Practical Conservatism

PRACTICAL CONSERVATISM

Geoffrey Brennan

Alan Hamlin

Australian National University Duke University University of North Carolina

University of Manchester King’s College London

1

I.

Introduction

We begin with two observations. The first is that conservatism involves a status quo bias – at least in cases where the status quo is reasonably stable1. Of course this bias is not necessarily the only feature of conservatism, but we take it to be a feature that is both significant and worthy of explanation.

Our second observation is that conservatism has a more extensive life in political practice and ideology than in contemporary analytic political philosophy. This fact suggests that arguments for a conservative status quo bias, may be more likely to be found in social science than in ethics: in the discussion of social and political feasibility rather than the discussion of ethical desirability.

In other papers we have explored explanations of a conservative status quo bias from within ethics by considering two possibilities: that the conservative takes a distinctive posture towards underlying ethical values, a position we term adjectival conservatism;2 and that the conservative identifies a distinctive ethical value (or values) not recognised by nonconservatives, which we term nominal conservatism.3 In this paper we consider instead possible explanations of a conservative bias that focus on the structure of political interactions and, in particular, on the role of conventions in political life. We term the forms of conservatism that may be located in this way ‘practical conservatism’ to distinguish those forms from ethical conservatism of either the adjectival or nominal type.

If conservatism is practical in this sense, it raises a question as to whether such a conservatism rightfully has a place in political philosophy at all. Hayek (2006) for example at the end of his famous critical essay on conservatism expresses a doubt as to “whether there can be such a thing as a conservative political philosophy.” He explains: “Conservatism may often be a useful practical maxim but it does not give us any guiding principles which can influence long-range developments.” The political philosopher, on Hayek’s stated view, “is 1

Oakeshott 1991 (p.408) refers to a preference for “the familiar” over “the unknown”.

Hamilton 2015 (section 1.1) refers to “fear of sudden change and tendency to habitual action”. 2

See for example Brennan and Hamlin 2004, 2006 and Brennan 2015.

3

See Brennan and Hamlin 2016, and for further discussion Brennan and Hamlin 2014. 2

not concerned with what is now politically possible, but rather with ‘general principles, which are always the same.’” (quoting Adam Smith) [Hayek, 2006, p 355]

There are two notable features of Hayek’s claim. One is that “general principles which are always the same” is the special province of values. The implication is that the realities of the human condition – and valid propositions in the social sciences – cannot provide such “general principles”, a suggestion that would surely have astonished Smith and, we would have thought, ought also to have astonished Hayek himself, if he had thought about it directly. For example, the fact that epistemic constraints will (on Hayek’s own view) afflict any attempt by social planners to secure the kinds of optimific outcomes that can be produced by the market order, seems to constitute just one such ‘general principle’. And Hayek surely believes that those epistemic constraints are capable of giving guidance on “long-range developments”!

So we think Hayek’s claim about the possibility of a conservative political philosophy ought to be rejected at the outset. We ought to acknowledge instead two propositions: first, that social science does provide some generalities4, which are sufficiently robust to provide a basis for judgments about political organization even in the long run; and second that generalities of this type do have a role in political philosophy qua political philosophy.

Now, one might concede the first of these propositions; but dispute the second. One might, for example, simply define political philosophy as being concerned exclusively with the ‘desirability’ aspect of political arrangements. The resultant disciplinary division of intellectual labour would assign pure desirability issues to political philosophy and leave all questions of how politics actually works (or might work) to social science. But that division would imply that neither political philosophy nor social science, each operating in its respective sphere, could ever produce any practical recommendations. There would, for example, be no point in Rawls’ avowed interest in producing a “realistic utopia” because political philosophers could not claim any authority in relation to claims about realism. 4

We refer to “generalities” here rather than “principles” because the latter term, since

Cohen’s influential 2003 essay, has taken on a technical meaning. But we have in mind general facts – and relations between facts – that are sufficiently stable to constitute effective constraints on what can be achieved in terms of any “fact-free normative principles”. 3

Besides, even if the disciplinary divide did track the desirability/feasibility separation, there would remain the important task of putting the two elements together. Taking it as given that providing action-guiding advice always includes some element of “doing the best possible in the situation that obtains”, then understandings of what “betterness” involves and “what is possible” are jointly required for any normative exercise that purports to be action-guiding. In other words if, as Hayek supposes, the object of a political philosophy is to provide “guiding principles that can influence long-range developments” then a sense of broad “feasibility considerations” in political philosophy seems indispensable. 5

In this paper we explore the extent to which general considerations about the nature of social organization and its political expression might provide a kind of defence for the systematic status quo orientation that we take to be characteristic of conservatism. We have designated this strand of conservative thought ‘practical conservatism’ because the conservatism in question is conceived as a practical response to certain facts about human organization. The challenge is to uncover those facts and explain why they may indicate a conservative bias, so understood.

