Formal Models of Authoritarian Regimes: A Critique

The very idea that authoritarian regimes (“autocracies”) may enjoy popular support is hard to fathom for democrats. Models of authoritarian regimes often entail tacit ideological assumptions, and many are driven by methodological fashions. They ignore the efforts of rulers to provide what people value. The psychology they assume is inadequate to predict actions. They are often too abstract to generate testable predictions. “Support” for any regime is difficult to assess.

W e now have a canonical view of authoritarian regimes. A dictator (autocrat), individual or collective, maximizes the probability of survival in office or the expected rents, using instruments that include co-optation, repression, censorship, propaganda, and the like. These instruments differ in their effectiveness and costs. In the (Stackelberg) equilibrium, the dictator chooses their optimal mix subject to some constraints. Comparative statics depend on the exogenously given cost and effectiveness parameters (for the best summary of this view, see Guriev and Treisman 2015).
This view is based on assumptions that are laden with ideological biases and are often driven by methodological fashions. The ideology is flagrant even at a purely semantic level. Game-theoretic articles on "persuasion" (Galperti 2018;Kamenica and Gentzkow 2011) use this concept in purely neutral terms. Yet, when such models are applied to authoritarian regimes, "persuasion" becomes "manipulation" (Edmond 2013;Gelbach and Simpser 2015;Rozenas 2016) in the same way as "information" becomes "propaganda" (Little 2017), without any difference in the structure of the models. But the ideological roots are much deeper. The very idea that autocracies may enjoy popular support is hard to fathom for democrats. Unless they are "brainwashed" or "indoctrinated," how can people conceivably support an autocrat?
Autocracies are assumed to be inherently brittle, surviving only because people are misled or repressed. This focus on regime survival diverts our attention away from their routine functioning. Certainly, all actions by government officials have some effects of regime stability. But this does not mean that all their actions are motivated by the drive to survive in power. Autocracies do collect garbage, regulate traffic, issue dog licenses, and fill street holes: they govern. In turn, the individuals who populate these models are assumed to have postures toward the regimes at every moment of their lives. But ordinary people are not politically hyperactive in any regime. People in autocracies do not incessantly live under the shadow of dramatic historical events; they lead everyday routine lives. As Pepinsky (2017) incisively observed, "Life in authoritarian states is mostly boring and tolerable." The target of my critique is neither formal modeling nor the "rational choice" approach in general but only some of their typical assumptions and the resulting conclusions. I realize that not all formal models suffer from deficiencies I identity and, conversely, that many studies not using this approach share these deficiencies. Ideological biases pervade our everyday language: we refer to very rich people in the United States as "billionaires" but to their autocratic equivalents as "oligarchs"; we speak of "police brutality" in democracies but of "police repression" in autocracies. Most of my comments are directed at models of incomplete information, whether about the state of the world, the competence of the rulers, or the preferences of others. My central complaints are that these models ignore the efforts of rulers to provide the material or symbolic goods that people value (in technical language, they model only the subgame in which the state of the world is given exogenously); that the psychology they assume ("beliefs," their relation to actions, the set of possible postures toward the regime) is inadequate to predict actions; and that they are often too abstract to generate testable predictions. In turn, I find models that predict regime change directly from the preferences of some actor, such as "the representative agent" or "the elite" (Acemoglu and Robinson 2006), to be too limited in their micro-foundations to elucidate the dynamics of these rare events. I do believe that these faults are redeemable, even if I do not know how to remedy them, so that my critique is not aimed at the formal modeling approach as such.
My dissatisfaction with these models has personal roots: they just do not "feel right" for someone who actually lived under such regimes. Moreover, it is fed by ethnographic studies, such as Wedeen's (1999) work on Syria on which I rely extensively, as well as personal accounts and novels, which are often more telling than models. I know that what does not feel right for some may be persuasive for others, but the list of questions that this view leaves open is long and some of the puzzles are deep.
