Legal Transitions without Legitimate Expectations*

SHOULD slave-owners have been compensated when slavery was abolished? Should the current owners of gas-guzzling SUVs be exempt from new climate regulations? Should workers in protected industries receive subsidized retraining when trade protections are liberalized? These questions, which pit the value of legal stability against the value of legal change, are but a few tokens of a general normative question that has been dubbed the ‘problem of legal transitions’: under what conditions (if any) should governments offset changes in value caused by changes in the law? In a world where laws change frequently—and many of our biggest contemporary challenges demand large-scale legal reform—the question is a vitally important, yet surprisingly under-theorized one. Recently, this neglect has begun to be redressed by liberal-egalitarian political philosophers who have explored the concept of ‘legitimate expectations’ (hereafter LE) as the basis for a principled solution to problems of legal transition. In this article, I seek to dampen the enthusiasm about the viability of LE for this purpose. Before previewing my argument, it will help to provide a sympathetic

reconstruction of the theoretical motivation for resolving legal transitions by reference to LE. To do so, it is necessary to first consider the role that expectations play in individuals' prudential good, and in the domains of interpersonal morality and private law (domains in which the application of LE is, in my view, entirely appropriate).
Since at least as far back as Bentham, 4 political philosophers have remarked upon the significance of expectations in how well people's lives go. Specifically, an agent's predictive expectations about the future have been linked to their prudential ability to undertake long-term planning, which is widely thought to be important to practical agency, autonomy, and wellbeing. 5 As Bentham noted, humans are temporally extended beings, with a characteristic desire to connect the past, present, and future elements of their lives into a coherent whole. 6 Having long-term plans is thus thought to provide us with present reasons for action, facilitating our day-to-day practical agency and rendering us autonomous, selfgoverning agents. 7 Some theorists, moreover, take living autonomously in accordance with one's life plans to constitute wellbeing, or one's personal good. 8 Without the capacity to form expectations about the future, our ability to planand, thereby, our ability to realize these various goods-would be drastically impaired.
Moral, political, and legal philosophers of LE, however, are not interested in predictive expectations writ large-encompassing the weather, or the laws of gravity, for instance-but, rather, in predictive expectations about the behaviour of other agents that are in some way normatively justified. 9 Only for this subset of expectations can the losses that occur when such expectations are frustrated be tied to a (moral, political, or legal) obligation on the part of the other agent: that is, an obligation to avoid frustrating the expectation, or to remedy the loss when the expectation is frustrated. It is only this subset of expectations that can, in other words, be classified as legitimate expectations in the relevant sense.
In fact, the domain of interest to LE scholars can be refined further still. Let us continue to leave aside for a moment the issue of legal transitions and think about normatively justified predictive expectations about the behaviour of agents in the domains of interpersonal morality and its legal cousin, private law. The moral and legal toolbox already contains numerous conceptual devices which function to render others' behaviour predictable-most notably promises and contracts. Accordingly, moral theorists interested in LE 10 have focused more narrowly on what Sidgwick called the 'dim borderland' of 'tacit understandings' and 'implied contracts' arising out of past behaviour. 11 On one side of this uncertain moral territory lie expectations generated by explicit promises or contracts, which are clearly normatively protected. On the other side lie the myriad expectations we develop about others' behaviour, on the assumption that it will continue to conform to a past regularity, which are simply frustrated in a world in which people and things inevitably change, but which are not protected because there is no normatively significant fault by another agent. 12 Sidgwick admitted that 'I know no intuitive principle by which we could separate valid claims from invalid, and distinguish injustice from simple hardship'. 13 It is the job of a conception of LE to furnish such a principled basis for dividing one from the other, and it is the normative core of such a conception that I shall refer to as the 'legitimacy basis'. 14 In the cases of promise and contract, the equivalent work of the legitimacy basis is done by the promise-or contract-constitutive speech act. It is the absence of any such explicit beacon that makes the borderland of tacit understandings so dim.
So what has all this got to do with legal transitions? The thought is that the law, being a pervasive set of rules regulating social interaction, plays an especially important role in shaping people's expectations, and hence enabling them to make plans. 15 Changing the law can therefore cause normatively significant losses to those who based their plans on the expectation that the relevant law will continue in its current form. Sidgwick thought that these kinds of losses occupied an uncertain terrain in political philosophy similar to the 'dim borderland' of expectations based on regularities in interpersonal conduct. 16 Others have gone further. Feinberg has likened legal changes to 'chang [ing] the rules in the middle of the game', 17 while Simmons refers to people having 'the rug pulled from beneath them by sudden institutional change'. 18 The recent crop of theorizing about LE in the realm of legal change, as shall be clear from my subsequent analysis (especially Section IIB), has a similar flavour to it, in the sense of at least starting with the presumption that many of the expectations of legal stability that people hold are legitimate. For example, Meyer and Sanklecha consider whether climate change laws that reduce agents' legally permissible greenhouse gas emissions violate those agents' LE about the level of greenhouse gases they will be able to emit (with adverse implications for the value of emissions-intensive projects on which they have already embarked). 19 The implication is that having one's legitimate expectations of legal stability frustrated by a change in the law gives one a normative entitlement to some kind of remedy from the state, such as compensation for resultant losses. 20 I argue that the attempt to scale up the concept of LE from the interpersonal domain to the high-political domain of legislative transitions is misplaced. 21 By way of preview, my argument encompasses the following three claims. First, the concept of LE is best understood as a species of special rights whose function is to adjudicate normative disputes about the frustration of expectations arising from practice-governed interpersonal interactions. Second, the features of this model of LE that make it successful in fulfilling that function (namely, what I shall call practice-dependence and expectation-dependence) give rise to the following two necessary 'conditions of application', 22 which effectively limit the domains in which the concept can be applied:  20 Precisely which kind of remedy is required is a separate question, and compensation is but one possibility among others, including grandfathering (exemptions for those who had relied on the legal status quo) and adaptive assistance; see Green, 'Who should get what when governments change the rules?', ch. 9. The focus of this article is, however, on the prior question about the conditions under which an agent has a LE of legal stability. I assume that having a LE grounds a normative claim to some state remedy. 21  2. that the relevant expectations can permissibly be identified by the relevant decision-maker (the expectation-identification condition).
