When do People Protest? – Using a Game Theoretic Framework to Shed Light on the Relationship Between Repression and Protest in Hybrid and Autocratic Regimes

What is the relationship between different levels of coercion and popular protest activity in hybrid and autocratic regimes?1 This is the question I want to address in this paper. An abundance of theoretical and empirical studies exists on whether state repression increases or decreases the incidence of domestic protest. However, findings have been mixed providing support for almost every possible relationship between protest and repression (e.g. Rasler 1996, Moore 1998, Carey 2006). The three dominant approaches are (1) the inverted Uhypothesis, (2) the backlash hypothesis, and (3) the non-linear hypothesis. The first perspective, the inverted U curve, argues that a shift toward lower levels of repression opens new opportunities for challengers to act collectively to demand their rights and to make claims against the state (Tarrow 1994). The second view, the so-called backlash hypothesis, is based on the opposite assumption contending that repression facilitates protest by nourishing a collective sense of defiance and intensifying organizational solidarity among diverse loosely connected movements (Meyer and Staggenborg 1996). A third array of studies claims that the relationship between coercion and protest might be more complex and more multi-dynamic than both the ‘inverted-U’ hypothesis and backlash assumption presume and advocate some sort of non-linear relationship (Kowalski and Hover 1992).


Introduction
What is the relationship between different levels of coercion and popular protest activity in hybrid and autocratic regimes? 1 This is the question I want to address in this paper. An abundance of theoretical and empirical studies exists on whether state repression increases or decreases the incidence of domestic protest. However, findings have been mixed providing support for almost every possible relationship between protest and repression (e.g. Rasler 1996, Moore 1998, Carey 2006. The three dominant approaches are (1) the inverted Uhypothesis, (2) the backlash hypothesis, and (3) the non-linear hypothesis. The first perspective, the inverted U curve, argues that a shift toward lower levels of repression opens new opportunities for challengers to act collectively to demand their rights and to make claims against the state (Tarrow 1994). The second view, the so-called backlash hypothesis, is based on the opposite assumption contending that repression facilitates protest by nourishing a collective sense of defiance and intensifying organizational solidarity among diverse loosely connected movements (Meyer and Staggenborg 1996). A third array of studies claims that the relationship between coercion and protest might be more complex and more multi-dynamic than both the 'inverted-U' hypothesis and backlash assumption presume and advocate some sort of non-linear relationship (Kowalski and Hover 1992).
The three hypotheses have been tested in multiple settings and there is an abundance of empirical studies (e.g. Mason and Krane 1989, Opp and Ruehl 1990, Choi 1999) that support any of the three hypotheses. In this paper, I will show through a game-theoretic framework that these contradictory findings can and should be harmonized. Through a game of complete information I will reveal that both the hypothesis of an inverted U-shape and the backlash hypothesis result in equilibria with empirical referents.

206
The chapter will adopt the following structure. Section two will review the literature on protest and coercion by presenting the competing hypothesis; (1) the inverted Urelationship, (2) the backlash relationship, and (3) the non-linear hypotheses. Section three will present and explain the game of complete information. This part of the paper will also present and discuss the different equilibria and assign them to either the backlash theory or the hypothesis of an inverted U-curve. Finally section four will summarize the main findings of this paper and will provide some avenues for future research.

