional change under the new constitution. In addition, control over public service, police and defence forces remained on White hands, preserving the power to undemocratically alter policy outcomes ex post facto. However, in October 1979, the Rhodesia Front compromised, revealing the preferences T > DC > CW > S (Prisoner's Dilemma). But Ian Smith himself refused to budge and had to be removed by his own party. This incident has been interpreted in two ways. The first is that he was sincere in his wish to go down fighting rather than compromise, giving Deadlock preferences T > CW > CR > S (Stedman, 1991). The second is that this was tactical ploy where he counted on being removed, allowing himself to be seen fighting to the bitter end, giving Prisoner's Dilemma preferences (Tamarkin, 1990).

The war in Rhodesia was about majority rule but not necessarily about democracy since an undemocratic Black regime was a possibility. Both ZAPU and ZANU officially endorsed democracy as a solution to end the war. The split between ZAPU and ZANU in 1963 had occurred over whether a Military strategy should be adopted, ZANU breaking away to begin the armed struggle. Nkomo tried to negotiate a separate agreement with Ian Smith both in 1976 and 1978, and the attempts only failed because Smith was not prepared to accept majority rule. Nkomo adopted the Military strategy only when compromise failed, and when military inactivity had become a political liability. Interestingly, Nkomo's did not opt for guerrilla warfare, building a conventional force to deliver Salisbury the final blow (Tamarkin, 1990, p. 100). It was no coincidence that Nkomo's military strategy led to very little actual fighting. Assurance Game players will opt for the Military strategy if others choose the Military strategy, but Nkomo's first preference remained the CR outcome, yielding the preferences, CR > T > CW > S (Assurance Game).

Robert Mugabe did want majority rule in Zimbabwe, but ideologically he believed that allowing multiple parties might cause ethnically diverse African countries to fall apart. He never thought that Ian Smith would yield on the crucial issue of majority rule before the military situation was ripe, and considered negotiating with him politically harmful. But Mugabe's reluctance did not stem ideology or principle, as one diplomat stated "He believed in armed struggle, because of Smith.(9) ZANU was militarily stronger and more active than ZAPU in the guerrilla war, suffering large losses -- 7000 dead in 1979 alone out of a total force of about 50,000. The morale of its forces was unbroken and Robert Mugabe was on his way to military victory, but he was not ideologically opposed to a settlement (Stedman, 1993, p. 138). His preferences were thus T > DC > CW > S (Prisoner's Dilemma).

Democracy and the Struggle for Economic Benefits

To say that Holden Roberto headed the FNLA is an understatement; he was the FNLA. The FNLA had virtually no fixed organisational structures. Roberto's style of leadership was entirely based on personal ties. His great grandfather and his maternal grandfather who had headed Bakongo nationalist exile groups in Belgian Congo (Zaire), and he divorced his first wife to marry the sister of Zaire's Mobuto Seke Seko, who repeatedly intervened military on Roberto's behalf during 1975. Personal patron-client relationships consist of the resources of patronage, combined with loyalty transcending mere interest, yet always remaining conditional. Before independence, Roberto depended upon foreign support to provide the necessary resources, and his attempts to secure it was the central motive in his policy from the late 1950s.(10) After independence, control of the main dispenser of patronage, government, was within his reach. FNLA was not a unitary actor in a trivial sense because of the game between leader and followers, but the organisation had the ability to formulate collective interests and to act strategically to pursue them. Since his demands on patronage was incompatible with the uncertainties of democracy, Roberto's preferences were T > CW > DC > S (Deadlock) and not T > DC > CW > S (Prisoner's Dilemma).

There were also certain elements of personal and family ties in the position of MPLA's central leader, Agostinho Neto, but far less pronounced than in the FNLA. The MPLA was mainly based on people who owed their position to how they filled a role within the organisation. In addition, the military organisation was part of a larger, formalised structure (Marcum, 1978; Marcum, 1987b, p. 18). UNITA also had an effective organisation, depending on Savimbi for charismatic leadership more than patronage. Like Holden Roberto, the leaders of the MPLA and UNITA preferred autocratic power to democracy, but since their leadership did not depend on patronage, the uncertainties of democracy was tolerable.

In Zimbabwe, both the ZANU and the ZAPU had a reasonably strong formal organisational infrastructure. Neither organisation depended on patronage for its existence. They were able to field large, effective military forces. The ZANU had about 50.000 men under arms in 1980, and they did not need external aid to maintain the organisations as such.

For the Whites, a democratic compromise was fully compatible with retaining economic stature. The possibility of preserving the White "way of life" under majority rule gave the Rhodesians incentives to compromise. Even the Sucker outcome might be acceptable since the economy would collapse if their property was confiscated. No future ruler of Zimbabwe could want to eliminate the most productive sector of the economy.

The Angola and the Rhodesia Game

The 'Angola game' was a three-person-game, where the MPLA and UNITA had Prisoner's Dilemma preferences and the FNLA had Deadlock preferences:

If FNLA chooses DS,

UNITA chooses
MPLA DS 2,3,3  1,1,4
chooses MS 1,4,1 1,2,2

If FNLA chooses MS,

UNITA chooses
MPLA DS 4,1,1 3,1,2
chooses MS 2,3,1 3,2,2*

The Deadlock player FNLA rejects the Democratic strategy (DS) in the Angola game, both when it is played once, and when it is reiterated. Given that choice, the MPLA and UNITA will choose the 'Always military strategy,' too, and the choice of (MS,MS,MS) results in the CW outcome with payoffs (3,2,2), forming the unique solution (*) regardless of the number of repetitions.