Thirdly, while these conditions can be met in the domains of interpersonal morality and private law, they cannot be met in respect of the most common and consequential kind of legal transitions, namely characteristic legislative enactments. With respect to condition (1), discussed in Section II, the characteristically general, institutional, and impersonal character of legislation makes it impossible to identify a practice-governed interaction that could ground the legitimacy of agents' expectations of legal stability. With respect to condition (2), discussed in Section III, the fact that legislation is an act of state that characteristically applies to large numbers of persons with heterogeneous expectations, plans, and projects means that the state would need to undertake intrusive, morally costly investigations of people's lives, including their mental states, in order to determine their transitional entitlements by reference to LE. I show that models of LE that define the legitimacy basis in a practice-independent way and that are not concerned with agents' actual expectations-models that thereby could be applied to legislative transitions-are poorly theoretically motivated and produce counterintuitive results.
The upshot of my argument is that attempts to apply the concept of LE to characteristic legislative transitions 23 involve a kind of category error. I conclude, in Section IV, that the development of satisfactory solutions to the problem of legal transitions should be addressed through other concepts, principles, and theories-ones that are better suited to the unique features of characteristic legislative transitions.

II. LEGITIMACY AND THE 'LEGITIMACY BASIS CONDITION'
A. Practice-Dependent Conceptions of LE: A Domain-Limited Defence The central feature to be specified in any conception of LE is the legitimacy basis. I begin by returning to the domains of interpersonal morality and private law to sketch the contours of a practice-dependent model of LE, which I argue is the best model for the legitimacy basis, insofar as it is theoretically well motivated and produces intuitively correct results. 24 The following two examples should help both to elucidate the core of the distinction between practice-dependent and practice-independent models of LE, 23 I will use the terms 'characteristic legislative enactment' and 'characteristic legislative transition' interchangeably, since any new enactment entails a transition from the legal status quo. 24 Elsewhere, I specify and defend a particular conception of LE based on this practice-dependent model; Fergus Green, 'Illuminating the "dim borderland" of tacit understandings: a unified conception of legitimate expectations', MS. For present purposes, however, all I need to claim is that the practice-dependent model is the right model of LE. and to establish the superiority of the former. The first example is given by Meyer and Sanklecha: Two housemates A and B enjoy having dinner together and for a long time have had dinner together on Fridays. They take turns in preparing dinner, and if A prepares dinner this Friday because it was her turn, then A has acted thus because of her expectation that she will have dinner with B … [One Friday evening] B does not turn up … 25 The authors use this example to pump the intuition that, when B does not turn up, 'A can be said to be harmed to some extent because of the frustration of her expectation', and B can be said to have done something prima facie morally wrong (and which is aptly captured by the notion of B's having frustrated A's legitimate expectation). 26 The second example is from A. John Simmons (writing on a different topic): 'a year of Kant's daily walks through town creates in the Konigsberg housewives the reasonable expectation that they will be able to set their clocks by his passing'. One day Kant decides to stay home to read Rousseau. 27 Simmons uses this example to pump the contrary intuition: whatever disruption this may cause the 'housewives', it involves no moral failing on Kant's part. 28 In both cases the expectations matter (albeit fairly trivially so) for the relevant agents' ability to plan, and therefore to realize the associated prudential goods of planning discussed in Section I. However, many readers will likely share-as I do-the divergent intuitions that the respective authors are trying to pump in each case. Wherever the borderline lies in Sidgwick's 'dim borderland', these two examples seem to fall squarely on opposite sides of it. We may safely surmise, then, that Housemate A's expectation was legitimate in the relevant sense, but the expectations of the Konigsberg women were not. The question of interest is: what is the normatively relevant distinction between the two cases? What makes one set of expectations legitimate, but not the other?
I propose that the distinction lies in the fact that in the housemate dinner case, the legitimacy of the expectation arises not merely from some past state of affairs or behavioural regularity per se, but from the fact that that state of affairs or behavioural regularity accords with the norms internal to a social practice in which the relevant agents were mutual participants at the relevant times.
To flesh this out, I adopt Haslanger's conception of social practices as: patterns of learned behavior that enable us … to coordinate as members of a group in creating, distributing, managing, maintaining, and eliminating a resource (or multiple resources), due to mutual responsiveness to each other's behavior and the resource(s) in question, as interpreted through shared meanings/cultural schemas. 29 Among the elements emphasized in this definition (and in Haslanger's surrounding discussion), two stand out as central: resources and social coordination. Haslanger defines resources very broadly to mean things that have a positively (or negatively) valenced social (dis)value, be it economic, aesthetic, moral, prudential, or spiritual. 30 For our purposes, we can assume that resources (in this very broad sense) are at stake, for otherwise there would be little reason to be concerned with the frustrated expectations caused by legal transitions; the more important definitional element is 'social coordination'. Practices coordinate human action by encouraging or enforcing some behavioural regularity; they have what Haslanger calls a 'descriptive normativity ', 31 or what might also be called an internal normativity (as distinct from the 'evaluative', or external, normativity more familiar to philosophers). 32 This internal normativity arises from the semiotic concepts, scripts, and meanings that comprise culture (also referred to as 'cultural schemas' or 'social meanings'), which set expectations about the right way for people to behave in a given context and evoke mutual responsiveness among practitioners. 33 In this way, practices 'set the stage' for human agency, enabling and constraining it by providing social roles to perform, reasons to perform them, and scripts and tools to perform with. 34 In short, we may say that, as a matter of social theory, our predictive expectations of other agents are (internally) justified by reference to the descriptive norms of social practices.