Review of the literature
The hypothesis of an inverted-U relationship implies that any state moving from middle or especially from high coercion to midrange to low coercion would confront a substantial rise in protest. For example, Kitschelt (1985: 300) argues that very open systems convert movements into pressure groups and very closed systems crush them. In his view, moderately intransigent systems are most conducive to high mobilization because they neither include the protesters nor destroy their movement. According to this perspective, a shift toward lower levels of repression opens new opportunities for challengers to act collectively to demand their rights and make claims against the state. Tarrow (1994) posits that especially when the state becomes unable to repress and thus becomes politically vulnerable, movements' organizations' opportunities to achieve goals increase. Conversely intense repression not only dissuades potential adherents from risking collective action but also deprives organizations of resources (Tilly 1978). A number of empirical studies including Olzak, Baesley and Olivier's (2002) analysis of collective action in South Africa indeed finds that severe levels of repression decreases collective action while medium levels escalate it. Their study suggests that on the one hand, antiapartheid protests in South Africa between 1970 and 1985 intensified when the government showed signs of weakness and passed reform bills. Reversely, Olzak, Baesley and Olivier (2002) claim that periods of state repression dampened protest. Muller's and Weede's (1990) analysis also triggers support for this inverted U-relationship in a largesample, cross-national test with political deaths as the indicator for political violence. Finally, Opp (1994) used data from interviews to test and confirm the inverted-U hypothesis for protests that occurred prior to the fall of the Berlin Wall in the former German Democratic Republic.
However, in other studies, the "inverted U" hypothesis receives less support than its "backlash" alternative, that is, that dissidents react strongly to extremely harsh coercion. In this view, repression leads to more protest by nourishing a feeling of collective solidarity among protesters and prospective protesters (Marks 1989 andBrockett 1995). According to Meyer and Staggenburg (1996) movement militancy intensifies when members encounter opposition. Mason and Krane (1989) advance a similar argument, contending that extremely harsh coercion might reduce protest temporarily, but would likely increase dissidence in the long run, in particular, when it is applied indiscriminately. In line with Mason and Krane, Opp and Ruehl (1990: 521-27) claim that the direct negative effect of repression on protest can be reversed if repression leads to micro-mobilization processes, which raise incentives for protest. These processes are facilitated if people perceive the repression to be illegitimate and if there is an organizational structure for protest activity in place (White 1989). Empirical examples of the backlash hypothesis are South Korea, Argentinia and the West Bank. In South Korea, during the famous Kwangju People's Uprising in 1980, paratroopers employed atrocious violence to break up the student demonstrations. This violence led to solidarity of fellow community members. In the end the entire citizenry united to resist the onslaught of the paratroopers against demonstrators and was able to rise up and fight off an elite force of three SWC brigades, which amounted to three-thousand Republic of Korea paratroopers (Choi, 1999). In Argentina tens of thousands of people were tortured, abducted, and murdered by their own government in the 1970s and 1980s. This cruelty spurred major protests; first by those that were directly affected by the junta's cruelty and human rights violations. Second these protests diffused in society, and let to a dramatic process of transformation from a "culture of fear" to a "culture of solidarity" which promoted the democratization of society (Brysk 1994). Finally, Khawaja (1993) andFrancesco (1995) find evidence that protest spiked under extremely harsh repression during the 1980s and the 1990s in the West Bank. In particular, Francesco`s study of the first Intifada highlights that Palestinians continued to protest under extremely harsh conditions (this coercion consisted of killing protesters).
Attempting to shed some lights on these opposite findings, scholars have tried to specify some circumstances under which either of the two hypotheses applies. Gurr and Lichbach (1986) purport that only consistent policies of state repression might decrease protest activity. In contrast, a mixture of increased repression and renewed concessions might have the opposite effect. Other scholars (e.g. Marwell andOlivier 1993 andOberschall 1994) adhere to so called bandwagon models, critical threshold models, or models of critical mass to explain the interaction between dissidents' responses and government repression. These models describe a chain reaction in which small numbers of people trigger the participation of larger numbers of people over time, suggesting that once a certain, usually unspecified, threshold of number of participants is crossed, the costs of mobilizing a larger crowd decline. In such a scenario, more people are likely to join because they feel encouraged by the protesters' commitment and willingness to dissent. Because this mobilization puts enormous pressures on leaders to reduce dissent quickly, they are likely to adopt conciliatory policies. The result is more dissent because successful collective action sustains the involvement of old participants while convincing sideliners of the usefulness of protest (Chong 1991: 116-125).