The 'Rhodesia Game' was a three-person game where the RF and ZAPU had Prisoner's Dilemma preferences, and ZAPU Assurance preferences:

If RF chooses the DS strategy

ZAPU chooses
ZANU DS 3,4,3 1,1,4
chooses MS 1,3,1 1,2,2

If RF chooses the MS strategy,

ZAPU chooses
ZANU DS 4,1,1 2,1,2
chooses MS 2,2,1 2,2,2*

If the Rhodesia game is played once, the RF and ZANU have the Military strategy (MS) as their best strategy regardless of the choice of the others. Furthermore, given that RF and ZANU chooses the Military strategy, ZAPU's best choice is to choose the Military strategy, too. Thus, the only Nash-equilibrium and the solution of if the game is played once is (MS,MS,MS) with payoffs (2,2,2).

In the iterated game, any individually rational outcome can be supported as an equilibrium outcome. Since actors with Prisoner's Dilemma and Assurance game preferences can guarantee themselves at least the Civil War outcome regardless what the others do, any outcome improving on that is in equilibrium. There are indefinitely many such outcomes, because mutual co-ordination to democratic compromise may occur (or unravel) at any given point in time. Let us see whether the magnitude of the payoffs T, CW, DC and S, influenced the outcome in the repeated political game under the constitution.

The Choice of Strategy in Iterated Games: The Magnitude of the Payoffs

The payoffs associated with the Sucker and Temptation outcomes are disregarded because they were constant. Two factors stand out in shaping the magnitude of the payoffs of the CW and DC outcomes in Angola and Zimbabwe; relative military strength, and relative electoral strength. Estimates of the military forces of the three Angolan movements at the time of the Alvor Accord vary, but all agree that the FNLA was numerically superior. Klinghoffer (1980, pp. 15-17) estimates that the FNLA had 10,000 men under arms, the MPLA 6,000 and UNITA about 2,000, while Marcum (1978, p. 257) estimates numbers at 21,000 (FNLA), 8,000 (MPLA) and 8,000 (UNITA). The military situation was complicated by a break-away faction of the MPLA headed by Daniel Chipenda that joined forces with the FNLA in February 1975, and a Katangese force that joined the MPLA in April 1975 because of the FNLA's close links with Zaire. Estimates of these forces also vary, the Klinghoffer estimate gives the figure 2,000-3,000 for the Chipenda force, the Marcum estimate says that the Chipenda force numbered 2,000-2,750 men and the Katangese force 3,500-6,000. On July 9, 1975, heavy fighting broke out involving all three Angolan movements. Although superior in numbers and armaments, the FNLA forces and leadership proved completely inept. Even with more support from Zaire and collaboration between FNLA and UNITA, the MPLA soon gained the upper hand.

The assumed military strength of the three Angolan liberation movements did not mirror their base of support. The movement with the largest ethnic base, the UNITA, was the militarily weakest. The movement with the smallest ethnic constituency, the FNLA, had most men under arms in the beginning of 1975. Consequently, the FNLA stood to lose the most in a democratic system, and the UNITA stood to gain. The UNITA initially advocated settling the conflict through elections.(11) The FNLA tried to win power in Angola by military means, a choice of strategy determined both by Deadlock preferences and by being stronger militarily than electorally. The MPLA, the only actor with a balance between military and electoral strength on the Angolan scene, did not initially play the Military strategy, but on the other hand never considered the Democratic strategy when violence had broken out. The unstable military situation increased the payoffs of the civil war outcome to the MPLA in July 1975, and to the UNITA in October 1975.

The war in Zimbabwe was a typical guerrilla war. The Rhodesian Army could move wherever it wished, but was unable to prevent the guerrillas from flowing into Rhodesian territory. The Rhodesian Forces had high morale and were well trained, enabling them to hold the guerrillas at bay though rarely fielding more than 4000 men. Total forces, all units and reserves included, were about 46,000 men (Rinehart, 1983, pp. 57-58; Butts, 1990, p. 28). Their kill ratio advantage over the guerrillas was up from 10 to 1 in 1974 to 14 to 1 in 1979 (Stedman, 1991, p. 74). The Rhodesian Army remained a formidable fighting force but could not win the war. The burden of financing the war was crippling expenditures classified as "war costs" accounted for 41% of total public spending in Rhodesia in 1979 (Rinehart, 1983, p. xxiv). The main asset of the Rhodesian Front in the negotiations was that the guerrilla force was too weak to succeed in a conventional attack on Salisbury. Mugabe and Nkomo (and their military advisers) had different beliefs about the military situation. Mugabe believed that he was winning the war and would defeat the Salisbury regime shortly. Nkomo, on the other hand, also believed that they would eventually win but at a prohibitive high cost, because guerrilla warfare could not bring victory, necessitating a conventional attack at Salisbury (Stedman, 1993, pp. 136-137). In an ambiguous military situation, because it ultimately rested on White reactions to economic hardship in an unwinnable war, the parties preferences influenced their assessment.

