Varieties of ‘rationality’ and the question of their continued theoretical relevance

Intentional action

For some (Hedström, 2005), human agents can be said to be rational simply because (or when) they form intentions and then behave due to those intentions. If an agent really wishes, i.e. intends or desires, to jump off a bridge and does so when the opportunity arises, such action is, according to this definition, rational. This is a very undemanding type of rationality. It only requires of an agent to generate goals and then act in a way that will ostensibly bring them about. Much of what people do, although not everything, due to the existence of unintended reflexes and, more prominently, unconscious motivations and processes (Kahneman, 2011), is rational in this sense. Such a definition of rationality therefore has the benefit of being descriptively realistic in many situations. However, it is also theoretically quite useless because it is much too general to aid us in formulating a more focused, precise explanatory hypothesis about human behavior which could make us understand why a certain action happened (in a given situation) and not a different one. It is also, and relatedly, useless in generating a narrower set of conditional predictions of behavior. One might object that such a broad concept of rationality is not really meant to be self-sufficient, and is merely a necessary starting point or an orienting framework under which more precise notions will be generated. Thus its vacuity is not problematic and might even be a strength. I recognize this but still point out the concept’s weaknesses (if used on its own) for the sake of analytical precision.

Although it was stated above that abstraction is the method for theory-development, not every abstraction will do. Such general ideas as ‘people are rational insofar as their intentions cause their behavior’ are, as said, not theoretically helpful. For example, imagine a dictator and her officer in a situation where it is given that: (1) the dictator asks the officer to collect some taxes for her, (2) a severe sanction is legally stipulated for anyone caught hiding the collected taxes, but (3) there are also no monitoring mechanisms which the dictator could use to find out whether the officer actually handed her all the gathered taxes. Nothing about the likely behavior of the officer can be theoretically deduced from these three given facts and the presupposition of rationality as ‘intentional action’. The notion is just too theoretically open, completely consistent with any course of action, to be of useful application for explanatory or predictive purposes. In contrast, and as will be seen later, stipulating a different kind of rationality, e.g. responsiveness to incentives or even self-interested motivation, would prove itself more theoretically useful.

Acting for ‘good reasons’

Raymond Boudon, a long-standing critic of RCT who was nevertheless a strong proponent of the concept of rationality and regularly praised its usefulness for social sciences (especially sociology) provides us with another general sense of the term. His theory of ‘ordinary’ or ‘subjective rationality’ (Boudon, 1989; 1998; 2003; 2011) holds that ‘people believe that X is true, acceptable, good, legitimate, etc. as soon as they have the feeling that X rests upon a set of acceptable reasons’ (Boudon, 2011: 18). Put differently, if we consider actors to be rational then we suppose they have ‘good reasons’ for their beliefs or actions.

‘Good reasons’ are not, for Boudon, necessarily or even typically derived from a self-interested cost-benefit calculation, nor do they imply that agents are fully informed when making a decision or forming a belief. However, having ‘good reasons’ for action is somewhat less general than simply intending an action. An actor has ‘good reasons’ for her actions if she has arrived at them by feeling that they would also be endorsed by other actors, had they been in her position. It seems that for Boudon being rational means perceiving the world and acting in a ‘trans-subjective’ way. Given an actor’s environment, her experiences and background knowledge, everyone or at least most people would think and act similarly. If this is violated, an action is not rational. This is Boudon’s criterion of rationality.

This formulation of rationality is, like the one before, quite realistic. Much of the time people do form beliefs and carry out actions based on their experiences of the surroundings and background knowledge, no matter how partial or false it is. However, as before, the theoretical output of such a formulation seems very weak. Does this concept of rationality, when it is applied to more concrete examples, help us arrive at precise explanatory hypotheses? What can be said, theoretically, about the dictator-officer example discussed above when relying on Boudon’s notion of ‘ordinary rationality’? The officer, being a rational agent, will have ‘good reasons’ for her behavior. This means she will do what anyone would do in her situation, given the officer’s actual goals, experiences and background information. But what, exactly, does this imply in the situation specified? It is impossible to say. The officer might steal from the dictator or she might be completely loyal to her. We cannot formulate in advance a more precise, non-ad hoc hypothesis about the likely behavior of the officer because Boudon’s rationality postulate is too broad and general. For Boudon it all depends on the goals, experiences and background beliefs of an agent – and he insists we cannot theoretically say what these might be. True, because of its openness the proposition is again realistic and hard, if not impossible, to falsify. But this is not such an unalloyed theoretical virtue as it might seem at first.