For the purposes of interpersonal interactions, assuming no problematic externalities, I see no good reason why the descriptive, internal normativity of practices should not constitute the legitimacy basis for conceptions of LE, and thus for LE-based moral and legal principles (I will address the 'no problematic externalities' assumption shortly). In other words, in these domains, the internal, descriptive normativity of the practice should be the source of the evaluative, external normativity ascribed by moral and legal theory. Why? Because participating in social practices is an important means through which agents intersubjectively shape their normative environment, and such normative shaping is an important feature of individual autonomy and wellbeing, including the capacity to form and sustain social relationships. 35 A society that values autonomy and wellbeing-and, in particular, its social-relational aspects-therefore has 29  reason to accord (external) value and protection to the (internal) normativity of social practices.
This kind of normativity is often characterized under the rubric of special rights and obligations. 36 Analogously, promises and contracts are conceptual devices by which individuals create special rights (with corresponding special obligations) that shape their normative environment. A function of our public doctrines of promise (moral) and contract (legal) is to provide external normative validation of the particular, explicit commitments that agents make to one another using these devices. 37 In advocating a practice-dependent model of LE, I am merely proposing that LE be understood in similar terms, as a species of special rights (with corresponding special obligations), similar in nature to the special rights that arise from a promise or contract, but which occupy and illuminate the 'dim borderland' of tacit understandings that lies beside these more familiar conceptual devices. 38 The practice-dependent model provides a compelling basis for distinguishing our two example cases. Housemate A can legitimately expect (predictively) that Housemate B will cook dinner on Friday, because they are involved in a social practice in which it is (tacitly) understood by both parties that this is what B should (normatively) do, where the normativity of the 'should' refers to the internal normative standards governing the practice itself. 39 In the case of Kant's walks, the housewives predictively expect Kant to walk by their houses on the day in question, and so they organize their affairs accordingly (to their detriment, on the occasion that Kant stays home). But there is (on the facts available) no intersubjective understanding shared by Kant that he should partake in his daily walk at this time every day; there is simply no social practice from which to generate any normative obligation on Kant to arrange his walks at the relevant time. The housewives' expectations therefore simply cannot be legitimate in the relevant sense. 40 Before considering how this practice-dependent model fares in legislative transitions, let me address a sceptical question of a kind that practice-dependent approaches to normativity must always confront: how can the internal norms governing a social practice be externally validated where the practice itself is 36 H. L. A. Hart, 'Are there any natural rights?', Philosophical Review, 64 (1955), 175-91, at pp. 183-4. 37 Seana Shiffrin, 'The divergence of contract and promise', Harvard Law Review, 120 (2007), 708-53. 38 The common law of equity also recognizes quasi-legal obligations of transactional consistency that stem from certain kinds of behavioural interactions between agents that fall short of contracts-e.g., via the doctrine of estoppel. 39 Of course, the practice must be interpreted in order to divine its internal normative standards and determine whether the expectation actually accorded with those standards. But on the facts provided by Meyer and Sanklecha, it seems to be a clear-cut case in which there was an established practice that Housemate B violated. 40 Meyer and Sanklecha's second example-'a thief steals a car and forms the expectation that he will get away with the theft' ('How legitimate expectations matter', p. 370)-is amenable to a similar analysis: there is no shared social practice among the thief and the victim (or the thief and the state) that could ground a LE. morally problematic? The answer is straightforward. As with other areas of interpersonal morality and private law, there are external, public-interest-based limits to what individuals ought to be able to do to and for one another in their private interactions. A key reason for setting such limits is that private interactions are rarely, if ever, fully 'private', since they will have implications for third parties, also known as 'externalities'. One purpose of political philosophy and public law is to set those limits, for example by determining when externalities are sufficiently problematic to warrant the curtailment or reconfiguration of private activity. 41 My point is that these limits are best understood as side-constraints on the otherwise permissible shaping by individuals of their normative environments. This side-constraint approach to problematic externalities is how lawyers and legal theorists typically approach the public limits of private law.
Consider, for example, how the common law deals with contracts the subject matter of which is the commission of a crime: should one of the parties seek to enforce such a contract in a court of law, it will be deemed null and void because it exceeds public interest limits on private attempts to shape the normative environment. This view acknowledges that there may be distinctive reasons to override particular contracts, such as the protection of important third-party interests, while preserving the distinctively private (or 'special') function of, and justification for, contracts in general.
Having sketched and defended the practice-dependent model of LE, I will proceed to consider the limits of its applicability. The practice-dependent model of LE is inherently limited to domains in which it is sensible to speak of the relevant agents (the expecting-agent and the agent whose change of position violates the expecting-agent's expectation) as mutual participants in a social practice. It is this consideration that motivates my first condition of application for conceptions of LE, viz. that the legitimacy of the relevant expectations is determinable by reference to the internal norms of a social practice in which the agents were mutual participants.
Characteristic legislative enactments cannot meet this condition. 42 Legislation is a public institution that characteristically applies generally and impersonally to all agents in the relevant jurisdiction. Accordingly, the entire corpus of a jurisdiction's laws exerts a pervasive influence over the expectations, preferences, and actions of a large, open set of individuals. The relationship between individuals and the state qua legislator is therefore not 'one-shot' or even 'repeat-play', but rather 'continuous-play', entailing the ubiquitous conferral by the state of costs and benefits to individuals across multiple domains and over their entire lifetimes. 43 In these circumstances, it is erroneous to speak, in each instance of a 41 The soundness of this response is not premised on acceptance of any metaphysical bright line separating public from private. It is simply premised on acceptance of the utility of distinguishing between public and private in the way the normative landscape is carved up. 42 I explore elsewhere whether LE can, in light of my proposed conditions, be scaled up to certain other forms of public decision making, including administrative decisions and even non-characteristic legislative transitions; Green, 'Illuminating the "dim borderland" of tacit understandings'. 43 Levinson, 'Framing transactions in constitutional law', p. 1333.