From 1974 to about 1978, the Rhodesia Front showed no inclination to compromise during the frequent negotiations. The Constitution adopted in March 1970 reserved 20 seats for Blacks and 50 for Whites in a parliament of 70. The Black seats fell well short of the blocking third needed to prevent constitutional change. The combination of military preponderance and an extremely narrow ethnic base explains the reluctance to accept democracy. When the military situation worsened in 1978-79, reducing the payoff of the Civil War outcome, the expected payoff of the democratic compromise was increased by a glimmer of hope that white overrepresentation and an alliance with the moderate Blacks would yield a sufficiently large constituency to preserve some political influence. The elections in April 1979, carried out under the internal settlement, had given the moderate UANC headed by Bishop Abel Muzorewa 51 of the 72 seats reserved for Blacks (62.3% of the votes), and the turnout was 64.45% of the electorate.(12) The guerrilla movements had boycotted the elections, and their strength at the ballot box was unknown. It has even been suggested that the RF and South Africa accepted the outcome of the elections only because ZANU's victory was a landslide, eliminating the possibility of crying foul play and shattering illusions of Black support for White rule. However, Zimbabwe is above all an interesting example of a conflict where a party was willing to settle in spite of having no chance of ever winning an election. The Rhodesia Front faced the choice between a certain military defeat in the long run or a democratic constitution with a hope of a continued say in Zimbabwean politics.

Table 1. Preferences, relative military and electoral potential and initial choice of strategy.
Preference Prisoner's Dilemma Prisoner's Dilemma Assurance Deadlock Prisoner's Dilemma Prisoner's Dilemma
Military Potential Above average Above average Below average Above average Average Below average
Electoral Potential 4% 


75% Shona 19% Ndebele 14% Bakongo 23% Mbundu 37% Ovimbun.
Mil/Elec. Balance Stronger Military Balanced Balanced Stronger military Balanced Stronger electoral
Initial Strategy  Military  Military  Democratic  Military  Democratic  Democratic

To conclude, the fortunes of war and electoral prospects strongly affected the actors' initial choice of strategy. Moreover, changes in the fortunes of war, i.e. the payoffs associated with the CW outcome, goes a long way to explain the changes in strategy among the actors during the conflict. In Angola, there was no costly war creating incentives to reach agreement when the transitional government began to negotiate agreement on a constitution. In Zimbabwe, on the other hand, the Lancaster House Conference was carried out during the bloodiest phase of the civil war, creating powerful incentives to end the fighting (Zartman, 1989, p. 260). Generally, the more destructive the conflict, the greater the incentives to end it. Let us consider the propositions of the game-theoretic analysis for a more complete understanding of the actors' choices between the Democratic and the Military strategy.

Discount Parameters
Leaders of armed ethnic groups operate in a highly unstable political environment, forcing them to concentrate on activities leading to immediate gains in security. There is a high probability of political marginalisation before democracy arrives, and consequently, the discount rate for future benefits that democracy may accrue is high. In Angola, the volatile military situation in 1975 resulted in an extreme version of this problem, rendering a negotiated transition to democracy impossible. The military situation was more stable in Rhodesia, creating an environment for prolonged negotiations. In addition, the White Rhodesians found the probability that their property would be confiscated after the elections low, because they knew that it would lead to economic disaster. Thus, promises to respect property rights were credible, lowering the White discount rate. But the uncertainty of future rewards proved a real problem also in Zimbabwe. According to one diplomat, "Mugabe was convinced that he would win an election, but was unsure he would have a chance to win an election."(13) Note that white economic privileges not only made the democratic compromise outcome more attractive, but, more importantly, offered a credible guarantee of long-term benefits for a group surely facing long-term political marginalisation. The economy would collapse if their property was confiscated, and no future ruler of Zimbabwe could want to eliminate the most productive sector of the economy. The electoral privileges that were to expire in 1990 were unconstitutionally removed in 1987-88, when Robert Mugabe abrogated the White minority constitutional rights ahead of schedule. However, White economic dominance has deterred any action so far. White economic privileges made the Democratic strategy rational even if the Whites were to become a permanent minority without any say in the iterated political game or if the iterated political game is cut short by the winners of the first elections, making it difficult to generalise this aspect of the Zimbabwean experience to cases where the parties can only be rewarded by the political game.

Agreement in Iterated Games and in One-shot Games
Lord Carrington's tactics at Lancaster House was to keep the issues strictly separated. The first issue was the constitution, then the transitional arrangements, and finally, the cease-fire. After six weeks of hard negotiations, ZANU and ZAPU, who had united politically under the Patriotic Front umbrella for Lancaster House, accepted the constitution on September 18, 1979. Agreement was reached with Lord Carrington fully in control of issues and proposals, in spite of many Patriotic Front attempts to wrestle the initiative from him (Davidow, 1984, pp.61-61). Moreover, no ultimatum from the Front Line presidents was necessary to make the Patriotic Front accept the Constitution, even with property rights enshrined in the Constitution: Land had to be voluntarily sold and paid for at once with full, market-level compensation. This provision made land reform impossible, setting aside the most important political issue in Zimbabwe.