It will be helpful at this point to distinguish between two ways in which a conservative or status quo bias can be understood. On the one hand, the bias in question may be essentially behavioural: certain patterns of behaviour may be temporally stable across the interacting individuals. In this kind of case, to talk of a conservative bias may be misleading. There need be no disposition favouring the status quo as such – the conservatism at issue is indistinguishable from the fact of the regularity. The external observer or social scientist may of course seek variously to explain and/or defend those patterns or regularities. But any normative defence of such patterns or regularities need not play a role in the explanation of their existence: the stability may arise “invisibly”, in the sense familiar from Adam Smith and more recently Nozick (1974, 1994) and Ullmann-Margalit (1978).

On the other hand, it may be that the pattern or regularity does in some measure depend on certain normative considerations, which the participants have internalized and which play a 5

The use of “feasibility” language has become somewhat contentious, especially in debates

over the relation between ideal and non-ideal theory. For our own views on how the term might be understood see Brennan 2013, Hamlin Forthcoming. 4

role in any explanation of the behavioural regularity. Note, however, that in any such case, the normative considerations that help explain the regularity may be distinct from the normative considerations that justify it – somewhat in the manner that rules in ruleutilitarianism may be sustained in part by non-utilitarian grounds for compliance.

The point might be rephrased in terms of recognising two rather different types of norm: moral norms deriving directly from ethical values,6 and social norms deriving in proximate terms from social practice rather than values.7 Social norms may still carry normative force in the sense that they provide reasons for action, even though those reasons are not directly grounded in basic ethical values. Of course, if such a social norm is to be justified, that justification may relate to underlying values, so that a justified social norm will carry both direct normative force as a matter of practice and indirect normative force as a matter of justification. However, social norms will typically carry some internal normative force even when they are not justified. The idea of practical conservatism based on conventions that reflect social practice operates in the world of social norms – with the question of justification, and the links between practical and ethical conservatism, to be explored separately.

Analogously, we will be interested in two distinguishable questions: first, can we throw light on the kind of conservative behaviour that might be understood either as a behavioural regularity or as a social norm? And second, to what extent is that behavioural regularity buttressed (explained, justified) by reference to ethical conservatism? In the latter case at least, we will have grounds for a genuine conservative disposition (rather than merely conservative behaviour).

The point of departure for our more detailed treatment of these ideas is a claim about the importance of convention as a source of social order. The core idea might be construed in terms of Hume’s response to Hobbes. On our reading, Hume can be seen as endorsing the Hobbesian conviction that social order delivers advantages, but as denying Hobbes’ appeal to 6

For current purposes we include prudential values within this set, so that self-interest is seen

as part of an ethical system. 7

This distinction between social norms and moral norms is developed in Brennan, et al.

2013. See also Southwood 2011. 5

any ‘social contract’ in connection with realising those advantages. Ordered society might be, as Rawls observes, a “cooperative venture for mutual advantage”, but neither its existence nor its stability rely on anything like the explicit cooperation that a social contract might be seen to embody. Social order, Hume seems to insist, emerges as a convention and is sustained by convention.

The standard treatment of conventions in the analytic literature involves an appeal to ‘coordination games’ – to structures of interdependence in which individuals’ coordination promises mutual benefits. And we certainly think that that approach provides important insights. But, as is often the case with analytically explicit treatments, some of its advantages lie in the exposure of its own limitations – and we shall stress those limitations in the ensuing discussion. One important limitation, as we see it, is that the game-theoretic understanding of rational action implies that the past as such is irrelevant to choices about the future (“bygones are bygones” as economists often put the point8). By contrast, a critical feature of the convention story is that these “bygones” constitute the central element in making coordination possible.

This then is the agenda – to build from a discussion of conventions in coordination games towards an understanding of at least some aspects of conservatism. In section II we say something about convention as a phenomenon – about the forces that make for securing initial coordination and for ensuring the stability of any coordination equilibrium once achieved. In section III, we expand the simple coordination game to allow for the possibility of rational non-compliance – and thereby open a door for the independent operation of conservative norms. Section IV offers a brief summary and conclusion.

II

Conservatism and Conventionality

The standard philosophical treatment of convention9 stresses path-dependency in selecting a coordination equilibrium. A standard example involves the choice of which side of the road to drive. All players benefit from coordinating on the same practice, whatever that practice is: whether “left” or “right” emerges as the common practice is irrelevant. The standard game is illustrated in Matrix 1 below. The pay-offs to each player are fully symmetric. If the players 8

See Brennan and Hamlin 2009.