legislative change, of individuals and the state qua legislature as participants in an interaction governed by a social practice. Of course, individual citizens may expect that a particular law will remain on the books, and, where the expectation is frustrated, losses may result (absent transitional assistance from the state). However, the generality and impersonal nature of characteristic legislative enactments precludes any intersubjective understanding-any tacit agreementbetween particular citizens and the legislature that could normatively underwrite those expectations. At the very least, the onus lies with those appealing to LE in the legislative context to substantiate the relevant practice and pinpoint the shared understanding. But when we consider what such a shared understanding could amount to, the possibilities don't seem promising. One potential candidate for the content of this supposed shared understanding is the notion that laws, once made, will never change; that they will remain on the books indefinitely. But this is implausible. It is in the nature of a legislature that it make, and hence change, laws-all the more so given that its composition changes as a result of regular elections or otherwiseand this renders untenable the idea of a tacit agreement between state and citizen that any particular law will never change. 44 An alternative candidate for the putative shared understanding between citizen and legislature might be something like the notion that laws will not be changed unfairly, unreasonably, unjustly, without due procedure, or the like. But notice that this move would moralize the notion of a shared understanding (and hence a practice). We would no longer be speaking of a practice in terms of a thick behavioural pattern-like cooking dinner on a Friday, or going for a walk at a certain time in the afternoon-but rather in terms of a thin, abstract normative standard. The notion of a social practice would thus become a mere vessel for an objective normative requirement binding the legislature. But if the legislature has normative obligations to act fairly, justly, and/or reasonably (and so on) with respect to transitional issues when it passes laws-and I think that it does-then such obligations should stand independently of whether any particular citizen expects the legislature to act in this way. Once one specifies the relevant standard of fairness, justice, reasonableness, or due procedure (and so on), then it is difficult to see what normative work is left for the concept of LE to do. Note that, in the course of this discussion, I will appeal to the intuitive correctness of providing state assistance to certain kinds of losers from legal transitions in certain situations. As flagged in the final paragraph of Section IIA above, I think it quite plausible that a theory of legal transitions could require the state to mitigate the transitional effects of legal change in some cases. This view is entirely consistent with the core argument of this article, which is that LE is not the right concept for this role. It follows that this role should be filled by other concepts and principlesones that are appropriate to the unique, public-institutional circumstances of legislative change. Detailed discussion of what those concepts and associated principles should be is beyond my scope here; however, towards the end of this subsection I will briefly sketch some plausible alternative candidates in order to dispel the idea that (practice-independent) LE exhaust the possible options. 45 On a practice-independent model of LE, the legitimacy basis has nothing to do with the internal norms governing social practices. The advocate of such a model must then locate the legitimacy basis in some other purportedly 'normatively salient feature' that recurs in cases of frustrated expectations. 46 Let us first consider one of the LE conceptions discussed 47 by Meyer and Sanklecha: the 'Normative Authority View'. 48 On this conception, '[i]f the political authority in charge of maintaining the background institutions and of ensuring widespread compliance is legitimate, then so are the expectations generated by those institutions and that compliance'. 49 Whether or not some agent's expectation of legal stability is legitimate, then, depends on the correct (or rather, the theorist's preferred) theory of legitimate authority.
The first concern to note about this view is its theoretical motivation. Why should the legitimacy or otherwise of the state authority that makes the law determine the legitimacy or otherwise of agents' expectations that the law will stay the same? Meyer and Sanklecha's answer to this question is as follows. The state plays a significant role in the formation of people's expectations; it does so through the operation of its coercive mechanisms, and the question of legitimate authority is essentially the question of the conditions under which such coercion is justified; 'if that coercion is justified, then the citizens of that state are correspondingly justified in forming expectations based on the legitimate actions of the state'. 50 But this chain of logic does not justify the agents assuming that the law will stay the same. On the contrary: if a key function of the state is to make 45 I develop a detailed theory of legal transitions in Green, 'Who should get what when governments change the rules?', chs 6-9. 46  and revise laws, then surely citizens' expectations about the law ought to include recognition of the risk that the law may change. Moreover, if the state is legitimate, and therefore (per Meyer and Sanklecha) justified in its coercive law making, surely citizens ought to expect all the more that the state will frequently use that power to change the law. The Normative Authority View is also objectionable in virtue of the results it generates when applied to concrete cases. Specifically, it suffers from what I shall call the Generality Problem: because the legitimacy basis is located in some macro-structural feature of the state that makes the law (in this instance, the legitimacy of the state authority), the verdicts it yields about the legitimacy of expectations are verdicts that apply generally to all agents adversely affected by a given legal change who share the expectation of legal stability. Consequently, the theory is insensitive to normatively relevant features of heterogeneous agents and practices that give us good reasons to treat different agents differently, even when they are affected by the same legal change.
To illustrate this problem, consider Meyer and Sanklecha's example of greenhouse gas restrictions. To fix ideas, let's say these take the form of an economy-wide 'carbon tax'. 51 Now consider an individual representative of a heavily industrialized country-let's call her Anna-who expects to be able to continue emitting greenhouse gases at previous (untaxed) levels; who expects, in other words, that the law pertaining to greenhouse gas emissions will remain the same. 52 If, on the Normative Authority View, the state that enacts the carbon tax law is deemed to be legitimate (that is, if one's preferred conception of legitimate authority, when applied to the relevant facts, yielded such a verdict) then Anna's expectation would be deemed to be legitimate. Moreover, and this is the key point, so would the equivalent expectations of all other agents-be they fossil fuel corporations, stock investors, SUV pleasure-drivers, coalminers, energy-poor consumers, and so on. Conversely, if, on the Normative Authority View, the state enacting the law is deemed to be not legitimate then Anna's expectation would likewise be deemed to be not legitimate. Again, so would the equivalent expectations of all other agents.
This feature of the Normative Authority View virtually guarantees that its application will yield a large number of counterintuitive results, in terms of both the classification of expectations and the entitlement of agents to transitional remedies such as compensation or grandfathering. For example, I suspect that many readers will share my intuition that, at the very least, any expectation on the part of fossil fuel corporations and their shareholders that the laws governing greenhouse gas emissions will stay the same cannot be legitimate. Yet, on the Normative Authority View, such expectations could, in principle, be classified as legitimate, entitling these entities to transitional remedies. While this result would depend on how legitimate state authority is specified and applied to the relevant governmental authority in whose jurisdiction the expectation was formed, the fact that this outcome is a live possibility on the Normative Authority View must count against it.