Responsiveness to incentives

A third, somewhat narrower definition of rationality holds that an agent is rational if (or when) her demand curves are downward sloping (Caplan, 2006). This means she is willing to buy less of a certain good if its price goes up, and vice-versa, holding everything else (i.e. the quality of a good) equal. More generally, to be rational in this sense means to act in a way that demonstrates an agent’s sensitivity or responsiveness to the incentives (and changes thereof) facing her. In other words, according to this definition an agent is rational if she is pursuing her intended goals in an effective, appropriate manner, given her goals and existing conditions.

For illustration consider, again, the dictator-officer example. If we suppose the officer is responsive to incentives then, whatever her goal-orientation is (be it self-interested or not), she should not be adversely influenced by the stipulated legal sanction. This is so because of an absence of a monitoring mechanism which makes the stipulated sanction toothless. Had such a mechanism been in place it would disincentivize the officer from her behavior. Or to take a more straight-forward example, a student who has the goal of getting to school in a timely manner and has the option to either drive there on a route that takes 5 minutes or on a route that takes 20 minutes, is responsive to incentives if she decides on the former.

It is important to note here that, in contrast to a common misconception, being responsive to incentives does not mean being self-interested. The former is consistent with, but logically distinct from the latter. Altruistic agents who pursue their goal of selflessly helping others in a way that does the most good, given their budget-constraints and what kinds of humanitarian agencies exist, are most certainly responsive to incentives. For example, an altruistic person who seeks out a high-paying, stable job so that she will be able to give even more money away for a longer time, and is then giving it away to a charitable organization which is known for helping the most people per dollar donated, is acting rationally even though not self-interestedly. This third definition of rationality as responsiveness to incentives is narrower than the first two (‘behaving with intent’ and ‘acting for good reasons’) but still broader than rationality as self-interestedness.

Still, as was said previously and as the two examples above intimate, responsiveness to incentives is not a strong enough theoretical assumption that it would enable the scholar, on its own, to formulate a precise explanatory hypothesis. It has to be complemented with a further stipulation of the actor’s goals (e.g. the pursuit of wealth or, more generally, material well-being) as well as a supposition that the actor is knowledgeable of the various incentives and disincentives she is faced with (e.g. the officer has to know that a monitoring mechanism is absent in order not to be disincentivized by a legal sanction). This will be further discussed below.

How realistic is this rationality assumption? Although there is a rich literature on how improperly structured incentives (especially monetary ones) can backfire, research shows people are, in general, responsive to changing incentives, i.e. they try to act so as to efficiently pursue their goals given situational constraints, or to at least pursue them in a ‘good enough’ way (see Kamenica, 2012; Prendergast, 1999). This is especially so if the stakes are of a higher order, if intrinsic motivation is low, and if they have the opportunity to learn from mistakes (Camerer & Hogarth, 1999).

Self-interested action

A canonical and much narrower conception of rationality equates it with the pursuit of self-interest in general and material self-interest in particular. Agents are said to be rational when the goals they are trying to achieve with their actions are such that their own wellbeing is increased as a consequence. There is no obvious empirical ² or purely logical reason why wellbeing should be conceptualized in material rather than psychological terms. Still, rationality as self-interest is usually conceived in terms of material wellbeing such that wealth- or power-maximization are the typical goals of rational agents. There is an important theoretical reason for this stipulation. If self-interest is conceptualized broadly as anything that is of psychological benefit to the individual, theoretical robustness is lost. As with the first two concepts of rationality discussed above, such a move tends to be too general for theoretical, explanatory purposes, even though descriptively it might be true. Because any action whatsoever, or almost any action, can be said to be of psychological benefit to an individual, the generation of clear and precise hypotheses becomes impossible. ³

Consider again the dictator-officer example. What can be deduced from the provided conditions, given the assumption of actors as pursuers of psychological benefits? Perhaps that the officer will steal a part of the tax for herself because this brings her the psychological joy of being rebellious. Or perhaps that the officer will not steal because she feels a ‘warm glow’ when she acts justly and obediently even in the absence of a monitoring mechanism. The situation is theoretically indeterminate if we assume only the pursuit of general psychological benefits.