I find particularly doubtful our assessments of political support for authoritarian rulers and regimes. 1 Mussolini claimed in retrospect that "strictly speaking, I was not even a dictator, because my power to command coincided perfectly with the will to obey of the Italian people" (quoted in Cassese 2011). Clearly, one can easily reject his claim as an ex-post excuse for repression. But is it true that the Italian people did not want to follow Mussolini? Is the popularity of Messrs. Putin or Xi due only to their use of nasty, nondemocratic, instruments? (see Frye et al. 2015 on Russia;Chu 2011 on China). Are postures toward autocratic regimes consistent with beliefs? What does it mean to "believe"? Are people duplicitous when they collaborate with authoritarian regimes? Can we tell what is rational to do under different choice sets? And in the end, what does it mean to "support" any regime?
As these questions make clear, some of them cannot be answered by using game-theoretic apparatus, and perhaps some cannot be answered at all. But I hope to persuade others that if we are to understand how authoritarian regimes function, we need to drop both our ideological and methodological blinders. In what follows, I first briefly lay out the ideological assumptions shared by most models of authoritarian regimes and then delve into issues that I find puzzling.

Assumptions
Rule of a Minority? I blithely wrote some years ago that "dictators are dictators because they cannot win elections, because their preferences diverge from those of the majority of the population" (Gandhi and Przeworski 2006, 2). I am now eating these words because there certainly must have been at least some periods in the histories of some autocracies when they enjoyed widespread popular support.
As Schumpeter (2010Schumpeter ( [1942, 221) observed, "Instances abound-perhaps they are the majority of historical cases-of autocracies, both dei gratio and dictatorial, of the various monarchies of non-autocratic type, of aristocratic and pluralistic oligarchies, which normally commanded the unquestioned, often fervent, allegiance of an overwhelming majority of all classes of their people." The authority of the rulers in such regimes did not originate from their capacity to persuade but from their preassigned place. 2 During the twentieth century, such autocracies were replaced by revolutionary regimes proclaiming to be a civilizational alternative to democracy: communism, fascism, "socialism with Chinese characteristics." At least during some period, these regimes probably enjoyed popular support. Yet even such autocracies are almost extinct by now. Modern autocrats claim that they rule not only on behalf but also on bequest of "the people." What is new is that, to validate this claim, modern autocrats hold elections, elections that they win.
We know from empirical research (Cox 2008;Geddes 2009;Przeworski 2015) that political leaders who hold elections of any kind, even those without opposition, have a longer expected tenure in office than those who do not. But reading popular support from results of controlled elections is tricky. 3 When communists produced 99% turnout in single-list elections, they were just showing off their grip over the societies they ruled: such elections informed everyone about their coercive capacity but not about their support. But when elections admit some opposition, how to separate the weights of coercion, manipulation, and authentic support is not obvious (Little 2012;Luo and Rozenas 2018). In the end, just assuming that autocrats cannot enjoy popular support is pure ideology, but interpreting the results of elections is difficult.

Everyone Hates Repression?
Is it possible that masses of people would support violent repression? They did during the Stalinist purges of the 1930s (Goldman 2007), during the Cultural Revolution in China (Walder and Lu 2017), and during the antisemitic campaign in Poland in 1968.