On November 12, during the negotiations on transitional arrangements, Carrington had tomake public that there would be a Commonwealth monitoring force to supervise the cease-fire. Although Carrington had been "loath to discuss the cease-fire before an agreement on the transition," he had to relinquish control in order to make it easier for the parties to see the coming advantages that would result from agreement on the transition (Stedman, 1991, pp. 195). A transition agreement was reached on November 15. On November 26, during the life-and-death issues of the cease-fire, Carrington issued an ultimatum on the British proposals, prompting Mugabe to fly to Dar Es Salaam to meet the Front Line presidents. "I am not going to stand for my forces being herded like cattle into these detention centres at the mercy of the Rhodesian army and air force," Mugabe said (Smith & Simpson, 1981, p. 141). The Front Line presidents persuaded Carrington to make concessions, including close monitoring of the Rhodesian forces, particularly the Rhodesian air force. This was the only time Carrington had to yield from arbitration during Lancaster House, but he still saw no need to prolong the cease-fire (Stedman, 1991, p. 199). On December 16, after having extracted the concession of a sixteenth assembly point 'somewhere in the centre of Rhodesia,' Nkomo was satisfied, but Mugabe still balked, for reasons that have puzzled analysts. The physical security of his troops was not an issue anymore. The military leader of ZANU's forces, General Tongogara, had been less concerned than Mugabe about the location of the assembly points, knowing that his men were close to the border in case the cease-fire should fall apart. It has been suggested that Mugabe mistrusted the British; that he was fearful of the designs of the Rhodesian military, that he was convinced that his forces would eventually triumph militarily, and that he may have been afraid for his life and of betraying the trust his people had invested in him. In the end, Mozambican President Machel, through his representative Honwana, allegedly told Robert Mugabe that if he did not sign he would be given a nice house on the beach in Mozambique (Davidow, 1984, p. 89).

Why was an ultimatum from the man who controlled his guerrilla bases necessary to make Mugabe sign the cease-fire agreement? After all, Mugabe had compromised on the more important issue of land reform in the constitution without any ultimatum from Machel. The explanation for Mugabe's reluctance is simple: The Constitution created a repeated game, in which the Democratic strategy was rational. The cease-fire happened only once, and Mugabe's dominant strategy was the Military strategy. Indeed, Machel had to remind that "he would win the elections" (Davidow, 1984, p. 89). In Zimbabwe, it proved more difficult to reach agreement about the transitional government and the cease-fire conditions than the Constitution, nicely illustrating that in the one-shot Prisoner's Dilemma game, preferences determine the outcome, in contrast to the repeated versions of the game where the magnitude of the payoffs influence the choice of strategy.

Let us now see how the actors dealt with the four limitations of contingent strategies in inducing co-operation in an iterated Prisoner's Dilemma for a more complete understanding of the conditions that affected the choice between a Democratic strategy and a Military strategy, i.e. how did they overcome that 'Always military strategy' is always a sub-game perfect equilibrium; how did they made reciprocal punishments credible; how did they choose between the many possible equilibrium outcomes; and how did they find what equilibrium superstrategy they were playing to co-ordinate punishments to enforce co-operation?

'Always Military Strategy' Is a Sub-game Perfect Equilibrium for Any Discount Parameter
In Zimbabwe, both the Rhodesia Front and the ZANU moved from the Military strategy with great reluctance. The Rhodesia Front only accepted the Democratic strategy when the military defeat appeared imminent, even though it was a certainty in the long run from the moment the war began in earnest in 1974. The ZANU, who was on its way to victory in 1979, would neither have entered nor compromised in the Lancaster House negotiation, had it not been for pressure from the Front-line states, in particular Mozambique. In Angola, both the MPLA and UNITA initially hesitated somewhat in choosing the Military strategy, but once they had done so they stuck firmly to it.

Contingent Strategies Need Credible Threats
As an Assurance Game player, Joshua Nkomo's first choice was democracy, preferring electoral defeat to military victory. The relatively benign conflict between the two ethnic groups assured the Ndebele individual social mobility and respect for their group identity if they lost the elections. But of course he preferred a prominent political position in a democratic Zimbabwe. Nkomo realised that ZAPU had a much narrower ethnic constituency than ZANU, and his strategy was to win the elections by being the man to introduce majority rule in Zimbabwe, bending over backwards in his attempts to negotiate democracy in 1976 and 1978, until Ian Smith had demanded so many constraints on democracy that it no longer was democracy. ZANU and ZAPU had retained separate military forces also when united politically in the Patriotic Front during the negotiations in 1979, and Mugabe was so confident of victory that he split the Patriotic Front before the elections. Joshua Nkomo evidently hoped that the Patriotic Front umbrella would be used in the 1980 elections, but he suspected all along that Mugabe might jettison him, so his endorsement of democracy was not due to misperceptions (Stedman, 1991, p. 202).

Ironically, Nkomo's moderation deprived him of a credible threat to induce the RF tocompromise, and they turned him down both in 1976 and 1978. Nkomo adopted the Military strategy only when compromise seemed impossible, and when military inactivity had become a political liability. Assurance Game players will opt for the Military strategy if others choose the Military strategy. Interestingly, Nkomo's did not opt for guerrilla warfare, building a conventional force to deliver Salisbury the final blow (Tamarkin, 1990, p. 100). It was no coincidence that Nkomo's Military strategy led to very little actual fighting. Consequently, Nkomo's threat lacked credibility as a punishment to force the RF to a democratic compromise (Stedman, 1991, p. 69). It was the Prisoner's Dilemma player, Robert Mugabe, with forces obviously willing and able to inflict damage who could credibly threaten the Rhodesia Front to abandon the Military strategy. A contingent strategy needs a credible threat of continued conflict to induce democracy. This explains the paradoxical situation where the most conciliatory actor can not get a deal.