9

Following Lewis 1969. 6

coordinate their actions the payoffs to each player are strongly positive, and invariant across (left, left) and (right, right). If the players fail to coordinate, the payoffs to each player are strongly negative and invariant across (left, right) and (right, left). (There is no particular significance attaching to the numerical details). Standard analysis identifies both (left, left) and (right, right) as Nash equilibria of this game.10 In a Nash equilibrium, the choice of action by each of the players is the best response to the choice of the other, so that there is no potential gain to either player from unilateral defection. In the simple two-person interaction depicted, the solution seems simple enough. All that is necessary to secure a Nash equilibrium outcome is a minimal amount of pre-play communication. Since Row and Column both seek to coordinate and are indifferent as to which of the Nash equilibria they coordinate on, either one can simply announce which action she intends to perform. The announcer has every incentive to tell the truth; and the other has every reason to believe her and every reason to choose the coordinating action. For the interaction to be “interesting”, it must be the case that pre-play communication is impossible (or extremely costly). The thought is then that, in the repeated game setting, current action becomes a resource by which players might communicate likely future action. Similarly, choice of action by the two players must be simultaneous. Otherwise, the secondmover will simply watch to see what action the first takes and mimic it. We shall maintain the assumptions of no pre-play communication and simultaneous moves in the remainder of this section.

Matrix 1 Column’s choice Row’s choice

left

Right

Left

(10, 10)

(-10, -10)

Right

(-10, -10)

(10, 10)

10

There is also a Nash equilibrium in mixed strategies where each player selects left and

right with probability 0.5. Here and throughout we ignore such mixed strategies. 7

While the convenient depiction involves just two players, the simple interaction in Matrix 1 stands for a more complex one in which there are many players and many actions, but where full symmetry still applies. In that case, there are many Nash equilibria – one relating to the universal adoption of each of the actions. It is that many-person, many-action setting we have in mind, even though we shall continue to illustrate the analysis in terms of a two-person, two-action game.

In the repeated game setting, the use of the term “equilibrium” may seem to suggest that if both players were to succeed in coordinating (say by choosing “left”) in a particular time period, we should expect them to continue to choose that “equilibrium” option thereafter. And describing such an equilibrium as a convention underlines that thought. After all, one might think, once players have settled on a Nash equilibrium (perhaps after playing randomly over several rounds) neither has any reason to alter her action in the next period. But the truth is that neither player has any reason to stick to her previous action either – no reason, that is, based on expected pay-off maximization. Achieving continuing coordination here depends on the structure of beliefs. If Row believes that Column will choose a particular action in the next round and Column knows that Row believes that, then Column will have pay-off maximizing reasons to act as Row believes Column will act, and Row will have analogous reasons to match Column’s selection. But that structure of beliefs is not supplied merely by a desire to maximize payoffs. That structure of beliefs has to come from outside the structure of the coordination game.

Salience provides one possibility in this connection: if a particular action is recognised as salient, that fact might generate or at least support the belief that others will choose that action, so that salience might provide the basis for a reason for action and for beliefs about the likely actions of others.11 So we might ask: can practical conservatism build on the idea of salience in supplying the relevant structure of beliefs? More generally, what role might conservatism play in solving coordination problems?

11

If one particular outcome “stands out” or is conspicuous in some way it is said to be

“salient”. The role of salience in the emergence of conventions is explored in more detail in Cubitt and Sugden 2003, Sugden 2011, drawing on the basic idea in Schelling 1960. 8

In exploring this question, it is useful to begin by distinguishing two aspects of the coordination predicament: one is the emergence of a common practice and the speed (the number of rounds of play) with which a common practice is established. The other is the maintenance of the common practice once established.

Consider first the emergence and speed of emergence issue, and consider some strategies that might be considered ‘conservative’ in an intuitive, behavioural sense. Suppose first that each player adopts the following strategy: if the outcome in a particular round is ‘satisfactory’ maintain action choice; if the outcome is unsatisfactory, change action choice. It should be clear that if this strategy is adopted by both players in a two-person game it will give rise to immediate (and continuing) coordination with probability 0.5 (that is, just in case the two players happen to select the same action in the first play of the game); and otherwise to perpetual non-coordination (as each changes her action at every round).

Suppose instead that each player is “conservative” in the alternative sense of being slow to adapt and that she always plays the same way for three rounds and only then changes her action if the outcome in the third round is unsatisfactory. It should be clear that if both players adopt that strategy, then nothing is altered: effectively each “round’ becomes a sequence of three identical plays; and coordination and perpetual non-coordination remain equally likely.