The standard view is that repression is used by autocrats to deter opposition, and probably most of the time it is 980 Perspectives on Politics used for this purpose. In turn, repression used against certain groups, rather than against individuals who are suspected to hold anti-regime postures, may well be supported by other groups. What is repression against some groups may be co-optation for others. 4 Rozenas (2020, 2) observes, "A group that is supportive of the regime becomes an even stronger supporter when another group is repressed, because it infers that the repressed group will be deterred as it does not expect the supportive group to reciprocate in opposing the regime." Yet the use of group-targeted repression may just cater to crass prejudices: antisemitic in Nazi Germany or Communist Poland, anti-Muslim in contemporary Myanmar, anti--Indian in El Salvador. Strictly distributive considerations may also play a role: restricting the access of Jews to some professions was welcomed by non-Jewish majorities in Nazi Germany because it created opportunities for them. The waves of repression in Stalin's Russia and Mao's China were unleashed to protect the leaders from the ire of the masses by scapegoating the "bureaucrats," and it appears that their repression was supported by many workers and peasants. 5

Bases of Support
An opening sentence of a recent paper summarizes in a nutshell, "How do autocrats govern? Roughly speaking, the literature has identified two broad sets of strategies that autocrats use to hang onto power and pursue policy goals: they can manipulate information, through censor ship and propaganda, and they can repress" (Gehlbach et al. 2021). Authoritarian rulers build their support by using repression and manipulating information. Does it make any difference whether they deliver material and symbolic goods that most people value? Is it irrelevant that average Chinese incomes increased sixfold since 1978? That Russians value the restoration of order provided by Putin? 6 That both Chinese and Russians find pride in nationalist adventures? 7 Some distinctions are necessary here. There is a large class of models that focus on structural factors, typically comparing autocracies with democracies. These models highlight the importance of accountability mechanisms, power sharing within the elites, particular institutional arrangements, and the difficulty of autocrats to commit (Acemoglu and Robinson 2006;Benhabib and Przeworski 2010;Besley and Kudamatsu 2007;Che, Chung, and Qiao 2013). In turn, models that focus on information analyze only subgames in which the distribution of the states of the world is taken as given. If the state of the world happens to be good, the autocrat does nothing. If it happens to be bad, the autocrat censors information, explains away the bad news that filter through the censorship by propaganda, and, if both censorship and propaganda fail, activates repression. So, autocrats do nothing to make states of the world more likely to be good: they do not promote growth, pay attention to income inequality, build armies, invest in the performance of sport teams in international competitions-nothing that people may care about. 8 Guriev and Treisman (2019) document the evolution of autocracies from relying on brute force to relying on manipulating information. But I find it striking that they never consider actual performance-only information about performance. Yet performance informs by itself: the poor peasants of Chongqing who now live in nicely kept city apartments with incomes that have doubled every five years do not need to be informed that they are better off. Smart autocrats may use performance to convey information (on Hitler's autobahns see Voigtlaender and Voth 2019), but they do so because performance speaks for itself. Bayesian models of persuasion are intellectually attractive, but it stretches imagination to think that support for the Chinese Communist Party is due to its manipulation of information. 9 The preoccupation with repression, censorship, and propaganda as props that maintain the regimes implicitly assumes that performance-the delivery of whatever people want-is insufficient for an autocrat to survive. Suppose it is not. The empirical question is whether the application of such instruments can be ever sufficient if things are truly bad. I suspect that information manipulation is at most a fig leaf when things go badly and a sweetener when they go well. At least, I know of no evidence that information manipulation has saved autocrats who would have otherwise fallen.

How Autocracies Fall?
Various counts of breakdowns of authoritarian regimes show that only about one-eighth occur as a consequences of mass protests "from below" (Svolik 2012). Yet, most of the literature models mass protests (Edmond 2013;Ginkel and Smith 1999;Lohmann 1994;Shadmehr and Bernhardt 2011), and one can only suspect that it is because we have several types of models that apply to such events.
The dichotomous distinction between "from below" and "from above" (of divisions within the elite resulting in coups) may be too sharp. In many breakdowns of authoritarian regimes, information about divisions within the elite incited popular mobilization, while popular mobilization generated divisions within the elite (Przeworski 1991). When the authoritarian elite is confronted with mass protests, it divides about the strategy of coping with them: "hardliners" want to repress, and "reformers" want to compromise because the support of a moderate opposition would increase their power within the authoritarian ruling bloc. When the opposition learns about the divisions within the elite, it senses that it has allies among the "reformers," so it mobilizes. Static equilibria of such situations have been studied by Casper and Tyson (2014), but to the best of my knowledge, their dynamics have not been analyzed.