But just how realistic is it to assume people are self-interested or even materially self-interested? Although we now know from laboratory and field experiments that material self-interest is far from the only thing people care about (Gintis et al., 2005), it is an important motivational goal. Most people are not selfless altruists but what has been called ‘conditional co-operators’ who co-operate only insofar as the other side does the same, never completely losing sight of their own interests. Moreover, according to World Giving Index, Americans make up one of the most charitable nations on the planet, yet they only give 2–3 % of their income to strangers (NCCS, 2015). Furthermore, only a quarter of Americans perform volunteer work and those who do so engage, on average, in only 50–52 hours of volunteer work per year (BLS, 2016). This means that the typical volunteering American works only an hour a week for strangers. Compare this to the standard 40–50 hours of work and more per week that the typical American performs for pay. Surely those percentages and volunteering hours should be much higher if self-interested motives were unimportant.

Full information

The last definition of rationality – a component of what is commonly known and criticized as homo economicus – is the most demanding and unrealistic one. It imbues the agent with possessing all the relevant information regarding her circumstances as well as with the ability to process the information in an unbiased way. This is theoretically implausible because, as will be further illustrated below, if an agent is also responsive to incentives she rarely has the incentive to actually gather all the relevant information or to process it without biases. It is also usually quite empirically incorrect as demonstrated by the now vast scientific literature on cognitive heuristics and biases which help people make a decision without considering all the information as well as preventing them to be unbiased in their considerations.

However, theoretically supposing social agents to be fully or at least fairly informed proves very useful. Without it many exemplary social-scientific explanations would break down, as will be seen in more detail later. This goes even for sociology which is ostensibly the last bastion of social inquiry that was not penetrated by rational choice scholarship. Take, for example, two of the most famous classical sociologists, Karl Marx and Max Weber.

Even though the former is one of the three ‘masters of suspicion’ (Ricœur, 2002), along with Sigmund Freud and Friedrich Nietzsche, and is thus known for his fascinating theory of commodity and capital fetishism, according to which agents are not and cannot be fully informed about their social surroundings due to the inherently mystifying nature of their respective social positions, he does not always embrace this notion. Some of his most fundamental explanatory claims all presuppose informed agents. Consider, for example, Marx’s and more generally Marxist claim that wherever capitalism exists workers will seek out employment – even an exploitative one – because they know in a capitalist society it is better to be employed and exploited than not being employed and exploited (Marx, 1982). Or that capitalists will either strive to enter or at least influence the halls of state power so as to increase their social bargaining position and reduce that of the workers they exploit because they know this is beneficial for them. Also, that they (the capitalists) enter and leave production lines with the up and down movement of the rate of profit because they know bankruptcy is otherwise more likely (Marx, 1991 [1895]).

The same goes for Max Weber (Kiser, 2006) who, notwithstanding his nuanced four-fold ideal-typical distinction between instrumental rationality, value-rationality, traditional behavior and affective behavior, in his most important substantive analyses of historical phenomena, usually presupposes, like Marx, that agents are informed (and also self-interested and responsive to incentives). This is not to say sociology (or any other discipline) does not recognize that people make mistakes, are moved by a-rational concerns and are biased in their reasoning. Far from it, as precisely the Marxist notion of fetishism, Weberian ideal-typical distinctions, or the famous and often used idea of unintended consequences (Merton, 1936) should make clear. But it is to say that the supposition of fully informed or, more realistically, fairly informed agents is more widespread – and more theoretically useful – than is usually thought. This should not be surprising as theorizing a social phenomenon is much cognitively easier, more transparent and tractable with such a supposition. Also and as mentioned before more generally, a simplified analysis based on full information should only represent the first step in research which can and should be gradually further complicated to gain needed realism. Such a method of ‘decreasing abstraction’ (Lindenberg, 1992) goes some ways to rescue us from irrevocably compromising between realism and abstraction.