Of course, if only one of the players is ‘conservative’ in the sense of playing in blocks of three and that player’s conservatism is common knowledge, then the fact that the nonconservative knows that the conservative will continue the same action for three rounds permits coordination on the second round with certainty. And then the strategy of sticking with success kicks in: the behavioural regularity is established. But it should be clear that dispositional conservatism is neither necessary nor sufficient to secure that outcome.

If we move from the two-person setting to the many-person setting (but retaining the structure of just two available actions) a natural ‘conservative’ analogue of the strategy of ‘continue if satisfactory outcome, change if unsatisfactory’ might be, ‘in each round choose the action chosen by the majority in the previous round’. This strategy then builds on the idea that the majority view carries some sort of salience and, since there will almost surely be a majority in the first round (i.e. the probability of an exact tie is small and decreases sharply 9

with the number of players), this strategy will almost surely lead to the emergence of an equilibrium in the second round. Equally if, in a many-person setting, conservatives play in blocks of three, there might be some advantage in having some conservatives (in this limited sense) around, but only if the identity of the conservatives is known and if all the conservatives manage to select the same action as each other initially – if there is initial coordination among the conservative it will spread rapidly to coordination in the whole population. We might relax the common knowledge stipulation somewhat. A nonconservative player might “discover” conservative players in this setting from the play, in that if a player is observed to play identically over two rounds, this will constitute some evidence that that player is conservative and hence provide grounds for mimicking the action of that player in the third round. And if the prediction is correct, coordination will be established. Still, it is not entirely clear that a similar conclusion would not arise if the conservative disposition here were replaced by any other that involves well-defined action and equivalent knowledge assumptions: predictability, rather than conservatism, seems to be the key.

If we move to the many-person, many-action version of the game, one possible generalization of the ‘conservative’ strategy of following the majority from the previous round would be to follow the most popular action and, to the extent that a single action can be identified as the most popular in round one, this would still lead to the establishment of a practice in round two. But this strategy seems considerably less salient than that based on simple majority in the two-action case. Similarly, if conservatives play in blocks of three, the introduction of multiple actions severely complicates any process by which a common practice might emerge.

Considerations of this sort lead us to conclude that the emergence and speed of emergence of coordinated common practice might be either inhibited or accelerated by the presence of behavioural conservatives12 depending on the precise understanding of how such conservatives act when the situation is out of equilibrium and on details of the game (including the number of players and the number of actions available to each player). That the presence of some behavioural conservatives might assist in the equilibration process is certainly conceivable, but this possibility seems to depend on coordination within the set of 12

Understood as those with a predilection to act in each round as they did in the last. 10

conservatives (for which conservatism itself provides no presumption). And there are other cases where a disposition to behave ‘conservatively’ simply slows the equilibration process. What is needed here is not just some form of behavioural conservatism (or predictability) but the identification within the context of the game of a salient action that may become the focus of conservatives.

As far as the maintenance of any established practice goes, the primary forces making for compliance seem to be: rationality (interpreted here in terms of the desire to maximize one’s own payoff13); and the belief that other players are more rather than less likely to act consistently with any practice that becomes “established”. As we have noted earlier, this latter requirement ascribes a certain kind of behavioural conservatism to agents, as a kind of presumption. “When you’re on a good thing, stick to it” is certainly a folk-conservative nostrum. But it is the fact that players believe this nostrum of others - rather than its truth that is relevant here. The dispositional conservative will, by definition, be disposed to follow that nostrum. Being so disposed will serve to make the established practice more stable - both in itself and because others will know that those who are conservative will follow the nostrum. But in the terms of the interaction as depicted, any such additional reasons do no real work. Once the practice is descriptively ‘established’, each will have reason to believe that others will follow it. (That is what it means to say that the practice is established!) And if each believes that others will follow the practice, each will have purely self-interested reasons to do so herself.

In any such situation, individuals will be observed to be behaving ‘conservatively’ – as if the status quo had independent normative force. But it need not be the case that any of the participants in the interaction has any independent inclination to follow the status quo simply because it’s the status quo. Participants do what they do because they are rational and hold the relevant beliefs. If there are some in the community who do have an attachment to the status quo for independent reasons and if that fact is common belief among the players, that may help account for the belief that (most) others are likely to adhere to the prevailing practice. And the presence of those with an independent attachment to the status quo may bolster the stability of the outcome against random shocks. Such an independent attachment to the status quo might arise as a result of recognising particular conservative value in the 13

It is important to stress that the payoffs need not be understood in self-interested terms. 11

status quo, so that individuals with the attachment would be nominal conservatives in our understanding of that term. All these things are possible; but here dispositional conservatism seems to operate as an unnecessary fifth wheel.