Authoritarian regimes do fall, but such events are rare in comparison to instances in which one autocrat replaces another. Sometimes such replacements result from coups within the ruling elite, as when General Viola removed General Videla; sometimes one autocratic regime is replaced by another, as when the mullahs succeeded the shah. Autocrats who maximize their expected tenure in power defend their regime because they cannot survive regime change: to the best of my knowledge, there was never an autocrat who democratized and remained in office after having won a competitive election. But the converse is not true: regimes stay while autocrats go. The question, then, is under what conditions will people living in an autocracy consider that they would be better off under democracy than under an alternative autocrat. Most of the time, democracy is not in the realm of possibilities, so all people can do is to attach their hopes to replacing one autocrat with another. Perhaps it is only when all the autocrats fail that regime change appears on the political horizon.
Puzzles Is Propaganda about Facts or Interpretations? All rulers-those selected in clean elections, those who hold this ceremony without putting their power at stake, and those who do not even bother to hold them-claim reasons that they should be followed, and people are willing to follow them if they believe these are good reasons. The leader (an elected president, el Caudillo, il Duce, ein Fuhrer), the Party, or the State defines who belongs to the collectivity and identifies the interests this collectivity has in common. The Leader offers to the people, in the words of Antonio Salazar, delivered in a speech in 1934, "a well-constituted order, rational as an expression of the nation organized, just in subordinating particular interests to the general, strong because of having as its basis the authority that cannot be rejected and should not be rejected"(emphasis added; quoted in de Oliveira Marques 1998, 432). Coercion is rationalized as "the subordination of particular interests to the general": according to an eminent fascist leader, Alfredo Rocco, fascism was inspired by the idea of "liberty conditioned on the protection of general interests" (liberta condizionata dalla tutela degli interessi generali; quoted in Cassese 2011). This is obviously "propaganda," but it is ideologically sobering to keep in mind that "propaganda simply means the presentation of facts and arguments with a view to influencing people's actions or opinions in a definite direction. In itself, therefore, "propaganda does not carry any derogatory meaning and in particular it does not imply the distortion of facts" (Schumpeter 2010(Schumpeter [1942, 4). 10 After all, the very ideological foundation of representative institutions is the myth that people rule when they are ruled by others (Morgan 1988, 13-14). Propaganda is an instrument of rule in every regime. Its function is not only to prop up a regime against potential threats but also to propagate shared beliefs, mobilize people toward shared goals, and coordinate actions, as in "United We Stand." Is propaganda about facts or interpretations? Is it Bayesian persuasion or model persuasion? It must be the latter: Otherwise, what would be the difference between propaganda and censorship? As Rozenas and Stukal (2019, 982) observe with regard to Putin's Russia, "Bad news is not censored, but it is systematically blamed on external factors." Propaganda must be a comprehensive framework that ties together the past, the present, and the future. It must be incessantly reiterated, providing a causal interpretation of all punctual events, making it costly for individuals to evaluate all its claims (on "exhaustion propaganda," see Horz 2017). Here is something I do recognize as propaganda: All frames promote party ideology to construct a political big picture that links individual frames and bestows an overarching meaning upon them. The country's social and economic development and respective policies in times of transition are thereby embedded in a long-term perspective, encompassing both a historic dimension and a future vision. The reference to the party's historic mission to lead China into a promising socialist future and the emphasis on the CCP's commitment to social and economic development that benefits the people accentuate the elites' efforts to serve the common interest. (Bondes and Heep 2013, 324) Why Does Censorship Work? The autocratic justification for censorship is that it screens harmful or false information. For example, the Portuguese Constitution of 1933 specified (in Article 3), "The censorship will have the aim to impede the perversion of public opinion as a social force and should be exercised to defend it from all the factors that disorient it from truth, justice, morality, good administration and common good." A July 1946 decree of the Polish Communist government instituted censorship "to avoid misleading public opinion by information that does not correspond to reality" (quoted in Tokarz 2012). The welfare effects of censorship might not be always negative: think of newspaper titles that report that someone, one of millions, had negative side effects from vaccination against COVID. Yet, governments that can censor some information can use this power to censor all information that is inconvenient to them, whether it is false or true.