Moreover, it was not a coincidence that the Lancaster House Conference was carried out during the bloodiest phase of the war, nor that two escalatory thresholds were crossed during the last months of the war: In mid-October 1979, Rhodesian forces, for the first time, attacked economic targets in Zambia and Mozambique; in the third week of October 1979, Mozambique began hot-pursuit raids into Rhodesia after Rhodesian attacks. The purpose of both escalations was to demonstrate the ability and willingness to punish defections. Middle-ranking Rhodesian officers thought agreement at Lancaster House more likely if the credibility of the Rhodesian Army as a future threat was maintained (Stedman 1991:227-230). The Mozambican raids were carried out after agreement had been reached on the Constitution, and just before President Machel threatened Mugabe with political oblivion if he did not sign the cease-fire agreement. It was the Salisbury regime, the militarily weakest actor, and Mozambique, who was relatively less motivated than the Zimbabwean nationalist, that felt the need to demonstrate military prowess and resolve.

Equilibrium Outcomes
Lord Carrington gained control over the agenda by preserving the privilege to present proposals. Moreover, he and his team engaged in a kind of shuttle diplomacy; the actors sat in separate rooms, bargaining with Carrington rather than with each other. This minimised the problem created by the many equilibrium solutions, because all actors were forced to concentrate on the one and same solution. Note that this does not necessarily mean that the parties at Lancaster House had a wide zone of acceptable solutions. Roughly the same amounts of utility may be represented in many concrete ways, giving considerable leeway even with the thinnest of acceptance zones. Davidow (1984:110) argues that Carrington's tactics enabled him to obtain concessions from each party that they would not otherwise have granted, but more importantly, the parties were able to converge on an acceptable outcome in only four months.

Equilibrium Strategies
The need for a shared conjecture about what superstrategy they were playing proved to be no obstacle in Zimbabwe. The escalations during Lancaster House, the Rhodesian bombing of economic targets outside Rhodesia and the Mozambican hot-raid pursuits into Rhodesia, was interpreted as a round of Military strategy used for punishment within an overarching Democratic superstrategy and not as, say, the first move in 'Always military strategy' because they could not win the war for either party. In Angola, on the other hand, the precarious security situation for all concerned forced the interpretation "Always military strategy" on any use of the Military strategy because even minor gains could tilt the balance to either side.

Implications for Transition to Democracy

In addition to serving as a plausibility probe for deductive propositions derived from game theory, the Angola and Zimbabwe cases are an empirical source for inductively finding mechanisms that facilitate or hinder transitions to democracy. The purpose of the following section is to discuss the implications of these mechanisms.

The consent of all actors is necessary for a transition to democracy to take place. Prisoner's Dilemma players who are relatively strong militarily and relatively weak electorally will reject democracy. The logical implications is that to avoid having one actor reject democracy the actors must operate in an environment of balance between relative military and electoral strength. That balance may come about in different ways: In Zimbabwe, ZANU and ZAPU were both balanced: the former was strong both militarily and electorally while the latter was weak on both scores. In Angola, only the MPLA was balanced with medium strength both militarily and electorally, making the situation difficult for a transition to democracy.

Imbalances between electoral and military force among the actors may be redressed either by manipulating the electoral potential through the electoral system or by manipulating military strength. The ultimatum the Front-line States gave Robert Mugabe and South Africa's support of the Lancaster House agreement were crucial in manipulating the military balance but had no long term effects. The other way is to increase the payoffs from participation in democracy, for example the twenty seats reserved for Whites in Zimbabwe in a ten year transition period, has long term effects.

The separation of substantive and constitutional issues. In Angola, the actors confronted both constitutional and substantive issues in the joint transitional government. The way the Angolan process was played out reveals two mechanisms. First, mixing substantive and constitutional issues made for a double one-shot game, preventing agreement on the constitution. Second, the intrusion of substantive issues in the constitutional process influenced the attitudes of the actors. Instead of bargaining and arguing in the constitutional process, only bargaining took place. Arguing means that the one with the best arguments prevails, whereas in bargaining the one with the most material resources prevails. Since all actors controlled military forces, bargaining meant civil war. The successful Lancaster House Conference had no administrative duties and avoided these pitfalls.(14) The process in Zimbabwe also reveals that separating substantive and constitutional issues enabled both parties to use substantive issues to enhance the credibility of future punishment that deterred the choice of the Military strategy and promoted the choice of the Democratic strategy. Thus, in Angola, the escalation of the war undermined the negotiations, whereas in Zimbabwe the escalation of the war was crucial in reaching agreement.

The role of credibility in the transition process can be understood by contrasting the efforts of the Portuguese troops in Angola with the efforts of the Commonwealth Monitoring Force (CMF) in Zimbabwe. The CMF was an incredibly small, only thirteen hundred men. Both the Rhodesian Forces and the Liberation movements disposed of far larger numbers some 50,000 to 60,000 men each. The Governor during the transition, Lord Soames, lacked the means to force either side to comply with the agreement. The forces at his disposal were in fact so small that he had to use his brief authority as formal head of state to call in some Rhodesian forces to supervise the disarming of the guerrillas. However, the CMF entered Zimbabwe when there was agreement on a Constitution creating an iterated political game between the parties. It had a clearly defined plan of how they should monitor the cease-fire and disarm the belligerent parties. There were 16 assembly areas, a pattern for deploying the force, and a time-table.