Bringing the different types of rationality together

To more clearly demonstrate the theoretical power of at least some of the reviewed definitions of rationality, a tri-partite conceptual marriage of rationality as responsiveness to incentives, self-interestedness and full information is in order. Consider, again, the first example of the dictator and her officer. What can we now say about the situation where it is given that: (1) the dictator asks the officer to collect some taxes for her, (2) a severe sanction is legally stipulated for anyone caught hiding the collected taxes, but (3) there are no monitoring mechanisms which the dictator could use to find out whether the officer handed her all the gathered taxes? A simple conclusion follows: the officer will collect as much tax as possible and keep most of it to herself. Now, what can be said about the likely behavior of the officer in a situation where there is a monitoring mechanisms and everything else remains equal? The officer will collect about as much tax as was required of her by the dictator, and she will relinquish roughly all of it. How about if the situation is still different and there is a monitoring mechanism but the legal sanction is absent? The officer will again act as in the beginning, i.e. as a cheater. She is, after all, responsive to (changing) incentives as well as self-interested and fully informed so she knows her deviance in the first case will not be punished (and will pay off), she knows her deviance in the second case will be prosecuted (and will not pay off), and she knows her deviance in the third case will not be punished (and will pay off).

This point regarding explanatory usefulness of the rationality postulate can be further strengthened by illustrating it with less trivial examples from the social-scientific literature. Consider first contemporary theories of democratization, such as the one proposed by Daron Acemoglu and James A. Robinson (2006). These theories usually start with the supposition that there exist two groups of people in any society, elites and citizens, and that the members of these two groups have radically different interests. The elites don’t want a transition to democracy, while citizens would like to see such transition come about. This is so because without democratic institutions elites find it much easier to enrich themselves at the expense of the citizens, while in a democratic society they are more constrained and the needs of citizens are more likely to be taken care of.

Now, the main conclusions of a theory like Acemoglu and Robinson’s are that a move to a more democratic regime is more likely (1) when crises and other shocks – e.g. ‘harvest failures, economic depressions, international financial or debt crises, and even wars’ – hit a society, (2) when landed elites are gone and replaced with industrial capitalists, and (3) when a sizable middle class is present. The idea is that these mediating factors increase the likelihood of democratization because they change the incentives and abilities of both elites and citizens for repression, concession and rebellion.

The first conclusion (1) can be deduced from at least three general conditions. Surprisingly, Acemoglu and Robinson do not provide these reasons in their analysis but they are not hard to come up with. First, crises are plausibly relevant for democratization simply because in times of crises elites’ income generally falls, and a reduction in revenue makes it harder for them to control the army and mercenaries with which they usually repress citizens’ protests and calls for democratization. Second, crises cause intra-elite conflict due to the scramble between them brought by dwindling income opportunities, which leads to less intra-elite solidarity and thus a weakened capacity to act collectively. This weakened elite capacity increases the chances of popular revolts succeeding. Third, in times of crises the regime cannot afford as many benefits (e.g. money or job offers) to dole out to the citizens in return for their continued support which means the opportunity cost of rebellion is lower.

The second conclusion (2) unfolds because industrial or post-industrialist capitalists tend to be less bitterly opposed to citizens’ calls for democratization than landed elites. This is so because in the event of a successful democratization their (i.e. the capitalists’) physical and financial capital, which can easily be moved out of the country or hidden in another way, is harder to tax and redistribute than land. Therefore, capitalists have less to lose by conceding democracy than landed elites. Moreover, ‘social and political turbulence may be more damaging to physical and human capital owners who have to rely on cooperation in the workplace and in the trading process, which makes landowners more willing to use force to preserve the regime they prefer’ (Acemoglu & Robinson, 2006: 32).

The last conclusion (3), which requires the theoretical addition of middle classes to the original analysis, makes sense not only because the comparatively more educated members of the middle class might play the role of leaders in protests and movements struggling for democracy. The middle class also plausibly increases the likelihood of democratization in a society because its presence lessens the degree of threat posed to the elites by the prospect of democratization. Because the middle class is propertied and more prosperous than the ordinary citizenry it will less likely support radical policies of complete wealth redistribution following the transition to democracy. The elites thus know they can count on members of the middle class to act as a kind of buffer between their (i.e. the elite) interests and the interests of ordinary citizens, which makes democratization more tolerable for them.

These theoretical insights into the workings of the democratization process, which are consistent with much historical evidence (Acemoglu & Robinson, 2006), are possible only due to the method of quite strong simplification through abstraction and, more concretely, the stipulation of one or another form of rationality. The theory works because it is assumed that only two classes comprise the society, i.e. elites and citizens, and that their members are responsive to incentives, are self-interested and knowledgeable about their (changing) social environment. This should not be taken to mean that these are the only two really existing classes nor that ‘individuals always act rationally’ (Acemoglu & Robinson, 2006: 19). Indeed, as has been pointed out, this very theory includes at various points in time the addition of middle classes and even inter- or transnational actors. But it does mean rationality had to be postulated and that complex social reality had to have been simplified with the aid of this postulate and others.