Other scholars (e.g. Tarrow 1989aTarrow , 1989b attribute special importance toward the timing of repression within a protest cycle. 2 The essential argument is this: when the concept of the protest cycle is wedded to state violence, harsh repression will only provoke further popular mobilization during the ascendant phase. In contrast, indiscriminate state oppression will deter popular collective action under normal conditions that is prior to the onset of a protest cycle. The same logic applies for the descendant phase of a protest cycle; harsh state violence 208 can bring protest to an abrupt end. Nonetheless, Brockett (1993: 475) contends that when elites are willing and capable to institute widespread killings on a sustained basis, they are often successful in ending the protest cycle (even in its ascending phase) and terrorizing the population back into political passivity.
In another study, Mason and Kray (1989, 177) give explicit attention to the ways in which differing mixes of benefits and sanctions (from both the government and the opposition) affect the political preferences and behavioral choices of the masses. They argue that the distribution of mass support, opposition and apathy toward the regime will vary depending on the targeting strategy employed by a regime's repressive activities. The regime violence might be directed (1) just against leaders of opposition organizations, (2) against the rank and file membership of opposing organizations, or indiscriminately at the public regardless of its involvement with the opposition. Violence targeted against just the opposition is postulated to reduce the willingness of non-elites to actively support the opposition. If opposition leaders are imprisoned or murdered non-elites begin to doubt whether opposition organizations can deliver collective goods and rather abstain from joining the opposition struggle against the regime. According to Mason and Krane (1989: 180), similar outcomes are to be expected if the regime targets not only leaders, but also rank and file supporters of the opposition; now the inactive masses are all the more likely to remain uninvolved because fear has been added to futility. However, Mason (1989) argues that this targeting strategy will have the opposite effect on non-elites who already support the opposition. Being inactive would probably not remove them from being at risk. As a result, sympathizers of the opposition are likely to shift to more protest and possibly more violent forms of collective action. Finally, Nam's (2006) analysis of pro-democratic protests in South Korea in the 1980s and 1990s uncovers variation in the dynamics of state coercion and protest according to types of dissident groups. Of the groups assessed, workers were particularly active in protest. Farmers were the least active, and the Korean regime responded with the least repressive approach toward them.
The above-mentioned studies hint that the relationship between coercion and protest might be more complex and more multi-dynamic than both the 'inverted-U' hypothesis and backlash assumption presume and advocate some sort of non-linear relationship. As Hoover and Kowaleski (1992) argue it seems plausible that different situations produce different relationships between protest and coercion. In a case study on Iran, Rasler (1996: 149) comes to an analogous conclusion claiming that there are important reciprocal relationships among concessions, strikes, and spatial diffusion that suggest an interactive link between government and non-government actors (149). Trying to model this complexity Tsebelis and Sprague (1989) present the dynamics of a nonlinear (predator-prey) model under the hypothesis that the relationship between protest and coercion oscillates (pp. 551-52). Their model is motivated by an interest in understanding the sequential response of states to dissident activity. That is the model poses the two questions: if dissidents protest what will the state do next? Vice versa what will the state do if dissidents are cooperative? Under this predator prey model the re-action of both state and dissenters depends on the most recent behavior of the state (accommodation or repression) and the dissidents' responses. The model produces the hypothesis that states substitute repression for accommodation and vice versa, in response to dissident protest (Francesco 1996, Moore 2000. The above review of the literature shed some light on the complex interaction between protest and repression. Depending on, for instance, the type of repression and its consistency, on whether the aims of the movement are revolutionary or reformatory, instrumental or identity oriented, repression may sometimes succeed in intimidating protesters, whereas at other times it may activate them. In order to answer the question why regime violence smothers popular mobilization under some circumstances, whereas at other times, similar or even greater levels of violence by the state will provoke mass collective action rather than pacify the target population, it is necessary to clearly lay down the conditions under which each of the two hypotheses applies (Carey 2006). Through a game theoretic model, I will show (temporary) equilibria, which will result in protest. These empirically relevant equilibria underline the conditions under which each of the two theories is valid.