But now let's assume that the bar for legitimate state authority were set sufficiently high that the relevant jurisdiction enacting the carbon tax was classified as not legitimate. The effect would be that no adversely affected agents who in fact expected the law to stay the same would be classified as having legitimate expectations, and therefore no agent would have a normative entitlement to a transitional remedy. This would generate the intuitively correct result (no compensation/assistance) in the case of the fossil fuel companies and their shareholders. But what about the workers in coalmines and coal-fired power stations who lose their jobs and whose skills decline in value? Or low-to middleincome consumers who own poorly insulated homes, whose electricity bills rise under a carbon tax? It seems intuitively plausible that at least some of these agents might have at least some kind of normative claim on the state for transitional assistance (though I do not think their claim could be based on LE; I suggest an alternative possibility towards the end of this subsection). In sum, however the notion of legitimate state authority is filled in, the Normative Authority View will inevitably generate a large number of either false moral positives or false moral negatives when applied to real-world legal transitions.
Another set of candidates for the legitimacy basis that has received considerable discussion in the recent LE literature refers to the 'justice' of the expectation. This can be cashed out in various ways, and I cannot consider all of the possibilities in detail here. It will suffice for present purposes to consider Meyer and Sanklecha's 'Simple Justice' and 'Complex Justice' views and the views of Matravers and Moore. On the Simple Justice View, 53 the justness, and hence legitimacy, of an expectation depends on the objective justice of the law to which it pertains: effectively, an expectation that an unjust law will continue will not be legitimate, but insofar as the legal status quo is just, then one can legitimately expect it to continue. 54 The upshot is that transitional claims fall to be resolved by an appeal to the correct theory of justice (or rather, the theorist's preferred theory of justice).
One problem is that a justice-based conception of LE would yield indeterminate conclusions in many of the cases that commonly arise in the day-to-day business of changing the law, from zoning ordinance changes to tax reforms. In such cases, the justice of the legal change will be far from obvious. Indeed, plausible views about the justice of the laws in question are likely to be heterogeneous, nuanced, and tentative. It is precisely these cases in which we would need a distinctive legitimacy basis if a theory of LE were to do useful normative work. It is true that using a justice-based legitimacy basis for LE would tend to generate intuitively plausible results in cases where obvious injustice is at stake-slavery abolition is perhaps a good example. But in these cases the value of justice would do all of the important normative work for us; appealing to LE would be redundant.
One possible response to this problem is to limit the application of a justicebased approach to LE to the most 'obvious' or egregious kinds of injustices. This is the move made by Matravers and Moore, who incorporate something like the Simple Justice View into the first stage of their two-stage hybrid conceptions of LE. Matravers argues that laws involving violations of fundamental democratic machinery, basic rights and liberties, and a 'social minimum' of socio-economic entitlements (he draws on Rawls to fill these out) can never generate legitimate expectations that such laws will continue. 55 According to Moore's first stage, expectations can never be legitimate if they are contrary to 'objective justice', by which she means 'rules or policies or practices that are egregiously unjust, that violate basic human rights, or some kind of moral minimum'. 56 However, even if we could agree on a general theory of basic justice and its application in particular cases, there remains room for doubt about whether all losers from 'obviously just' legal change should be denied transitional assistance. For one thing, the respective conception of justice at work in each author's account is a rights-based one that ignores the duty-based dimension of justice. This leaves them vulnerable to the Generality Problem: even in cases where laws obviously violate basic rights, the agents who would be made worse off by the just reform of those unjust laws will often have widely varying degrees of moral responsibility for the instantiation and maintenance of the unjust laws, and this seems relevant to the allocation of transitional entitlements.
Climate change is a good example here. Insofar as we consider the effects of climate change on people's basic rights (for example, to food, water, and shelter), greenhouse gas emissions seem like a clear-cut case of basic rights violation. Measures to mitigate climate change would presumably, then, be classified by Matravers and Moore as obviously just reforms, with the result that agents' expectations of unrestricted entitlements to emit are not legitimate. Yet reforming the laws pertaining to the emission of greenhouse gases affects a wide variety of agents, from Exxon Mobil and its shareholders to the workers on its oil rigs, as we have seen. Accordingly, determining transitional entitlements by reference to an LE conception that rests on a general determination of the justice of the reform will imply a false equivalence among those on the wrong side of history.