The same can be said of an important recent line of research into the workings of authoritarian regimes (Ezrow & Frantz, 2011; Frantz, 2018; Geddes et al., 2018). For example, there is an interesting theoretical observation about the differential rates of survivability of different kinds of dictatorships. Theory predicts and empirical evidence confirms that military dictatorships are the least durable, going on for an average of only nine years (Ezrow & Frantz, 2011: 59), while single-party regimes are the most long-lasting, and personalist dictatorships land somewhere in the middle. Why is this so?

Turning first to military dictatorships, the ruling military elite has a much greater incentive to leave office than a party or a personalist elite because professional soldiers have different interests than civil politicians. The former, because they have secure military careers and posts, prioritize the survival of the military, while the latter, because they have no other secure source of income or power, are primarily concerned with attaining and holding office. This means that military elites are more worried about the collapse of the military than of the regime. As a consequence, if during a military dictatorship the elite starts to splinter and infight ‘most officers prefer to return to the barracks than risk the disunity of the military’ (Ezrow & Frantz, 2011: 59). The probability of a ruling military elite voluntarily leaving office when signs of trouble appear is thus theoretically higher than the probability that a ruling civil elite would do so.

In single-party dictatorships splits and faction formations are comparatively less dangerous because ‘all factions within the party elite corps, and even the main rivals of the leader, are better off if the party remains in power.’ (Ezrow & Frantz, 2011: 59) Because members of the elite in such a regime have a relatively lower chance of gaining income or power in case of regime breakdown, they – even oppositional members of the elite – have a lesser incentive to cause serious strife. Finally, personalist dictatorships differ somewhat from both military and single-party regimes, which could theoretically explain their middling survival status. Even though they are in some respects structurally similar to party dictatorships, the fact that in a personalist dictatorship power is concentrated in only one person and a few of her cronies means that their closest rivals are often marginalized. In the event of a dictator’s death, this institutional vulnerability has the consequence of increasing the probability that the regime will fall and be replaced through a chaotic power-struggle by the marginalized rivals, instead of enduring a smoother intra-regime replacement as is often the case with the more stable party dictatorships.

Rational ignorance

Consider, for example, the long-established fact that most voters are badly informed about how politics works, which policies are proposed by which politicians, whether these policies will most likely work and even about basic economic and political facts, such as the unemployment level, the rate of economic growth and the branches of government (Caplan, 2007; Somin, 2013; Brennan, 2016). The most succinct way of presenting the state of electoral knowledge was provided by the political scientist Philip Converse (2000: 331), who stated: ‘[t]he pithiest truth I have achieved about electorates is that where political information is concerned, the mean level is very low but the variance is very high.’ More precisely, ‘the top 25 percent of voters are well informed, the next 25 percent are badly informed, the next 25 percent are know-nothings, and the bottom 25 percent are systematically misinformed’ (Brennan, 2016: 32).

These empirical findings can, in part, be made sense of analytically by what is called rational ignorance theory (Downs, 1957). It is not that voters are simply cognitively impaired and so cannot either gather nor process the relevant information. After all, the same voters seem to be quite informed when it comes to buying a house or choosing a place to move. It is also not, as some Marxist intellectuals pointed out in the past, that the masses are ideologically duped or manipulated by ‘false consciousness’ (see already Abercrombie, Hill & Turner, 1980). The problem is not so with cognition but with the problematic incentive structure that democratic systems provide the electorate with. An individual has astronomically low chances of ever affecting the outcome of elections in bipartisan systems. Even where the decision comes down to a couple of hundred votes a single individual’s vote is inconsequential. The situation is somewhat different in proportional systems, but even there a single individual’s vote has virtually zero impact on the result (Brennan, 2016).

When this fact of a voter being inconsequential is paired with the fact that acquiring the relevant social-scientific knowledge and properly deliberating on the merits and demerits of particular candidates and proposed economic or social policies, is a time- and energy-consuming process with little intrinsic pleasure, a situation arises in which ignorance prevails for rational reasons. It is not sensible, in other words, for the ordinary citizen to go through months or even years of complex, unpleasant and costly information gathering and processing only to, in the end, have no chance of actually reaping the benefits by getting the optimal candidate in office. This holds regardless of whether the voter is self-interested or altruistic. In either case the vote she cast is inconsequential and will not produce good consequences either for herself or for her compatriots.