According to Stedman, "The British plan can best be likened to transforming the conflict into a one-shot game of chicken: there may have been incentives to cheat and seek advantage of the cease-fire to cheat, but the costs of total breakdown were prohibitive" (1991, p. 208). I disagree because a one-shot game of chicken is not affected by the magnitude of the payoffs associated with the outcomes. In a game of Chicken, both players have an incentive to be the first to move to the Military strategy, presenting the other player with the choice between the S and CW outcome, obviously leading to the transition to democracy to fail. The importance of the CMF was that it transformed the transition and cease-fire process into a repeated game. First, the task of disarming the actors were deliberately left out of the mandate of the CMF, the force only carried out the task of geographically distancing the parties (Ginifer, 1995, pp. 52-55). Thus, both parties retained their arms and their ability to deter defection from the peace agreement. ZANU had faced a dilemma in the locations of the assembly points: Centrally located assembly points were an advantage in the election campaign, but, on the other hand, it was easier to escape across the border from peripherally located ones. But ZANU's military leader, General Tongogara, had made sure that sufficient numbers of guerrillas would escape to deny Rhodesia Front military victory, and he was able to keep some of his best fighters out of them altogether. The Rhodesians fear that sufficient numbers of guerrillas would turn up to stay at the assembly points disappeared when Governor Soames called in Rhodesian forces and police to monitor them. The geographical space between them reassured against a surprise attack, preserving their deterrence capability. Second, Lord Soames attributed the success to the fact that the neutral troops were located with the opposing forces of either side, not between them (Soames, 1980, p. 413). The parties could neither leave their own assembly areas nor attack the enemy's without getting into conflict with the CMF. The CMF was a hostage to guarantee that parties reneging on the Lancaster House agreement needed to take long term external hostility into consideration as well as the admittedly limited immediate costs of taking on an additional 1,300 men. The merger of the three forces into a new army did not go smoothly, in March 1981 there was a mutiny and fighting between ZANU and ZAPU elements, but the situation stabilised (Berdal, 1996, p. 55).

Third, the CMF was capable of some coercion. For example, one ZANU candidate was banned from the elections, and the CMF did apply force in order to keep things as calm as possible during the elections. On the other hand, when the question arose of banning Robert Mugabe's party from the elections because of alleged intimidation, Lord Carrington aptly summarises the position of Lord Soames, "It [banning Mugabe] was obviously an option, but it was not an option" (Charlton, 1990, p. 143). Thus, the Third Party contribution to the peaceful transition to democracy in Zimbabwe had three elements. The transition was credibly transformed from a one-shot game to a repeated game by gathering but not disarming the parties and by risking 1,300 hostages to guarantee that reneging the agreement would not be forgotten by the outside world. In other words, spontaneous compliance was carefully nurtured. Finally, the CMF added some enforced compliance during a critical phase of the transition to democracy.

The Portuguese force of 24,000 men was the militarily strongest force in Angola during thewhole of 1975. In contrast to the Commonwealth force, the Portuguese could easily have withstood the combined military forces of the three Angolan liberation movements (Marcum, 1978, p. 255). The force was responsible for running the country and monitoring the cease-fire during the transition period until the new government had been elected. The Portuguese military commanders lacked a sense of purpose and direction on how to achieve these general objectives. The Portuguese commanders only wanted to withdraw from their own lost cause in Angola, a desire that did not lead them to create a plan themselves. To be effective, external military forces need a concrete political plan creating an iterated game between the local belligerents in place. Their own role has to be reduced to reassure the parties that their mutual deterrence does not enable any party to defect, while adding a little deterrence of their own. The fate of the Portuguese forces in Angola is a reminder that the organisation of the transition, not credibility and numbers, is the key to success, although the willingness to commit large numbers of troops can of course be a way to gain credibility.

The transition is a one-shot game where the magnitude of the payoffs of the alternative outcomes does not influence the outcome, meaning that the terms of the transition does not influence whether the probability of successfully implementing the agreement. The use of Rhodesian armed forces and police, under CMF supervision, was tolerable to the guerrillas. The problem is that terms rarely are purely transitional. For example, the location of the assembly point had long-term influence because ZANU used guerrillas for Mao-style political mobilisation, and lack of assembly points located in central Zimbabwe hurt ZANU's election campaign. Another example is the Patriotic Front's demand of power-sharing and a six months transition period to establish themselves in the competition with Muzorewa's UANC. When Muzorewa and the Rhodesia Front would only accept a shorter transition period without any participation in the government, they also unintentionally favoured ZANU vis-à-vis ZAPU. Nkomo needed the extra time and the status in the election campaign more than Mugabe, whose guerrillas had liberated and mobilised large parts of Shonaland long before the transition.


The different fates of the attempts to solve the conflict in Angola and Zimbabwe are neither explained by the intensity of the ethnic conflict, nor by the actors' preferences about democracy, nor by their political identities. Political outcomes such as transitions to democracy are influenced by but not determined by social forces. Societies with the same amount of ethnic strife, in which the leaders continue to have the same preferences for belligerence, may choose different political alternatives. The chance that the leaders of armed ethnic groups choose democracy increases if constitutional issues are kept separate from substantive ones. Furthermore, if democracy is an iterated Prisoner's Dilemma, each player must succeed in striking a balance between the need for establishing a credible threat to punish future defections from democracy and the need to alleviate the fear of defection from democracy. Third Party intervention may thus aid the parties to achieve spontaneous compliance during the transition, but is no substitute for it.


1.  Thanks to Dag Anckar, Dag Harald Claes, Jan-Erik Lane, Christopher R. Mitchell, Inga Marie W. Nyhamar, Bjørn Erik Rasch and Anne Julie Semb for valuable comments that by no means make them responsible for the contents of the article.