Consideration of various problems with the Simple Justice View leads Meyer and Sanklecha to propose the Complex Justice View, which is more sensitive to interpersonal differences among agents affected by the same legal change. 57 The authors invoke a variant of Rawlsian 'pure procedural justice' 58 to determine an 55  agent's range of just expectations. Specifically, they propose three 'consistency requirements' to delimit the applicable range of just expectations that an agent may hold. 59 An agent's expectation is just if it falls within the range of expectations that: (1) the agent could form through impartial reasoning; (2) respects relevant substantive considerations of justice that are basic and uncontroversial; and (3) respects relevant substantive considerations of justice that the agent herself has endorsed. 60 This is an interesting proposal that merits a more wide-ranging analysis than I can provide here. But for present purposes the core problem with this type of approach is that it makes the legitimacy of expectations too agent-specific. By determining the legitimacy of an expectation by reference to other features of the expecting-agent's mental state, the Complex Justice View fails to provide the other agent-here, the state-with a reason for acting consistently with the expecting-agent's expectation. Why should other agents (the state and, derivatively, whichever individuals ultimately bear the costs of compensating, or providing another transitional remedy to, the expecting-agent) bear the costs of satisfying expectations that meet these 'consistency requirements'? 61 Finally, consider the second stages of Matravers' and Moore's hybrid conceptions of LE. 62 For Matravers, features that are potentially relevant to the legitimacy of an expectation include the position of the agent within the social structure, and whether the agent ought to have foreseen that their conduct (that is, that which is threatened by the legal change) was morally wrong and likely to be precluded by a future legal change. 63 For Moore, determining the legitimacy of expectations (at the second stage) is a matter of determining the fairness of the expectation. 64 Moore does not give us a thoroughgoing account of fairness, but in the course of discussing some example cases she identifies numerous desiderata: (1) whether the agent developed a reliance interest based on their expectation that a law would continue; (2) whether the agent ought to have foreseen the relevant change to a law; (3) the suddenness of the change; (4) whether the agent suffered a 'serious' loss or disadvantage as a result of the change; and (5) countervailing distributional considerations, such as the agent's pre-existing wealth relative to others. 65 What interests me about these proposals is that they involve the kind of considerations (for example, reasonable foreseeability) that one would expect to 59 Meyer and Sanklecha, 'How legitimate expectations matter', pp. 385-6. 60 Ibid. 61 Perhaps consistency requirement (2) could rescue this legitimacy basis from my criticism that it is too agent-specific. However, if the 'basic justice' requirement were strong enough to dominate the other two considerations then the view would collapse into a Simple Justice-type view, and would thus be vulnerable to my earlier criticisms. 62 These second-stage considerations come into play when the expectations in question do not pertain to matters of basic justice (the first stage, discussed above). 63  find in private law and interpersonal morality-considerations that seem to cry out for a practice-dependent interpretation of LE. And yet, as I argued earlier, it is problematic to think about individuals as engaged in an interpersonal interaction with the state qua legislator, or to think of changes in the law as violating tacit arrangements that laws would stay the same. For example, what practice-independent standards could one appeal to in order to evaluate whether it was, ex ante, 'reasonably foreseeable' that the state would change the law in the way it ultimately did? As Barbara Fried has argued, legislative changes have a complex political aetiology, making it virtually impossible to assess the reasonableness of expectations about the law at any significant temporal remove from their actual occurrence. 66 In short, insofar as Matravers' and Moore's second stages include such interpersonal-morality-like (or private-law-like) considerations, I think they are hitting on the right kind of considerations for a concept of LE, but wrongly applying them to the domain of legal transitions. Insofar as Matravers' and Moore's second stages include structural factors, such as Matravers' appeal to the agent's position within the social structure, or Moore's recognition that the pre-existing distribution of wealth should also affect agents' transitional entitlements, we see further recognition that the private-law/ interpersonal-interaction model of LE does not travel well to the domain of institutions and institutional change. Structural factors such as these are clearly relevant to the determination of who should receive transitional assistance when laws change. But what do they have to do with people's expectations? Packing such structural considerations into an overloaded conception of LE that also contains interpersonal-interactive, private-law-like considerations (not to mention the first, 'obvious injustice', stage of these authors' hybrid theories) seems like a recipe for theoretical incoherence and practical indeterminacy.
In sum, the legitimacy bases that have been proposed in the existing LE literature applicable to legislative transitions are vulnerable to various objections: some because they are too general and macro-structural; some because they are too specifically focused on individual agents; and some because they combine these two problematic approaches. These criticisms illustrate the inappropriateness of using LE to address transitional problems arising from characteristic legislative enactments. Alternative concepts and principles, better suited to the institutional context of legislative transitions, are needed.
At this point, an objector, noting that my critique in this section has been aimed at particular practice-independent conceptions of LE, might press the objection that I have not conclusively demonstrated that no practice-independent conception of LE could fulfil this role at the core of an institutionally oriented theory of legal transitions. Could we not have a practice-dependent model of LE for interpersonal cases and a practice-independent model for institutional cases? 67 66 Fried, 'Ex ante/ex post', pp. 141-4. 67 I thank an anonymous reviewer for pressing me to respond to this set of challenges.
In response, if the concept of LE were destined to lead this double life, then two radically different models of LE would coexist, governing different domains of activity. We should be concerned about the potential, inherent in such a proposal, for conceptual confusion-especially when it comes to LE, given how much confusion the concept already attracts. While perhaps not a conclusive consideration, it seems to me that this kind of proliferation of rival conceptions of concepts should be avoided in normative analytical political philosophy, all else equal. 68 Perhaps all else would not be equal if we otherwise lacked the concepts needed to formulate good principles of legal transitions-principles that, at the very least, consistently yield intuitively plausible results in transition cases. If this were the case, then having different conceptions of LE for the different domains might be a price worth paying for having a concept (and associated principle) that could do the important normative work of sorting out who should get what, in terms of transitional assistance, when laws change.
However, it is not the case that we lack other concepts (and associated principles) that can fulfil this role. For example, I have argued elsewhere that the values of wellbeing and responsibility are independently relevant to the normative analysis of legal transitions. 69 First, the ex post consequences of a legal change on the affected agents' wellbeing should be taken into account. The second factor to be taken into account is a normative judgement about whether the individual was, ex ante, responsible for managing the risk of the particular kind of legal change (or whether the state was so responsible), which turns on considerations about the agent's obligations of justice and fairness. Judgements about the appropriate form and extent of transitional assistance in a particular case should, I argued, primarily be informed by aggregating the wellbeing and responsibility considerations into an all-things-considered judgement. 70 It is beyond the scope of the present article to specify and defend these concepts and principles, and the process of their aggregation, in any detail. 71 For the purpose of addressing the present objection, it will suffice for me to illustrate that they are, on the face of it, viable candidates, capable of consistently yielding intuitively plausible results in transition cases.