3. The game theoretic model 3 I will start the discussion of the game theoretic model by laying out the assumptions and the building blocks of the game. There are two unitary actors, the state and the protesters. Nature decides whether the government is strong 4 or weak 5 . No matter whether it is strong or weak, the government makes the first move in the model by determining the degree of repression. 6 It can either decide to repress strongly 7 or it can opt for moderate degrees of repression 8 . Contingent on whether the government is strong or not and on whether or not the government engages in high or moderate degrees of repression the protestors can decide to protest or not to protest. 9 3 I assume that protesters know whether they are confronted with a strong or a weak government, which means that they recognize on which side of the tree they are. In reality, it is plausible that in many situations civil society knows through the media, leaks in the government, foreign sources or good networking how strong and determined the government is to quell possible protest. In such a scenario it is sufficient to evaluate one side of the tree, because citizens act under complete information. They calculate the costs and benefits of protesting, compare them to the status quo and if their expected utility of protesting minus the expected costs of protesting is greater than the status quo then they protest. 4 A strong government is characterized by government coherence and considerable governmental resources to quell protest and punish and oppress the protesters 5 A weak state or weak government is characterized by splits within the government and/or between the government and the military complex. A government is also characterized as weak if it has few resources to effectively fence off protest 6 I have the government move first because both inverted U-relationship and the backlash relationship explicitly see protest as a reaction to state repression 7 High degrees of repression are characterized by extensive political imprisonment of a considerable number of the population. Murders disappearances and torture are a common part of life. Generally, the leaders place no limits on the means of thoroughness with which they pursue personal or ideological goals. 8 Moderate degrees of repression are characterized by restrictions of personal and political freedoms. Regime brutality and arbitrary imprisonment occur, but not on a large scale basis. The human rights violations are also not as flagrant as in a regime that engages in harsh repression. 9 I refer to protesters quite loosely. In some cases, they include the entire citizenry, while in other cases they include one subset of the population of a country. The group against which repression is directed can also be one region, or religious group. The government must not only calculate the benefits of repression, it must also calculate the likely costs under the possible scenario of protest. These costs, which are labeled K(S) if the government is strong and K(W) if the government is weak, involve the costs of repressing the protestors in case of collective action. Due to its increased military capabilities, government coherence and monetary resources, we can assume that the costs of repressing are lower for a strong government than they are for a weak government (K(W)>K(S)). The government is also faced with a final set of costs, the opportunity costs for high repression. Continuous levels of high repression involve a high degree of monetary and personal resources, which the government has to spend. Opportunity costs can also be reputational and economic, as other states might isolate the regime or stop trading with the country as a result of harsh repression. Thus, at least in the medium or long term, opportunity costs may harm the regime rendering it hard for those in power to sustain these costs over a long period of time. I label the opportunity costs OC. Similar to the protracted costs of putting down a rebellion; the opportunity costs of harsh repression are higher when the state is weak than when the state is strong.

OC(W)>OC(S). 10
The protesters, as the second actor, face costs and benefits from protesting, as well. I denote the expected utility of protesting by E and the costs by C. The expected utility can be derived from the expected individual and group benefits. Individual rewards can be tied to some political or social benefit (more economic or political rights, higher wages or more social benefits) that the individual hopes to reap through protesting. Group benefits are linked to some benefit a deprived group might gain (e.g. more political representation or an end of discrimination against a certain ethnic group or region) (Lichbach 1994). Protest activity can also be pushed by peoples' conviction that group action can be successful and from beliefs that their own participation may trigger subsequent participation from more people in collective protest activities (Gamson 1992).