It may be instructive to ask how we might reflect the relevant dispositional conservative value in the terms of the coordination game. One natural thought is to add to the status quo action an additional “conservative” payoff: this extra payoff will motivate behaviour in the interaction but need not figure in the justification (which can remain a matter of the original payoffs). Matrix 2 illustrates for the case in which “right” is taken to be the status quo action and only Row is a dispositional conservative. Note that the conservative payoff α is to be added to the relevant action, not a particular outcome, so that Row receives the additional conservative payoff even if Column does not ‘comply’. Note further that unless α exceeds the payoff differential (20 in this case) then it cannot serve to ensure that conservatism is entirely rational even for Row14. However, the introduction of α does change the game somewhat, making it asymmetric and creating a situation in which (right, right) is a Pareto dominant equilibrium. This change in itself may be sufficient to render the choice of ‘right’ salient (if not fully rational) for Column as well as Row. Nevertheless, this cannot in itself explain the fact that α attaches specifically and only to the action that the ‘convention’ isolates. In other words, a convention identifying which action is to be viewed as the status quo must already exist before α can be deployed. In this way the attribution of any conservative payoff α depends upon history; bygones are not bygones in such a world. The basic point is that, when we say that the real story about coordination involves a mechanism that lies outside the game theoretic structure, we mean that that story cannot be properly told entirely within the standard language of game theoretic representations: it requires a history and a story that explains how that history shapes current beliefs! Without some sense of the relevance of precedent, it is not easy to see how any such story can plausibly be told.

So far, we have taken it that the conservatism relevant for conventions is behavioural in the sense of action choice – so that α is a function of the specific action (“left” or “right”) chosen in the past. But that understanding of conventions may be too restrictive – a convention might require alternating actions, or actions that vary over time in some specific pattern. Indeed, a 14

If Row has grounds for believing Column will choose left in a particular round of play,

Row’s conservative inclinations will be overcome unless α exceeds 20. 12

convention might be defined in almost any terms provided: (i) that the action that the convention prescribes in any particular situation is a matter of common knowledge at the relevant point of choice; and (ii) that there is a common belief that individuals are likely to abide by the convention. So the convention will be ‘stable’ even though the action choices themselves may not be. It may well be that condition (i) is much easier to apply if the action chosen is always descriptively the same. But no such assumption is necessary for our earlier remarks to go through: what is relevant for securing coordination is the propensity to abide by the prevailing convention, whatever that convention may be. Elsewhere, we have noted that a challenge for any conservative political philosophy is identifying that aspect of the status quo that is to be preserved – whether the substantive outcome; or the prevailing policies; or the rules for making policy choices. In the current context, that question must be answered by reference to prevailing beliefs about prevailing practices. 15

Matrix 2 Column’s choice Row’s choice

left

right

left

(10, 10)

(-10, -10)

right

(-10 +α, -10)

(10 + α, 10)

It is worth emphasizing that focussing on the desirability of coordination cannot induce the relevant behavioural adjustments. It looks as if solving coordination predicaments in repeated game settings requires a propensity to follow rules – either the rules emergent from success in previous plays, or the rules stipulated by some appropriately recognized rule-making authority.16 What is crucial is the (widespread) belief that others will follow those rules. However, the fact that such a belief is sufficient for the convention to prevail casts any form

15

As noted in Brennan and Hamlin 2004 the definition of the status quo and which specific

feature of the status quo is relevant for conventionality is a genuine problem for conservative analysis. What a status quo orientation requires is ambiguous whenever the status quo is not well-defined. 16

See Hamlin 2014 and references therein. 13

of dispositional conservatism in a second-order role, if indeed it does not render dispositional conservatism entirely irrelevant.

III

Anti-conventionalism

In order to establish a role for conservatism as a normative disposition in coordination settings, we need to move away from the simple coordination predicament assumed in the previous section.

The first variant of an ‘impure’ coordination game we shall consider is one in which the various Nash equilibria of the coordination game offer different returns – one equilibrium is Pareto superior to the other(s). Initially we retain the assumption of symmetric payoffs. So suppose, in an analogue of the situation depicted in Matrix 1, that one equilibrium (left, left) has a larger payoff to both players than the other (right, right). We show this case in Matrix 3.