In such a situation it is precisely because the voter is responsive to incentives, i.e. rational, that she remains ignorant. Whatever a person, individually, does at the voting booth makes no difference for herself or the society at large – so why become informed? Consider how different the incentive structure is when a person is buying a house or choosing to move. There it is most certainly not rational to remain ignorant, because the benefits of an informed decision (and costs of an ignorant one) will solely impact the individual, which is why most people usually seek out information. Of course, this analysis holds only if we assume voters are responsive to incentives and at least vaguely aware of the fact that their vote is practically powerless.

Rational irrationality

The theory of rational ignorance plausibly explains why voters are ignorant. However, voters not only lack information, but are also epistemically irrational (Caplan, 2007). They are not just uninformed, but misinformed. It is not just that they don’t know much about politics and therefore remain politically inactive. A significant proportion of voters entertain unfounded beliefs, speculations and even wild conspiracy theories about policies and political candidates and apply them in their decision on who to vote for. To account for this theoretically Bryan Caplan (2001) has devised the notion of ‘rational irrationality’. There are some social situations in which being epistemically irrational actually serves one’s interests. Because a single voter cannot influence the outcome of an election in a positive or negative direction with her vote, it not only pays for her not to become informed but also to entertain beliefs which either make her feel good in general or, more specifically, enable her to confirm her identity or signal her allegiance to a certain political tribe.

The notion of ‘rational irrationality’ brings together standard RCT, which assumes instrumental rationality, and behavioral economics, which catalogues psychological heuristics and biases. Caplan’s notion posits that whenever people cannot harm their material interests (i.e. wealth, power etc.) or when the probability of harm is low, they are more likely to indulge their natural biases, such as my-side bias or confirmation bias, and come to epistemically irrational beliefs in order to satisfy their needs for identity, recognition and the like. Here we have agents who are rational in the sense that they are responsive to incentives and are sensitive to what their (material) interest are, but who deviate from ‘rational expectations’ and are neither fully informed nor do they carefully, objectively process what information they do possess.

Miscalculation

However, in many social situations it turns out that agents are less than fully (or even fairly) informed not because of rational ignorance or rational irrationality but due to sheer miscalculation and error. Humans are highly fallible creatures which is why the postulate of full rationality seems so outlandish and unrealistic in the first place. But even here rationality in its various guises can be of theoretical help.

Consider the case of Polish top-down, elite-caused transition to democracy in 1989. A standard theoretical model, based on the three rationality postulates, can be developed along the following lines (provided by Clark et al., 2013: 290–304). ⁴ We start with a few basic substantive suppositions. First, there exist two intra-elite groupings, soft-liners and hard-liners, the former preferring to open up the dictatorship and absorb democratic opposition forces into it (i.e. broad dictatorship), the latter preferring the current exclusive dictatorship (i.e. narrow dictatorship). Neither soft-liners nor hard-liners want democratization however, and can use repression to stifle it should the prospects arise. Second, there exists a democratic opposition which wishes the society to democratize and is willing, under certain circumstances, to take action and organize in order to achieve this outcome. Third, the opposition can be either weak or strong. If it is weak it is unable to organize and push for democracy, if it is strong it is able to do so as long as a liberalizing opening, provided by soft-liners, reveals itself. Fourth, if soft-liners are faced with a strong opposition which has organized and is pushing for democracy they will not repress it, because even though they prefer dictatorship to democracy, they also prefer democracy to full-blown insurgency.

These suppositions along with the three rationality postulates suggest that as long as hard-liners are in power, status quo cannot be broken through a top-down, elite-caused process. However, in the event that soft-liners gain an upper hand on the hard-liners, perhaps due to an economic downturn and subsequent breakdown of intra-elite solidarity, and are thus in a position to begin their project of liberalization, the following two outcomes are possible (Clark et al., 2013: 290–304). The first outcome is a move to broadened dictatorship. This happens in case soft-liners are dealing with a weak democratic opposition. In such circumstances, they will go on with liberalization as they know that the opposition cannot organize and transmute the process of liberalization into full-blown democratization. The opposition likewise knows its own weakness and simply accepts the soft-liners’ liberalizing proposal. The second possible outcome is the continuation of status quo. This happens in case soft-liners are dealing with a strong democratic opposition. In such circumstances, they will not liberalize as they know that the opposition can – and is going to – organize and most likely successfully transform liberalization into democratization. The opposition does not get its opportunity and status quo remains.