With regard to the protesters' calculation of benefits we can reasonably assume that they can derive higher gains from protesting under more repression. In a situation, where the government indiscriminately represses, citizens are often tortured and discriminated against; a scenario which renders it very appealing for people to try to get rid of the regime. In contrast, under medium or lower degrees of repression, citizens have some rights and normally do not have to fear for their lives. This not only signifies that they can gain less through protest; it also means that they can lose the few rights they possess if the protests fail. Based on this rational, I assume that the expected utility from (successful) protests E is higher under harsh repression E

HR) than under lower degrees of repression E(MR). (E(HR)>H(MR)
However, people also face serious costs (C) from protesting. These costs include personal expenses such as the time and money individuals have to devote to protesting and more importantly possible costs that the regime might inflict on the protesters (e.g. death, imprisonment, torture and losing one's job). Organizations and associations also play a role in determining the costs and benefits of protesting. Well developed organizations and networks might decrease the costs of participation for individuals by providing protection against arbitrary or cruel state action. Networks can also supply protesters with resources or they can simply facilitate collective action (e.g. McAdam and Paulson 1993, Opp and Gern 1993, Oegema and Klandermans 1994. The costs of protesting are higher when the government is strong. Protesters know that a strong government has the monetary and logistic resources to quell any revolt or demonstration. They also know that the government might inflict harsh punishment upon them, if they dared to rise. In contrast, if civil society actors know that the government is weak, they can hope that those in power are somewhat restrained in their ability to push down the protesters. Those that weigh the pros and cons of protesting also know that a weak government might not be certain of the loyalty of the police and armed forces. In such a situation, civil society actors might also be aware that a forceful reaction might overstretch the resources and/or create (further) rifts within the regime. Finally, protesters may recognize that a weak government is vulnerable to outside intervention. All these factors make protest action less scary when the government is weak. This gives us the following calculation C(S)>C(W).
For the protesters there is a last component, which they have to take into their consideration before deciding on action. They have to compare the cost benefit ratio of protesting to their status quo. They will protest if overall they think, that despite the high costs of protesting they would gain more through collective action than they have at the status quo. The status quo is denoted by S. I assume that under harsh conditions the status quo is worse and involves more sacrifices and suffering than under medium degrees of repression. Thus S(HR)>S(MR).

Possible equilibria
So far, I have created the game tree and established the payoff structure. No matter on which of the two sides of the tree the regime is, it has two choices; it can either harshly or moderately repress. The protesters response is conditional on the governments' initial choice. Confronted with harsh/ moderate repression civil society actors can either protest or not protest. In order to proceed from this point on and in order to calculate equilibria, I have to specify some further conditions. In this section, I will calculate all equilibria which result in protest and assign them to one of the theories. 11 I will proceed in the calculation of the equilibria from the left side of the tree to the ride side of the tree. 12

B(HR)-K(S)-OC(S)) +(B(HR)-OC(S)) > (B(MR)-K(S))+B(MR).
For the second player, the protesters, the expected utility of protesting under harsh conditions minus the expected costs of collective action must be higher than the status quo (E(HR)-C(S)) > S(HR). On the regime's side, this scenario often applies when the opportunity costs are low and when the government has a strong military and security apparatus. On the side of the protesters, it is safe to assume that they must be very desperate and infuriated; otherwise they would not dare to rise against a strong state.
Some recent empirical examples that confirm this equilibrium can be found in Eastern Asia. For example, the recent protests of the Tibetan monks fit this scenario. China has perpetually opted for hard repression and denies any quest for autonomy to the Tibetan people. By continuously suppressing the population and the monks, in particular, China aims at maintaining its territorial integrity, its ideology and attempts to prevent other regions and/or religions from rising. For China the benefits of oppression are high and the opportunity costs are relatively low. China has the military power to maintain an iron grip on Tibet. Due to its importance as an international trading partner, it also hopes that despite its human rights violations it would not suffer from severe blows in its relations with foreign countries. For the monks on the other side, the potential benefits of protesting could be high. Being deprived of personal and group rights and restricted in practicing their religion, they have a lot to gain from protesting. Even if action right now, might not lead to more autonomy and greater freedoms, collective action can at least increase the opportunity costs of the regime. The monks' protests can attract foreign attention (e.g. from human rights groups), and might increase the costs for the government to maintain law and order and they might pave the way for future protests. Another example, where extremely harsh coercion led to manifestations were the 2007 anti government manifestations in Burma, again led by Buddhist monks. In general, protests events under a harsh repressive state normally do not turn into a civil war, but rather remain quite contained because the government is strong and has the power to repress the protesters efficiently.