Matrix 3 Column’s Choice Row’s choice

left

right

left

(12, 12)

(-10,-10)

right

(-10, -10)

(8, 8)

Clearly there are consequentialist reasons to favour the higher payoff equilibrium over the lower payoff equilibrium. Now, common knowledge of a Pareto dominant equilibrium may be sufficient to establish the salience of the relevant action and so may lead to the establishment of that equilibrium as a convention. But the lower payoff equilibrium remains an equilibrium, and if the lower payoff equilibrium happens to be the prevailing equilibrium, and so constitutes a convention, there will be a tension between that convention and the movement to a superior equilibrium (which then might become established as a convention). This remains true in the case of the larger coordination game involving more actions and more players. As before, the very fact that the prevailing equilibrium is an equilibrium provides it with salience once achieved and some grounds for sticking with it. No individual can unilaterally effect the transition to the target equilibrium, and in the absence of some

14

agency that has the power to mandate the behaviour of all individuals17, there can be no assurance that the process of change from the prevailing equilibrium to the target equilibrium will be either smooth or rapid – it may involve an extended period of “out of equilibrium” behaviour with attendant negative pay-offs. Even if a relevant powerful agent exists – the state being the obvious possibility – we might still expect some compliance issues during any transitional period. Some individuals may doubt that others will treat the powerful agent’s dictates as decisive, and so have reason to abide by the prevailing practice.

A classic example here is the Swedish case in which the issue involved a change from driving on the left to driving on the right. The advantages of driving on the same side of the road as neighbouring countries were self-evident. But many commentators predicted that a change in the road rules in Sweden would cause additional accidents so that the short-term costs of the change might make the change undesirable on balance18. Individuals who did not doubt their own capacities to adjust could reasonably doubt the capacity of other road users to do so. For this and other reasons, the proposed change was not popular: in a referendum in 1955 over 80% of voters voted for the status quo. And when the Riksdag voted in 1963 to proceed with the change, the policy remained widely unpopular with the electorate. In fact, when the shift was finally implemented in September 1967, the accident rate went down slightly, though that effect was temporary. As far as we know, no attempt was made to calculate the overall costs of the change ex post.19 The point here is that even if the transition was relatively smooth and low-cost, and even in the case where both the identity of the target convention and the benefits of a change in that convention were clear, popular opposition to the proposed change could be perfectly rational. Many in the population were, we would suggest, practical conservatives on this matter, exhibiting a status quo bias that can be understood and explained in terms of the salience of the prevailing substantive practice. And here practical conservatism can have genuine normative force. To be sure, that normative force is derivative from various perceived facts. What would distinguish conservatives from progressives here is an alertness to the social capital that is embodied in prevailing practices. Progressives, so the 17

In which case the coordination problem is effectively assumed away.

18

This is precisely the case of conflict between two rival conventions - follow the rule-giver

vs follow the established action. 19

The costs of change are not restricted to the accident rate. Increased travel time and more

rapid car replacement are relevant considerations. 15

conservatives would argue, attend excessively to how the world could be under different circumstances; they attend too little to the costs of change; they conduct their analysis as if the choices were simply ones between alternative social equilibria. Conservatives by contrast are inclined to emphasize that society inevitably starts ‘somewhere’ – in a location that involves prevailing practices on the basis of which agents are able to coordinate their activities. On this view, conservatives and progressives alike share the same basic consequentialist values: and they perceive the various payoffs within the game similarly. But conservatives think that progressives are guilty of an intellectual error – they take a ‘view from nowhere’; they abstract from transition costs, or they under-estimate their size. And, so the ‘conservatives’ will say, this error has significant normative upshots: and so practical conservatives of this type will have some common cause with adjectival conservatives and nominal conservatives who are likely to stand against change in such cases for independent reasons20.

A second variant of the impure coordination game lies in the possibility that different players’ payoffs may differ among outcomes of the game. Indeed, the case in which interactions are exactly symmetric across persons must surely be rare in practice. In general, different persons/groups have different stakes in which equilibrium emerges and there is no single Pareto dominant equilibrium. We illustrate this situation in Matrix 4. Here, Column prefers the (left, left) equilibrium and Row the (right, right) equilibrium. There are three inter-related points about this kind of interaction that are worth distinguishing here.

Matrix 4 Column’ choice

20

Row’s choice

left

Right

left

(8, 12)

(-10,-10)

right

(-10, -10)

(12, 8)

As a referee observes, one difference between those who seek to change to a superior

convention and those who seek to maintain an imperfect one may lie in different discount rates: the latter could have higher discount rates so that ‘transition costs’ will figure more highly in their calculus. 16

One is that players will in general have different preference rankings over coordination equilibria. They all agree that coordination is better than its absence, while disagreeing about the ranking of different coordination equilibria. If an equilibrium arises, and so offers a potential convention that would ensure the continuation of that equilibrium, different individuals may now take different views of the potential convention. For some, the prevailing equilibrium may rank very highly so that they will have instrumental reasons for continuing the status quo over and above the practical conservative reason outlined here; but for others – those who rank the prevailing equilibrium so low that they view it as scarcely better than a failure to coordinate – their instrumental reasons to reform the status quo and move to a preferred equilibrium may overwhelm their practical conservatism. This point illustrates the idea that the status quo bias associated with practical conservatism will not be the only relevant consideration in determining behaviour, and that asymmetries in the structure of the coordination game may also provide a basis for anti-conventionalism or radicalism.21