Under the stipulated assumptions a process of top-down, elite-caused transition to democracy is not theoretically possible. However, in the real world such processes do tend to happen and are empirically well documented. This is so because the specified model is too stringent. What happens if the assumption of full information is relaxed and we assume, instead, that soft-liners are not always aware of whom they are dealing with? If they think they can safely push for liberalization because the opposition is weak, when in fact it is not, the opportunity will be created for the opposition to organize and push for democratization.

A-rational behavior

In many cases rationality can break down to such a degree that the concept is of even lesser theoretical use. Perhaps the most famous example is the paradox of voting. An individual’s vote is inconsequential which should, if people were always responsive to incentives and fully informed, discourage her from ever showing up at the voting booth. Yet not only do a few individuals vote, but (depending on the country) hundreds of thousands, millions and even tens of millions do so, and do so election after election. This is true partly because people are not always responsive to incentives but act in a deontological way to fulfill what they perceive to be their duty. An ostensible citizen’s duty to vote is, for some, such that it should be carried out even though no negative or positive consequences follow from either participating or abstaining. Others vote not because they feel they have a special duty to do so but simply because they mistakenly believe their individual vote to have a significant impact (Blais, 2000). Research does show that closer elections – elections for which the voters know (due to news reports on polling data) that the race will be tight and whose eventual results will be close to be even – draw higher voter participation (Blais, 2000) which suggests that the notion of rationality might not be completely useless, but surely its power is significantly lower than in other cases which have been reviewed above.

This example and others like it demonstrate that some behavior is not only miscalculated or irrational but simply a-rational, non-rational. More generally speaking, norm-following is a whole class of social actions which tends to be seen as a-rational in nature. Sometimes norms are, of course, followed for prudential reasons or rational ones having to do with incentives and interests (more on that below). But other times they are followed for their own sake, independent of the likely consequences their acting out will have for the actor or anyone else. Rationality and norm-following can be completely orthogonal to each other. Not surprisingly, then, a common criticism of scholars who promote the notion of rationality for theoretical reasons is that because cases of norm-following are not sufficiently explained by it, social-scientific theories based on rationality are necessarily incomplete. This is true. But critics sometimes also imply that there are simple theoretical alternatives to rationality which can be used to cover norm-following (e.g. Shapiro & Green, 1994). It is suggested we should turn to notions of socialization, ideology or culture for theoretical aid in such cases.

This, in fact, is not as satisfactory or promising as it might seem. As Dennis Chong (1996: 47) pointed out in his response to Donald Green and Ian Shapiro’s criticism of RCT:

However, the speculative factors they name are not explanatory theories. They are variables that might have causal impact via some unspecified mechanism. [. . .] Green and Shapiro provide no substitute psychological theory of individual agency nor any discussion of what gives institutions and norms their causal power over social action (e.g., do they influence behavior by altering individual incentives, changing preferences, providing information, coordinating beliefs, or what?).

To point out that voters vote out of a sense of duty, that a war was started due to the presence of militaristic culture, or that youth obedience is on the rise because of different socialization patterns might be descriptively accurate, but it is of little explanatory insight. This is the difference between knowing, naming and describing the variables or factors of a phenomenon, on the one hand, and explicating a fine-grained theoretical mechanism through which these variables or factors ostensibly act, on the other.

Jon Elster, a prominent contemporary critic of RCT, has been painstakingly cataloguing many cases and variables of norm-following for decades. However, when he asks himself the theoretical question of what are the general mechanisms through which norms operate or ‘why do specific norms exist in specific times and places’, even he admits that he has ‘nothing to offer by way of an answer. I do not even know where to begin to look, or what general form an answer might take.’ (Elster, 1991: 127) He suggests a simple descriptive, inductive strategy whereby one would collect and examine many different cases of norm-following from different societies. This, he says, may in future ‘provide tentative generalizations, which in turn may suggest possible causal mechanisms.’ (Elster, 1991: 127)