2. Therefore, if the number of rounds is known, both players will play 'All MS'. Playing DC requires incomplete information about the payoffs of the other player, but uncertainty about whether your opponents knows that you know that she knows that one of the two is rational, is sufficient to generate co-operative strategies in equilibrium if the game has a finite number of repetitions (Kreps & Wilson, 1982; Cohen, 1994, pp. 49-52).

3.  A sub-game perfect equilibrium consists of strategies that are not only a Nash equilibrium for the whole game but also are a Nash equilibrium for all sub-games. The whole game in this case is a series of Prisoner's Dilemma games, and since MS is the dominant strategy in Prisoner's Dilemma K must be a sub-game perfect equilibrium. Whereas Nash equilibria allow players to make non-credible threats provided that they never have to carry them out, sub-game perfect equilibria eliminate such solutions from the game. Such Nash equilibria are based on behaviour off the equilibrium path, which prevent them from being established with the stability desirable to count as a solution of the game. For more on sub-game perfect equilibria, see Hovi & Rasch, 1993, pp. 71-72 and Morrow, 1994, pp.128ff).

4.  For analyses of nationalist ideology of the three movements, see Marcum, 1978 and Legum (ed.), Africa Contemporary Record 1975-76 p. B422ff.

5.  For social and economic conditions in Rhodesia before 1980, see Lan, 1985, pp. 121-123, Stoneman & Cliffe, 1989, p. 124; and Tamarkin, 1990.

6.  Regarding ethnic relations in Zimbabwe, see: Rinehart, 1983; Kaplan, 1983; Lan,1985, pp. 217-228; Chainawa,1985, pp. 202-219; Stoneman & Cliffe (1989 nationalist ideology of the three movements, see Marcum, 1978 and Legum (ed.), Africa Contemporary Record 1975-76 p. B422ff.) and Ngcongo, 1989, pp. 106-111.

7.  For the text of the Alvor Accord, see Legum, Colin (ed.) Africa Contemporary Record 1974-75, p. B221-226. For analyses of the events in Angola in 1975, see Legum, Colin (ed.) Africa Contemporary Record 1975-76, p. B421ff; Marcum, 1978; and Klinghoffer, 1980.

8.  Legum, Colin (ed.), Africa Contemporary Record 1979-80, p. B977.

9.  British Diplomat A. 1987. Interview with Stephen John Steadman (Stedman, 1993, p.138).

10.  I draw on the description of the beginnings of Angolan nationalism in Marcum, 1969. On Roberto's support over the years from the US, Great Britain, various other Western European states, some Eastern European states, the Soviet Union, various Arabic states and other African states, see Marcum, 1978, p. 122.

11.  UNITA's preferences over strategies have changed repeatedly, but its preferences about democracy have not. In 1992, Savimbi preferred elections to continued military stalemate, clearly counting on winning the elections. However, although the largest ethnic group in Angola, the Ovimbundus remain a minority. The UNITA lost the elections to the MPLA, advocating a more encompassing Angolan nationalism. When the hope of electoral victory had been substituted with electoral defeat, and the military balance changed because the MPLA had disarmed while the UNITA had not, Jonas Savimbi adapted to the new military situation by switching back to the Military strategy.

12.  Legum, Colin (ed.), Africa Contemporary Record 1978-79, p. B996-B977.

13.  British Diplomat A. 1987. Interview with Stephen John Steadman (Stedman, 1993, p. 138).

14.  Elster, 1993a, p. 15), Elster, 1993b, pp. 169-217. The problems with substantive and constitutional issues are discussed in Przeworski, 1991, p. 79ff; Colomer, 1991, pp. 1283-1302; and Colomer, 1995, pp. 74-85).


Anderson, B. 1990. Imagined Communities: Reflections on the Origins and Spread of Nationalism. London: Verso.

Berdal, M. 1996. Disarmament and Demobilisation after Civil Wars, Adelphi Paper 303. Oxford: Oxford University Press.

Bridgland, F. 1988. Jonas Savimbi: A Key to Africa. London: Coronet Books.

Butts, K. 1990. "Zimbabwe's Low Intensity Conflict: Background and Implications." Paper presented at the 31st Annual Convention of the International Studies Association, April 12 in Washington.

Chainawa, D. 1985. "African Initiatives and Resistance in Southern Africa."In A. Adu Boahen (ed.), General History of Africa VII: Africa under Colonial Domination 1880-1935. Oxford: Heinemann International, pp. 194-220.

Charlton, M. 1990. The Last Colony in Africa: Diplomacy and the Independence of Rhodesia. Oxford: Basil Blackwell.

Colomer, J. M. 1991. "Transitions by Agreement: Modeling the Spanish Way." American Political Science Review 85, 1283-1302.

Colomer, J. M. 1995a. Game Theory and the Transition to Democracy: The Spanish Model. Aldershot: Edvar Elgar.

Colomer, J. M. 1995b. "Strategies and Outcomes in Eastern Europe." Journal of Democracy 6:2, 74-85.

Cohen, Y. 1994. Radicals, Reformers and Reactionaries: The Prisoner's Dilemma and the Collapse of Democracy in Latin America. Chicago: Chicago University Press.

Connor, W. 1994. Ethnonationalism: The Quest for Understanding: Princeton: Princeton University Press.

Dawidow, J. 1980. A Peace in Southern Africa: The Lancaster House Conference on Rhodesia 1979. Boulder: Westview.

Elster, J. 1993a. "Arguing and Bargaining in the Federal Convention and the Assemblé Constituante." In Raino Malnes & Arild Underdal Rationality and Institutions: Essays in Honour of Knut Midgaard, Oslo: Scandinavian University Press.