Consider first the greenhouse-gas/carbon-tax cases I have discussed already. The intuitively easiest case is that of fossil fuel companies, which, I take it, have the weakest claim to transitional assistance. Since (assuming normative individualism) corporate agents cannot have wellbeing, the wellbeing impact of any legal transition on a corporation is zero. And in competitive markets where legal risk is treated by corporate agents as just another business risk to be managed, it is fair to hold these agents responsible for managing that risk-enjoying the gains when laws change in ways that suit their business strategy, and suffering the losses otherwise. Both factors therefore point strongly in the direction of leaving these companies' losses to lie where they fall. 72 When it comes to workers in high-carbon industries and to energy consumers, we might think it generally fair, in a market economy, that they too bear responsibility for managing risks to their interests from legal change. For highly skilled employees and relatively wealthy individuals, moreover, the wellbeing shock from a carbon tax would likely be negligible, again suggesting a fully or largely reformative transition policy is appropriate, all things considered. Yet, for certain subgroups of these populations-workers whose skills have been made redundant by the legal change, and energy-poor consumers, for example-the adverse wellbeing shock resulting from the introduction of a carbon-tax is likely to be significant, absent transitional assistance. For these subgroups, the combination of opposing considerations (reformative responsibility considerations and conservative wellbeing considerations) suggests the appropriateness of an intermediate response, somewhere between a fully reformative and a fully conservative position-possibly in the form of partial grandfathering, partial compensation, or 'adaptive policies' (state assistance to facilitate their adaptation to a carbon-constrained economy). 73 Again, these results are intuitively plausible. In none of the carbon tax cases, then, do we seem to need a practice-independent conception of LE to do the normative work required.
What about a seemingly clear-cut case where our intuitions point strongly in favour of a fully conservative response? A good example is the following case given by Margaret Moore concerning 'the requirements that a student has to have in order to be admitted to a university': Suppose society A requires that all students have to take A-levels, and write exams on the basis of this; and society B has an international baccalaureate system. Either system-A levels or IB assessment-is fine: there are advantages and disadvantages with each, but it would seem permissible for different states or different jurisdictions to have different requirements for this sort of thing, and no reason to think that one is uniquely required by justice. Now, let us imagine a case where a state permits both kinds of systems-some schools use A levels and other schools follow an IB curriculum and both are accepted as prerequisites to get into university. Even though both systems are acceptable, we would still think that there is a problem if the state suddenly announced that no IB certificates will be accepted. This rule-change seems to place an unfair burden on all those pupils who worked hard under an IB system, and did so at a time when both kinds of systems and both requirements were acceptable. Moore's normative analysis of this case proceeds as follows: Since either [system of assessment] is consistent with justice, the problem with the sudden change is that it severely disadvantages one group of people who couldn't be expected to foresee this big change in the rules by which they live their life, and have incurred costs-at the minimum, opportunity costs-by orienting their lives to a rule or policy that has now changed. This seems unfair. Even if there are good reasons for the rule to change, it seems that the individuals in this position should be protected [through conservative transition policy]. This seems not a requirement of ideal justice, but a requirement of fairness. 75 Up to the point where my quotation of Moore's analysis ends, my preferred approach to legal transitions-in terms of wellbeing and responsibility considerations-is quite similar to Moore's. First, the students who would be denied university entry on the basis of having followed the old rules would suffer a large, adverse wellbeing shock that would dramatically affect their life prospects (compare Moore's 'severely disadvantages one group'). Second, this is a context in which the individuals are not able to manage the risk of legal change on their own-diversification and hedging strategies are always difficult when it comes to lumpy 'human capital' investments, and third-party insurance is unlikely to be available. Moreover, children's educational pathways, at least pre-university, are often effectively chosen for them, or at least highly constrained, by the state. It seems a clear-cut case in which fairness dictates that ex ante responsibility for managing the risk of this kind of legal change should be borne by the state. Accordingly, both wellbeing and responsibility reasons point towards a conservative response, such as grandfathering the admissibility of IB certificates for affected students.
Where Moore and I differ is that, for Moore, all of these considerations are adduced in the service of an ultimate judgement about legitimate expectations. Her analysis continues with the following italicized text: 'This seems not a requirement of ideal justice, but a requirement of fairness in so far as the person had reasonable expectations that the IB certificate would count as a prerequisite to get into university'. 76 I have already critiqued Moore's hybrid conception of LE in general on the ground that it stacks too many independent normative considerations (of both a structural and an interpersonal kind) into a single concept. Now we can appreciate how this plays out in the context of a particular case: while it is probably true in this case that the students in question expected that their IB certificate would be acceptable for university entrance, and in the circumstances this expectation seems 'reasonable', neither the fact nor reasonableness of such an expectation is necessary for an adequate normative analysis of the case. 77 75 Ibid., pp. 239-40. 76 Ibid., p. 240. 77 Additionally, I discuss problems with the 'expectation' aspect of LE in the next section.
In sum, the analysis of legal transitions in terms of practice-independent LE is not the only game in town. If I am right about practice-dependent models of LE being superior for addressing interpersonal cases, but inapplicable to characteristic legislative transitions, and about the merits of avoiding using a radically different (practice-independent) model of LE for the institutional domain, then the presence of viable alternative concepts and principles for addressing legislative transition cases suggests that these should be explored more fully before embracing a practice-independent model of LE in the domain of legal transitions.

III. ExPECTATIONS AND THE 'ExPECTATION-IDENTIFICATION CONDITION'
Political philosophical work on LE tends to assume that we know or can easily find out what agents' expectations are in any given case. But once we leave the contrived situation of philosophers' hypothetical examples, in which all of the relevant information is transparently described, and enter the real world of moral, political, and legal conflict for which constructs like LE are intended to provide normative guidance, we are confronted with the fact that a real-world decision maker lacks such an epistemically privileged position. The concept of LE deals in expectations, so the relevant decision maker must have access to reliable information about the relevant expectations of agents. Of course, no one can literally access another person's mental state, so expectations need to be inferred. Nonetheless, inferences about an agent's expectation can be drawn from an investigation of the evidence available-the agent's and others' testimony and conduct, the surrounding circumstances, and so on-much as a civil or criminal law trial will involve making inferences of mens rea from admissible evidence. However, carrying out the necessary investigations entails costs, and these costs may render those investigations morally problematic-at the limit, impermissible. It is this concern with the permissibility of identifying expectations that motivates my second condition of application for the concept of LE, viz. that the relevant expectations can permissibly be identified by the relevant decision maker.