The second scenario that can lead to protest is theoretically possible, but empirically rare, if not non-existent. It consists of a situation where the regime is strong but nevertheless only opts for medium degrees of repression. Under this scenario the regime would calculate its costs and benefits as follows: (

B(MR)-K(S)+B(MR)) > (B(HR)-K(S)-OC(S) +B(HR)-OC(S)).
For the government this calculation makes sense under two scenarios: First the regime might opt of medium repression, when the benefits from harsh repression are only marginally higher than the benefits from lower degrees of repression. Second, when the opportunity costs for strong repression are extremely severe, it might be a sound choice for those in power not repress to indiscriminately. Under a situation where the government is strong, but abstains from harsh repression, civil society actors protest if E(MR)-C(S) > S(MR). This situation should be extremely rare because the protestors have comparatively little to gain, as they know that the regime is capable of crushing them. They also know that if their revolt is pushed down the regime might renege on all or some of the few benefits. Thus, if a strong regime decides not repress too severely, civil society actors are frequently better off by not rising. The example of Singapore and other Asian tigers, which are characterized by medium degrees of repression and high stability, underlines this point.
The third equilibrium, which also supports the backlash hypothesis, prevails when the government is weak, but despite its weakness chooses to repress harshly. This situation applies under the following scenario B

(HR)-K(W)-OC(W)+B(HR)-OC(W) > B(MR)-K(W).
Confronted with a weak and repressive government, civil society actors decide to protest if E is (HR)-C(W)> S (HR). For this equilibrium to be valid the government must considerably gain from high repression. First, these gains can be monetary in that the government can extract certain natural resources from a region. Second, these gains can also be ideological in that they allow one religious or ethnic group to dominate the political and social life.
Confronted with a weak government that harshly suppresses, the oppressed people or groups have a comparatively high tendency to resist and protest. They know that the state is weak and are aware that the government might not sustain perpetual challenges.
Empirical examples of this situation, which follow the backlash hypotheses, are multifold. Protests in such a scenario are often violent and turn into civil wars. It is often unclear, who prevails in these violent conflict, the regime or the protesters. A situation were a repressive regime prevailed was the Kurdish uprising against the repressive Iraqi government. The ruthless oppression of the Kurds by Saddam Hussein's regime provided incentives for these tribes to rise and fight for autonomy once they perceived a slight weakness in the regime. In 1991, after the first Gulf War these tribes saw the possibility for a successful upheaval and took up their arms. However, they were defeated by the Iraqi army in a very cruel way.
Sudan is another example, where a weak repressive state has been confronted with protest. However, contrary to the Iraq protests, the oppression of the non-Arab populations in the south by the Arab dominated government of the North, has led to perpetual conflict since 1983. In Sudan, the relatively weak government or government-related forces have not only repressed minorities, but have also engaged in atrocities and human rights violations for economic and ethnic reasons, thus spurring violence. This spiraling of protests and stronger repressions has led to a situation, where the government lost control over some parts of the country and has been unable of keeping up the order. In Sudan, as in other states including Somalia, the interaction between a weak state, high degrees of repression and protests have led to failed states.
Under a third scenario, protests against a weak, but strongly repressive regime can lead to a successful revolution. In such a situation the protesters prevail against the weak repressive government. An empirical example of this case would be the February Revolution in Russia in 1917. Weakened by World War I and confronted with a ruthless economic crisis (e.g. a severe famine), Tsar Nicolas II continued his autocratic and repressive rule. This infuriated large segments of the population and broke the loyalty of many civil servants including police and military officers. The Tsar was finally overthrown by massive protests, which he was unable to quell.