Censorship is an observable signal that the monopolistic communicators fear that people would coordinate on rival beliefs or that rulers are hiding something or seek to mislead the public. The immediate question is why censorship would ever be effective. Snyder and Ballentine 982 Perspectives on Politics (1996,(14)(15), for example, thought that people in communist regimes discounted government messages because they were censored. Yet censorship is widely used, and it seems to work often (Enikopolov and Petrova 2015).
Models of censorship assume that nature generates a stream of states of the world that may be favorable or not to the regime, and the autocrat then adopts a strategy of communicating these states to the public. One reason why censorship works may be that, even if people have discounted censored messages, they do not know what has been censored. Shadmehr and Bernhardt (2015) argue that when citizens do not receive information about some state of the world, they do not know whether there was nothing to report or that information bad for the regime was censored. 11 Censorship works because people do not think about information they do not see. 12 In the extreme, there is one kind of information that is never disclosed, because it is most dangerous to the regime: information about divisions within the elite. Authoritarianism is a system in which there is only one authority that can communicate in public. Yet even if preferences have been manipulated by propaganda and censorship, the claim that people would have held different preferences had they been exposed to a plurality of reasons is counterfactual, and counterfactuals are unverifiable by construction. Hence, although the presence of propaganda and censorship can make us skeptical about the "authenticity" of people's beliefs, it does not offer a test of whether people's preferences induced by such beliefs are caused by these mechanisms. 13

Public and Private Beliefs
Here is a Polish joke about a meeting of a local cell of the Communist Party: Comrade Secretary delivers a speech on "The Dangers of American Imperialism." Then all the comrades in the room express their opinions. Comrade Kowalski did not believe that American imperialism is dangerous, but he had the opinion that it is. What was his "belief"? He did not update, despite having received a public signal from the Comrade Secretary, and instead stuck with his prior. But he signaled publicly that he is willing to act as if he did update.
Comrade Secretary was not trying to change beliefs about American imperialism. He was just demonstrating to all the comrades in the room that their beliefs do not matter. His prudential advice was, "Do not act on your beliefs. Believe whatever you want, but this is how you should act." Game theorists would reply that there is nothing mysterious in people acting in ways not consistent with their beliefs if they fear sanctions. Yet does it mean that I still believe, or does the impossibility of acting on my true beliefs affects them, and I come to believe what I am told to believe? Or do I believe both at the same time? Preventive repression-repression intended to prevent anti-regime individuals from acting or even from communicating their sentiments in anonymous interactions-is insidious. Fear becomes a habit, and when we act out of habit, we do not rationalize every act. Some actions are not to be undertaken, some thoughts are not to be communicated: this becomes instinctive. Compliance may entail fear, but this fear is so deeply internalized that it does not appear as a motor of actions and not even as a restriction of freedom.
The question is, Which beliefs do people reveal when they answer survey questions saying that they support the Chinese Communist Party or President Putin? This is a question to which neither external observers nor the rulers have an answer. As Sen observed already in 1973 (258; emphasis added), "People may be induced by social codes of behaviour to act as if they have different preferences from what they really have." Wedeen (1999, 82; see also Wintrobe 1998) observes, "Requiring citizens to act 'as if' leaves the regime in the predicament of having to evaluate popular sentiment through the prism of enforced public dissimulation." The regime does not know how contingent is the manifest support that it receives.
The question can be posed in a different language. When political science was dominated by social psychology, we spoke in the language of "attitudes," defined as "behavioral predispositions." What was the attitude of Comrade Kowalski? As Kuran (1995) observed, "preference falsification" implies that actions cannot be predicted from professed beliefs.

Cognitive Status of "Beliefs"
What does it mean to "believe"? Many years ago, a friend and I were being hosted by the dean of the Economics Department of the University of Havana. My friend asked the dean what she teaches, to which her answer was "The Capital." "And do you believe that what Marx says in The Capital is true?" my friend continued. The dean's shock was betrayed by an uncomprehending blank gaze. Shrugging her shoulders, she replied, "It is in the program, so I teach it." It is expected of me to do it, so I do it: no "beliefs" are entailed. Is it true that actions must be supported by "beliefs"?