Note too that in a situation of such heterogeneity, it will not be straightforward to identify any target equilibrium as the basis for any proposed reform away from an existing conventional equilibrium. Even if a majority would be in favour of abandoning the status quo, members of that majority will typically identify different target equilibria and different reforms. This point suggests a further source of stability in existing conventions based not on their simple salience but the lack of salience of any specific or particular reform proposal. It is clear that salience operates in relative terms.

The third point lies in recognising the differential costs of adjustment for different persons in the event of any shift from one equilibrium to another. Clearly those who face high adjustment costs will have greater reason to resist change, so that the adjustment cost argument will work with the grain of their practical conservatism, whereas those who actively enjoy change may find that the “novelty factor” overwhelms any practical conservatism that might be present. To take an example close to our hearts, the specification of particular structures and systems within an organisation such as a university is a form of 21

We repeat the point that payoffs do not have to be understood in self-interested terms –

they can be driven by different substantive normative considerations. 17

coordination game: it is clearly desirable for all members of the same organisation to use the same systems, and it may well be that the shift from one system to another offers net benefits. But it seems clear that the transitional costs of any such shift are unlikely to fall equally on all members of the organisation. Some, perhaps including those who are closest to the design of the systems, may actually enjoy the novelty of a new system over and above any marginal advantages it offers; others may face very substantial transitional costs.

These differential adjustment costs are not readily captured within the coordination game structure itself. The relevant differences do not lie in different payoffs to different people from different systems or conventions once those conventions are fully in place. They are rather reflected in the differential time and cost involved in making the transition. These costs can be captured within the payoff structure by focusing on the non-coordination payoffs. For those for whom the costs of non-coordination are low, the gains for getting the preferred equilibrium are more significant in any rational calculus. Obversely, for those for whom such costs are high, sticking with the status quo will remain the best option.

IV

Summary and Conclusions

Our aim in this paper has to been to explore various aspects of what we have termed practical conservatism – a commitment to the status quo that arises not directly from either the substantive content of values (nominal conservatism), nor from one’s posture towards values (adjectival conservatism) but rather from certain perceived facts about the world. The very idea of practical conservatism presupposes that interaction between issues of ethical desirability and facts about how the world lies are important in any attempt to generate action-guiding recommendations about appropriate political and social institutions; and consequently that relevant facts derived from social science can provide general principles that can constitute elements of a genuine political philosophy.

The kinds of facts that we have taken to be relevant here include the following claims: a) That a significant range of social and political interactions take the form of coordination games; and b) That these coordination games are usefully analysed in terms of the emergence and stability of conventions (or ‘prevailing practices’); and

18

c) That there is an interesting and non-straightforward relation between such conventions and a form of conservatism.

More specifically, our conclusions are as follows:

1.

Agent rationality as such does not give us any clear grounds for thinking that stable

outcomes will emerge or persist in coordination games. The fact that others chose a particular action in the past provides no player with rational reason to think that those others will choose that action in the future; and considerations grounded in the common knowledge of rationality alone cannot lead anyone to choose that action either. Collective success in coordination predicaments requires more. It requires a predilection towards people behaving ‘conventionally’.

2.

Such a predilection can be argued to arise from the salience of the status quo – that is

the salience of a prevailing practice over those that merely could prevail. If we grant that, then fully symmetric coordination predicaments seem to present no major challenge. Everyone will recognize the salience of an equilibrium when it arises, will endow the relevant actions with the status of a convention and stability can be achieved.

3.

The stability of any such convention does not however rely on any independent

normative force ascribing authority to the status quo. It depends rather on the belief that, if there is a prevailing practice, others will tend to follow that practice in the future. This belief may be buttressed if there are some dispositional conservatives in the community but dispositional conservatism is neither necessary nor sufficient for coordination.

4.