The last equilibrium that leads to protest happens when the government is weak and moderately represses and the civil society actors protest despite merely moderate degrees of repression. This scenario, which confirms the hypothesis of the inverted U-curve, occurs when the regime is better off by weak repression than by strong repression (B(MR)-K(W)+B(MR)) > (B(HR)-K(W)-OC(W)+B(HR)-OC(W)). The protesters will turn out and engage in collective action if the expected utilities from protesting minus the costs from protesting are higher than the status quo: (E(MR)-C(W)>S(MR). This equilibrium is quite a common scenario because under some circumstances, a weak regime might be forced to lower its levels of repression.
Reasons can be financial problems, impending sanctions or military interventions, a lack of qualified spies and supervisors, economic needs and reputational costs. If any or several of these scenarios apply the regime is often forced to make concessions by decreasing its degree of repression. In this regard, a decrease of oppression is often a means of preventing protest actions from happening (e.g. with more rights the benefits for protesting decline). In some cases this rational might play out and regimes can in fact prevent protests from occurring. However, in other cases, the benefits are too marginal to fundamentally change the cost benefit ratio of the possible protesters. Despite being granted some rights, civil society actors might still opt for protests, because they think that they can gain much more than the few rights that the government granted them.
Thus, this last equilibrium describes a situation under which the inverted U-curve applies.
There are many empirical examples in which protests follow this last equilibrium. Very relevant and important are probably the protests that preceded the collapse of the Soviet Union and the fall of the Berlin Wall. Shattered by an economic crisis and incapable of keeping up the arms race with the United States the leadership in Moscow and the dictators of the satellite states could not uphold high degrees of repression any more. In the 1980s, people in Eastern Germany and elsewhere in the Eastern Bloc could critique the government; they could watch Western television and political prisoners could be bailed out by Western governments, most notably Western Germany. However, these concessions were not enough to appease the people. Confronted with a weakened state, civil society actors saw the change of gaining complete political and economic freedoms. This appeal minus the anticipated costs for protesting outweighed the status quo benefits and people turned out in larger and larger numbers. For those in power the costs of pushing down the protests became too costly and they finally had to step down, first in Hungary and Germany and then in all the other former Eastern countries.

Conclusions
Through a game theoretic framework I have shown the conditions under which both the inverted U relationship and the backlash hypothesis frequently occur. The inverted Urelationship applies when a weak regime engages in moderate degrees of repression and the civil society actors decide to protest. In many cases, these protests further weaken the government and trigger an overthrow of the regime. The backlash hypothesis prevails under two scenarios. Under the first scenario, the government is strong and represses harshly and civil society actors protest despite the strength of the government. As the examples of China and Burma reveal, these protests are often violently quelled and trigger subsequent periods of relative calm. Under the second scenario, the government represses harshly despite being weak. This scenario can lead to very volatile situations because the protesters have a lot to gain from protesting and the government has a lot to lose. Such a situation might spur protests and violent reactions, which can easily spiral into a civil war. The empirical examples of Sudan, Somalia, or the Democratic Republic of Congo fit this scheme.
Thus, protest can occur under three scenarios. None of these equilibria explains the repression protest nexus in all or the majority of the cases. Yet, each equilibrium has empirical referents and explains some events of protest action. Which of the two theories actually holds depends on the dynamic cost-benefit calculation of both actors -the state and civil society actors. As I have shown in this paper, this calculation depends on a plethora of factors. Based on the benefits of repression, the possible costs of pushing down protests and the opportunity costs of high repression, the regime determines its degree of repression. Depending on their expected utility from protesting, the expected costs of protesting and status quo, civil society actors then decide whether or not to turn out. To determine which of the two hypotheses apply, it is necessary for researchers to establish the conditions on the ground. Only with on the ground knowledge can scholars predict which of the two approaches applies but as the situation on the ground changes, so might the behavioral strategy of the actors.