Did the Cuban dean act of fear or only conformity? In Wedeen's (1999, 81) view, Although the power involved in compelling such performances may seem to be of a lesser order than the power to alter beliefs, people are nevertheless required to enact a prescribed, politically congenial self for public presentation. Symbolic practices of power interfere with people's political "subjectivities," with their sense of themselves as political persons. Moreover, people can learn to be comfortable with their involvement, to identify with it as though it were something natural and inevitable and, ultimately, so they may-with no external urging-come to treat any non-involvement as an abnormality, as arrogance, as an attack on themselves, as a form of dropping out of society.
In the cafeteria of the Central Committee of the Polish Communist Party, over lunch, everyone used the ritualized official language: "the working class," "class conflict," "Comrade Secretary" (Szlachcic 1990). In their world there were no "workers" but only "the working class." If asked, they could not even define what "the working class" is, but they actually did see the world in such terms. Were they pretending, conforming, or what?

Complicity or Duplicity?
We all know this story: "Driven by hunger, a fox tried to reach some grapes hanging high on the vine but was unable to, although he leapt with all his strength. As he went away, the fox remarked, 'Oh, you aren't even ripe yet! I don't need any sour grapes.'" How do I justify my complicity to myself? 14 My choices are to act against the regime, undermining its stability; to do nothing; or to actively collaborate, contributing to its stability. The decision to support must be psychologically tormenting for someone who does not like the regime. Perhaps this is why Wedeen thinks that supporters must hide their motives from themselves: why they are complicit but not duplicitous. Wedeen (1999, 77) writes, "Complicity is doubly degrading when people deny to themselves that they are behaving in complicit ways. Ideology is insidious because it allows people to hide their reasons for obedience from themselves." Active support for the regime is an act not of deliberate deception but of some kind of rationalization of their motives. This may well be true, but the psychological mechanism remains opaque. "Rationalization" is a mechanism that generates logical consistency between beliefs and actions but causality may still run from actions to beliefs (Satz and Ferejohn 1994). 15 The empirical status of Freudian psychodynamics is doubtful these days, but some defense mechanismsdenial, displacement, repression, sublimation, or what not -are required if I am to successfully hide my true motives from myself.
Active collaboration is often rationalized by the belief that working from within to improve the regime is the only effective course of action, even when it involves heinous acts. 16 When all resistance is fu tile, one can at least try to steer the regime toward one's values rather than remaining passive. Moreover, the material rewards for collaboration are often high. As a French TV series, A French Village, admirably depicts, both collaboration and resistance can be driven not only by ideological commitments and opportunistic motivations but also by purely emotional, sometimes even sexual, impulses. In situations of conflict, what are reasons for some are rationalizations for others, and they are hard to distinguish. Both the motives and the postures of people who populate these regimes are complex: there are some active opponents, most people do not think every day about politics, some collaborate actively for opportunistic reasons, and some act for authentically ideological motives. Distinguishing these postures matters because they imply different propensities to respond to signals and ultimately different "attitudes": predispositions for different types of actions. 17 Can the psychology of postures toward regimes be captured by expected utilities? The expected utility is the same when the reward from the successful outcome is very high (someone truly hates the regime and believes revolution would bring paradise) while the probability of success is miniscule and when the reward is low while the probability is high. Yet, if you agree with Sen (1988) that people value having a choice, these utilities cannot be the same. In Sen's classic example, I may starve because I am poor or because I decided to fast: the outcome is the same, but the utility of this outcome is different. In his words (1988,292; emphasis in the original), "Doing x and choosing to do x are, in general, not equivalent." But even without relying on Sen, consider an ardent regime opponent who realizes that no one can do anything to bring the regime down. Living in a status quo that one finds odious, and seeing no possibility that it could be changed, one feels impotent, despondent, and perhaps angry. 18 It is just psychologically implausible that someone who feels intensely but cannot do anything would value this state of the world equally as someone who cares less and sees a good chance that it would happen.

Are Some Beliefs Invalid as a Criterion of Representation?