In order to make a broader case for dispositional conservatism, there have to be forces

that can disturb the prevailing equilibrium in coordination games. Such forces can arise because:

a. The emergent ‘convention’ may involve lower payoffs for everyone than an alternative possible outcome. That is, there is no guarantee (under whatever processes exist for “equilibrium selection”) that a Pareto-dominated coordination equilibrium cannot emerge. (And even if the emergent convention is initially Pareto 19

optimal, it may not remain so if circumstances change.) The ‘conservative’ claim here is that this fact is not sufficient to justify change. Since we must start from somewhere, comparison of payoff structures is not enough for normative assessment. Comparison of payoffs might seem to suggest that there are benefits on offer; but the fact that a convention is a convention means that there will be potentially significant transition costs in any move from one equilibrium to another and these costs may be large enough to outweigh any advantages in moving to the apparently superior outcome. Anyone who understands the value of rules qua rules will accept the force of this point. And those who are constitutionalists in this sense will have common cause with those who are conservatives for other reasons.

b. Individuals may benefit differentially across different possible equilibria. In such cases, those who do less well under the prevailing convention than they might under some alternative may have incentives to undermine that convention, because ‘anarchy’ increases the chance that their preferred convention will emerge. The motives for change here need not be self-interested. Issues of distributive justice (or other substantive normative considerations) may provide grounds for preferring some other convention to the prevailing one. When practical conservatives stand against such changes, they need not be weighing individual preference satisfaction against other normative considerations. The case for maintaining a prevailing convention can be grounded in the same norms as generate the case for some alternative outcome – in other words, transition costs may be measured in exactly the same normative terms as are used to evaluate alternative equilibria. The practical conservative’s point is just that the transition costs have to be reckoned into any proper calculus and those transition costs may be very considerable.

We should emphasize that there are two tasks that we have not attempted here:

First, we have not shown that political life is dominated by coordination predicaments - only that (prevailing) conventions have a specific role in such settings.

Second, we have not indicated why practical conservatism, as described here, is anything more than intelligent consequential analysis. Of course, practical conservatives will be inclined to emphasize transitional cost considerations that more radical spirits will be inclined 20

to overlook. But practical conservatives may wish to go further. They may want to insist that the relevant practical arguments are not accessible through simple consequentialist reasoning. After all, many self-styled conservatives are inclined to mistrust the kind of consequential framework deployed here: they are inclined to see consequentialist arguments as representing a concession to a kind of foundationalism or rationalism about which the conservative tradition has long been uncomfortable. There may be good reasons for such misgivings, but exploring such reasons must await another occasion.

21

References Brennan, Geoffrey. 2013. "Feasilbilty in optimizing ethics." Social Philosophy and Policy 30: 314-29. ———. 2015. "Buchanan’s anti-conservatism." Public Choice 163: 7-13. Brennan, Geoffrey, Lina Eriksson, Robert E Goodin, and Nicholas Southwood. 2013. Explaining norms: Oxford University Press. Brennan, Geoffrey, and Alan Hamlin. 2004. "Analytic conservatism." British Journal of Political Science 34: 675-91. ———. 2006. "Conservatism, idealism and cardinality." Analysis 66: 286-95. ———. 2009. "Bygones are bygones." Rationality, Markets and Morals: 157-75. ———. 2014. "Comprehending conservatism: Frameworks and analysis." Journal of Political Ideologies: 1-13. ———. 2016. "Conservative value." The Monist 99. Cohen, Gerald A. 2003. "Facts and principles." Philosophy & Public Affairs 31: 211-45. Cubitt, Robin P, and Robert Sugden. 2003. "Common knowledge, salience and convention: A reconstruction of david lewis's game theory." Economics and Philosophy 19: 175-210. Hamilton, Andy. 2015. "Conservatism." In The Stanford Encyclopedia of Philosophy (Fall 2015 Edition), ed. Edward Zalta, N. http://plato.stanford.edu/archives/fall2015/entries/conservatism/. Hamlin, Alan. 2014. "Reasoning about rules." Constitutional Politcal Economy 25: 6887. ———. Forthcoming. "Feasibility four ways." Social Philoosophy and Policy 34. Hayek, FA. 2006. "Why i am not a conservative." In The constitution of liberty. London: Routledge. 341-55. Lewis, David. 1969. Convention: A philosophical study. Cambridge, Mass.: Harvard University Press. Nozick, Robert. 1974. Anarchy, state, and utopia. New York: Basic Books. ———. 1994. "Invisible-hand explanations." The American Economic Review: 314-18. Oakeshott, Michael. 1991. "On being conservative." In Rationalism in politics and other essays. Indianapolis: Liberty Press. 407-37. Schelling, Thomas C. 1960. The strategy of conflict. Cambridge, Mass.: Harvard University Press. Southwood, Nicholas. 2011. "The moral/conventional distinction." Mind 120: 761-802. Sugden, Robert. 2011. "Salience, inductive reasoning and the emergence of conventions." Journal of Economic Behavior & Organization 79: 35-47. Ullmann-Margalit, Edna. 1978. "Invisible-hand explanations." Synthese 39: 263-91.

22