Frustrated with this inconvenient fact, some scholars have devised ingenious analyses, based on one or another kind of rationality, to nevertheless try and explain why certain norms exist and are adhered to. A notable example from the past is the anthropologist Marvin Harris (1974) who famously rationalized away Hindu cow worship and post-medieval beliefs in witchcraft. He did so by demonstrating that these beliefs were at least socially, collectively beneficial in some roundabout way, even if they were not individually rational to pursue. A contemporary example is the economist Peter T. Leeson (2012a; 2012b; 2013; 2014; 2018; Leeson, Boettke & Lemke, 2014) who provides elegant rationalist explanations for such irrational or a-rational seeming phenomena as witch trials, beliefs in oracles, the practice of wife-selling, divine ordeals, monastic maledictions, or, perhaps most bizarrely, vermin trials. In the latter insects and rodents were accused of committing property crimes and were tried as legal persons in Italian and other ecclesiastic courts. Leeson not only spells out theoretically what the ostensible social (and, in contradistinction with Harris, even individual) benefits of such practices are but also tries to corroborate his explanations with copious qualitative empirical evidence and statistical regressions.

Although interesting, the main problem with such theorizing usually turns out to be similar to that which pervaded functionalism. It is not all that hard to come up with at least a somewhat plausible account of how many norms which are individually irrational might be beneficial to the society as a whole. What is much harder to do is to provide a mechanism which would causally link these putative collective benefits to the perpetuation, and especially emergence, of a norm. (Elster, 1991: 125) Of course, the preferred mechanistic building blocks of those who theorize with the concept of rationality, i.e. responsiveness to incentives and individual self-interest, are in many cases not allowed ‘since the benefits are collective rather than individual.’ (Elster, 1991: 125) But in the absence of any alternative mechanism we have no falsifiable explanation at all and the collective benefits of a norm might simply be accidental, while the real causal pathway which accounts for the emergence or perpetuation of a norm remains unknown.

This is not to say that explaining norm-following as rational works only in marginal cases. As stated before, so long as acting in accordance with norms confers individual benefits rational individuals will follow them. In such situations rationality is a plausible, although to be sure not necessarily true, explanation for norm-following behavior. ⁵ For example, usually each individual is personally better off following norms of safe driving on the highway. Here, collective and individual interests are not misaligned, and so there is no theoretical mystery why rational individuals would obey such norms. Most of Leeson’s, although not Harris’, cases fit in this category. Furthermore, even in cases of norm-following where there is a disjunction between collective and individual rationality, it can still be argued that people follow norms because of individual incentives that might emerge in certain interactions. For example, those who are not following various norms of cooperative behavior – which is otherwise collectively beneficial –, might provoke angry reactions, even confrontations and ostracism, from bystanders and will thus incur high costs, i.e. direct punishment or loss of a good reputation. Here, even though it is at first only in the collective (not individual) interest to act in accordance to the norm, subsequent active norm-enforcement by other people helps to align group and individual incentives. Holding to a norm is, therefore, again rational.

However, this same example reveals the main two theoretical issues that hobble at least some cases of norm-following. For we first have to ask what incentive do norm-enforcers have to enforce the norm against shirking which, if not enforced, is only collectively rational and thus will not be followed by rational individuals. If people have an incentive to free-ride, and if they abstain from free-riding only when incentivized to do so, why would norm-enforcers themselves not free-ride with regards to the act of punishment? Two solutions come to mind. First, the nature of the situation might just be such that certain people gain more by enforcing the norm than it costs them to perform such punishment. On net, they gain. By punishing others and thereby incentivizing them to obey norms of cooperation, they themselves end up better off. The trouble is that this will not always be the case. Sometimes the situation people find themselves in is simply such that it is universally rational to free-ride. The cost of enforcing the norm is higher than the benefits accrued, so norm-enforcement is not rational. However, and secondly, a rational solution still might be possible. What if someone comes along and incentivizes would-be enforcers to enforce the norm? If there exists a class of people who could be called ‘enforcers of norm-enforcement’ the puzzle is temporarily solved. But not really, because the problem is only pushed one layer down where the same issue will emerge again and lead to an infinite regress. What, in other words, is their incentive to ‘enforce norm-enforcement’?

Moreover, and relatedly, what about cases of norm-following where the norm is obeyed even though nobody was watching, ready and willing to (rationally or not) punish deviants? For example, some people simply pay their taxes, even though they know they could not get caught if they broke the law and even though they are aware that their individual contribution to the state budget is virtually inconsequential. I conclude that in certain cases and at some point at least some people have to either follow norms, or enforce norm-following on others, purely out of an emotional reaction, a conscious or unconscious deontological duty or an ingrained habit.