The fundamental assumption of liberal conceptions of representation is that the preferences to be represented are exogenous with regard to the relation of representation. The "will" of the people is a primitive, given before anyone does anything. Pitkin's (1967, 140) claim that "the represented must be somehow logically prior" leads her to reject the possibility that someone can be represented if the wills are affected in the process of representing. Her red line is "manipulation": representation "is perfectly compatible with leadership and with action to meet new or emergency situations. It is incompatible, on the other hand, with manipulation or coercion of the public" (233).
Pitkin's assumption is patently unrealistic. Representation is never a unidirectional relation in which people 984 Perspectives on Politics want something and the representatives do or do not implement it. It is a dynamic relation, a tatonnement, by which individuals adjust their preferences based on beliefs induced by the representatives, and the representatives make decisions that anticipate the reactions of the represented to their messages. As Schumpeter (2010Schumpeter ( [1942, 215) put it, "The will of the people is the product and not the motive power of the political process." Moreover, all regimes use coercion to prevent actions harmful to the collectivity. Even Bentham (1988Bentham ( [1776) argued that the legislators should be promoting "social actions" at the cost of "self-regarding" ones. But this is nothing other than using coercion in defense of common interests, so it cannot serve to disqualify a relation as representative.
Endogeneity of preferences, even if shaped by manipulation, is not sufficient to disqualify a relation as representative. But this need not imply that all endogenous preferences are equally valid as the criterion of representation. Here is Bay (1958, 322): "If support for the regime is manufactured by way of a monopoly control over the media of mass communication, supplemented by severe coercion against oppositional elements …, one must conclude that no amount of public support to the regime can prove that the people's genuine interests are not being exploited in the interest of the ruling few." Yet note that although Pitkin makes a normative claim, Bay makes an epistemic one: claims that a regime enjoys support are unverifiable when the streets are quiet. When people cannot object, the absence of protests is uninformative: we cannot tell whether they do not protest because they believe that the government is acting in their best interest or because they fear repression. The usefulness of the concept of representation is limited: when Paris is burning, we see that people do not accept the government as representative, but when Beijing is quiet, we cannot tell whether that holds true.
Is "Support" Information-Contingent or Alternatives-Contingent? In models going back to Schelling (1973) and extending through global games, the calculus made by each individual monotonically depends on the proportion of other individuals opposed to the regime. Such models make prima facie sense, but they rely on the assumption that the size of the opposition is informative about the probability that the regime will fall. The question is whether "support" for the regime is contingent on information or on perceived alternatives, whether people support it because they believe that the regime is good or because they believe there are no alternatives. In Communist Poland most people opposed the regime, but no one believed that communism could fall because the Soviet Army would not allow it. Only when it became clear that the Soviets could not or would not intervene did numbers begin to matter.
As Sen (1993) pointed out, not choosing the same element from different choice sets is not logically inconsistent. What one chooses depends on the options one faces: "An imperative assumption underlying the definition of consistency is that the choice context must be the same when determining whether two or more acts of choice are contradictory" (Mahmoud 2017, 5). 19 People may not like the available options (Sen 1997, 745; think of Sophie's Choice), but one can choose only from the available options. Only when the survival of a regime becomes uncertain does the option to act against the regime become conceivable.
Moreover, how can people evaluate states of the world that they never experienced (Sen 1993)? People in democratic countries are told that autocracy is bad because it is repressive; autocrats tell their people that democracy is bad because it is disorderly; and most people under both types of regimes believe what they are told. Would Mr. Smith from Iowa support the CCP if he lived in Guandong; would Ms. Zhou from Guandong be an ardent democrat if she lived in Iowa? I know no way to answer such questions. The question whether democrats bring about democracy or democracy generates democrats has been studied, but the evidence is scarce and hard to interpret (Maravall 1995, chap. 5). All we know is that we should be distrustful of preferences endogenous to regimes, which unfortunately holds for democracy as well.