Elster, J. 1993b. "Constitution-Making in Eastern Europe: Rebuilding the Boat in the Open Sea." Public Administration 71, pp. 169-217.

Frendreis, J. P. 1983. "Explanation of Variation and Detection of Covariation: The Purpose and Logic Comparative Analysis." Comparative Political Studies 16:2.

Ginifer, J. 1995. Managing Arms in Peace Processes: Rhodesia/Zimbabwe. Geneva: UNDIR.

Gottlieb, G. 1993. Nation Against State: A New Approach to Ethnic Conflict and the Decline of Sovereignty. New York: Council of Foreign Relations Press.

Hampson, F. O. 1996. Nurturing Peace: Why Peace Settlements Succeed or Fail. Washington DC, United States Institute of Peace Press.

Hobsbawn, E. J. 1990. Nations and Nationalism since 1780. Cambridge: Cambridge University Press.

Horowitz, D. L. 1985. Ethnic Groups in Conflict. Berkeley: University of California Press.

Horowitz, D. L. 1991. A Democratic South Africa? Constitutional Engineering in a Divided Society. Berkeley: University of California Press.

Hovi, J. & B. E. Rasch. 1993. Strategisk handling. Oslo: Universitetsforlaget.

Hroch, M. 1985. Social Preconditions of National Revival in Europe. Cambridge: Cambridge University Press.

Kaplan, I. 1983. "The Society and Its Environment." In Harold D. Nelson (ed.) Zimbabwe: A Country Study, Washington D. C: The American University.

Klinghoffer, A. J. 1980. The Angolan War: A Study of Soviet Policy in the Third World. Boulder: Westview Press.

Lan, D. 1985. Guns and Rains: Guerrillas & Spirit Mediums in Zimbabwe. Berkeley, University of California Press.

Legum, C. (ed.). 1974-80. Africa Contemporary Record. Annual Survey and Documents. London: Rex Collins.

McGarry, J. & B. O'Leary (eds.). 1993. The Politics of Ethnic Conflict Regulation. London: Routledge.

Marcum, J. A. 1969. The Angolan Revolution Vol. I: Anatomy of an Explosion. Cambridge, Mass: MIT Press.

Marcum. J. A. 1978. The Angolan Revolution Vol. II: Exile Politics and Guerrilla Warfare (1962- 76). Cambridge, Mass: MIT Press.

Marcum, J. A. 1987a. "UNITA: The Politics of Survival." In Helen Kitchen, Angola, Mozambique and the West. Washington: The Center for Strategic and International Studies, pp. 3-16.

Marcum, J. A. 1987b. "A Quarter of Century of War."In Helen Kitchen Angola, Mozambique and the West. Washington: The Center for Strategic and International Studies, pp. 17-35.

Meckstroth, T. 1975. "Most Different Systems and Most Similar Systems: A Study of in the Logic of Comparative Inquiry." Comparative Political Studies 8, pp. 132-157.

Morrow, J. D. Game Theory for Political Scientists. Princeton: Princeton University Press.

Ngcongo, I.D. 1989. "The Mfecane and the Rise of New African States." In J.F. Ade Ajaydi (ed.) General History of Africa VI: Africa in the Nineteenth Century until the 1880s. Oxford: Heinemann International, pp. 190-93.

Nodia, G. 1992. "Nationalism and Democracy." Journal of Democracy, 3.

Ohlson, T. & S. J. Stedman. 1994. The New Is Not Yet Born: Conflict Resolution in Southern Africa. Washington DC: The Brookings Institution.

Przeworski, A. 1986. "Some Problems in the Study of the Transition to Democracy." In G. O'Donnell, P. C. Schmitter & L. Whitehead, Transitions from Authoritarian Rule: Prospects for Democracy, Baltimore: John Hopkins University Press.

Przeworski, A. 1988. "Democracy as a Contingent Outcome of Conflicts." In J. Elster & Rune Slagstad, Constitutionalism and Democracy, Cambridge: Cambridge University Press.

Przeworski, A. 1991. Democracy and the Market: Political and Economic Reforms in Eastern Europe and Latin America. Cambridge: Cambridge University Press.

Rinehart, R. 1983. "The Origins of Zimbabwe." In H. D. Nelson (ed.) Zimbabwe: A Country Study. Washington D. C: The American University.

Smith, A. D. 1986. The Ethnic Origins of Nations: Oxford: Blackwell.

Stedman, S. J. 1991. Peacemaking in Civil War: International Mediation in Zimbabwe, 1974-80: Boulder: Lynne Rienner Publishers.

Stedman, S. J. 1993. "The End of the Zimbabwean Civil War." In R. Licklieder, (ed.) Stopping the Killing: How Civil Wars End. New York, New York University Press, pp. 125-163.

Stockwell, J. 1978. In Search of Enemies. London: Andre Deutsch.

Stoneman, C. & L. Cliffe. 1989. Zimbabwe: Politics, Economics and Society. London: Pinter Publishers.

Tamarkin, M. 1990. The Making of Zimbabwe: Decolonization in Regional and International Politics. London: Frank Cass.

Tsebelis, G. 1990. Nested Games: Rational Choice in Comparative Politics. Berkeley: University of California Press.

Zartman, I. W. 1989. Ripe for Resolution: Conflict and Intervention in Africa. 2. rev.ed. New York. Oxford University Press.