In the End
Although this essay may seem highly critical, it is not intended to question recent advances in our understanding of authoritarian regimes, which are incisively summarized by Gehlbach, Sonin, and Svolik (2016). This is just a friendly nudge to delve behind ideological and methodological assumptions that in my view limit our grasp of complex, sometimes bewildering realities of these regimes. My doubts concern whether the models elucidate the puzzle of popular support for autocratic regimes. I fear that, driven by methodological fashions, several models focus on autocratic practices that have a minimal effect on the functioning and the survival of these regimes. I doubt that the conceptual apparatus of game-theoretic models that entail "beliefs" can cope with the psychological processes that make people remain passive, actively collaborate, or actively oppose autocratic regimes. 20 I had a lifelong friend who, after consuming a half-liter of vodka, still exclaimed, "The Party is always right." I could never understand him, and I have not seen a model that would help me do so.
Which of these doubts apply to democracies as well? What is "opinion" in "public opinion"? Is propaganda more prevalent under autocracy than under democracy? Is centralized political repression more dissuasive than fear of decentralized social opprobrium? Are actions based on beliefs or driven by a desire to seek safety in a confirming group membership? Are the preferences that people hold more stable when they are exposed to conflicting messages, as J. S. Mill (1989Mill ( [1859) believed? We do not know the answers to these questions.

Notes
1 See also Przeworski (2019) from which the following passage in taken. 2 Whether authority results from places occupied in some hierarchy, such as the father in a family, or from being able to persuade has been a subject of controversy between Arendt (1954), who held the first position, and Friedrich (1958) who advocated the second. 3 On this topic, I have learned from Arturas Rozenas. 4 I owe this formulation to Carlo Horz (personal communication). 5 The irrationality of this support is wonderfully captured by Victor Serge's (2004Serge's ( [1949) The Case of Comrade Tulayev: the vice-director conspires to have the director sent to Siberia, so as to become director, whose vice-director conspires, and so it goes. 6 Using a series of public opinion data, Matovski (2018) argues that Russians care above everything else about stability and order, which are provided by strongarmed leadership, and they are averse to any alter natives that may be destabilizing. 7 Greene and Robertson (2020, 3) report that the surge of emotional engagement in Russia after the annexation of Crimea "did much more than boost the fortunes of the regime: it transformed the way Russians saw their lives, their futures and even their past, creating a wave of positivity." 8 The only model I found that considers both policy improvements and information manipulation is by Chen and Xu (2015), who argue that they are substitutes. 9 For a blood-curdling allegory of the Chinese regime see the story "Gubaiku Spirit" in Chen (2020). 10 Mercier and Sperber (2011) argue that the main function of reasoning is not to pursue truth but to find arguments that persuade others: "reasoning has evolved and persisted mainly because it makes human communication more effective and advantageous." Propaganda is then just a political extension of this inherent human predisposition. 11 In the Shadmehr and Bernhardt (2015) equilibrium, the ruler never completely censors all bad news but ends up censoring more than it would have had he or she been able to commit to a scheme of information disclosure. 12 I owe this formulation to Andrew Little (personal communication). For the experimental evidence for this assertion, see Enke (2020). 13 See Chen and Yang (2019) for an experimental study designed to get at this question. 14 Czeslaw Milosz's, The Captive Mind (1953), is admirable is sketching the variety of personal trajectories that led Polish writers to embrace communism. 15 In Ernest Jones's classical definition, rationalization is "the inventing of a reason for an attitude or action the motive of which is not recognized." According to Festinger (1957), people invent such reasons to reconcile logical inconsistencies among beliefs or between beliefs and actions. 16 Sofer's (2020) Man of My Time is an exceptionally painful account of such collaboration. 17 Ginkel and Smith (1999) distinguish between two types of protesters. 18 For the despair at the futility of resistance, see Haffner's (2003) 1939 memoir, which was horribly mistitled in English as Defying Hitler; it was called Memoir of a German in the German original. 19 At stake here is the axiom of the independence of irrelevant alternatives. 20 On the general difficulty of distinguishing beliefs driven by "directional motives" from those based on different priors, see Little (2022). On the importance of emotions, see Greene and Robertson (2020).