In the domain of legislative transitions, it is the state that must ultimately determine which agents are entitled to transitional assistance when laws are changed. On an LE-based theory of legal transitions, then, it is the state that must investigate the expectations and relevant circumstances of all affected agents. This gives rise to two concerns which, though perhaps obvious, have been underappreciated in the LE literature. First, the state would have to expend (potentially immense) economic resources to conduct the necessary investigations. Financing the investigation effort would require raising taxes, raising debt, or cutting existing public expenditures, which, all else equal, would have morally relevant costs in the form of burdens on those agents adversely affected by the revenue-raising mechanism. Call these financier moral costs. Second, there would be morally relevant costs incurred by the agents whose expectation is being investigated. A number of liberal egalitarians have highlighted the moral costs associated with overly intrusive practices engaged in by the state in order to determine citizens' entitlements. A particular concern of these scholars is with state practices that entail the revelation and evaluation of agents' inner livestheir metal states, mental capacities, rationality or reasonableness. 78 Empirical investigations into agents' expectations about the law would do precisely that. Call these agent moral costs. I shall refer more generally to the financier and agent moral costs associated with the state's investigations as the Moral Costs Problem.
In characteristic legislative transitions, the scale of the Moral Costs Problem is particularly great. This is because characteristic legislative enactments affect large numbers of agents with heterogeneous expectations, and apply impersonally. Accordingly, the range of affected agents is not only large but wide open, such that investigations would need to be made into all potentially affected agents. Consider again the example of an economy-wide carbon tax law. A carbon tax affects producer and consumer prices for fuel, electricity, agricultural products, steel, and cement, among other products. If we now contemplate the moral costs of the state having to investigate the actual expectations about the legal status of greenhouse gas emissions among all potentially affected agents, 79 we should get a rough sense of the scale of the Moral Costs Problem. To take another example, imagine the scale of the economic and moral costs involved if the British government were normatively required to determine transitional entitlements and obligations associated with Brexit legislation by reference to all potentially affected agents' (legitimate) expectations.
It is true that not all legal changes are so far-reaching, and that representative structures can assist in ascertaining evidence of agents' expectations. The Moral Costs Problem is thus perhaps better conceived in scalar terms: the more agents that are (potentially) affected by a legal transition and the more heterogeneous their expectations and circumstances, the more morally costly it will be for the state to ascertain the information needed to make decisions about transitional assistance if such decisions are to be based on LE. The moral costs would be more manageable in less significant instances of reform. But it is the cases of legislative change that have the most far-reaching transitional implications in which the need for good principles of legal transitions is most acute, and yet it is precisely in such cases that applying LE principles would be the most morally problematic. 78  Now consider an objection to this line of argument. So far, I have been assuming a model of LE that is concerned with the legitimacy of agents' actual expectations (let us call this an expectation-dependent model of LE). But one might object that this is mistaken. Rather, according to this objection, the compound term 'legitimate expectations' should be understood merely as a kind of shorthand label for a normative entitlement that falls short of a right (call this an expectation-independent model of LE). The latter notion is surprisingly commonplace in both philosophical and legal usage of the term. 80 The general idea behind the expectation-independent model, as applied to a theory of legal transitions, I take it, is this: to say an agent has a LE is to express that the agent is normatively entitled to the benefits of legal stability, or at least entitled to a remedy in response to losses incurred as a result of the state having changed the law, whether or not the agent actually, predictively 'expected' the law to stay the same.
The concept of LE, when understood in this expectation-independent sense, serves an analogous function in theories of legal transitions to the concept of 'hypothetical consent' in theories of state authority. 81 This analogy illuminates some reasons why we should reject expectation-independent conceptions of LE. Recall that what motivates LE theories is that agents' actual expectations facilitate the long-term planning that is thought to be important to their practical agency, autonomy, and wellbeing. What is normatively compelling about protecting such expectations against changes in the laws is the thought that frustrating such expectations violates one or more of these interests of the agent. While those expectations need to be legitimate in order to be normatively protected (and this is the task of the legitimacy basis), it is the distinctive kind of expectation-linked harm to the agent that motivates the concern to protect expectations in the first place. 82 If experiencing such harms is no longer a necessary condition of having a LE, then it is not clear what motivates the concept or why the language of expectations is invoked at all. The motivation must stem from something else, and we are owed an account of what that 'something else' is. The worry, then, is that there is something opaque about an expectation-independent model of LE: it leverages the rhetorical power of frustrated expectations, in just the way that hypothetical models of consent are draped in the rhetorical power of 'consent', but on closer inspection it in fact turns out to derive its normative force from some other, mysterious source.
Notably, similar concerns have been raised by legal scholars concerning the use of an expectation-independent model for the purpose of applying the administrative law doctrine of LE that is found in various common-law jurisdictions. Scholars have registered concern with the very real possibility, entailed by expectation-independence, that one could be deemed to have a legitimate expectation about X despite having no actual expectation about, or even knowledge of, X. As one leading administrative law scholar puts it: 'If the individual did not expect anything, then there is nothing that the doctrine can protect'. 83 Similarly, Justice McHugh of the High Court of Australia (as he then was), wrote: 'If the doctrine of legitimate expectation were now extended to matters about which the person affected has no knowledge, the term "expectation" would be a fiction'. 84 An expectation-dependent model is therefore preferable. But, for the reasons discussed above, the Moral Costs Problem is sufficiently serious to sink the prospects of applying it to resolve normative questions arising from characteristic legislative transitions; such applications would be impermissible. If I am right about this, then LE principles must be limited to those domains in which the Moral Costs Problem can be acceptably managed.
In the domains of interpersonal morality and private law, the characteristically small-n, closed nature of private interactions ensures that the costs of information gathering in these subdomains would be manageable and, in any case, borne by the parties in the event of a moral or legal dispute among them; the legislature and the executive are simply not involved. As such, these investigations would entail minimal moral costs. 85 Indeed, it is in the nature of interpersonal interactions and private law that the resolution of normative disputes requires fine-grained analysis of facts and circumstances, as we saw in Section IIA, since the analysis must focus on the specific ways in which the relevant agents have chosen to shape their normative environment inter se. We should therefore embrace a model of LE that is equipped for the epistemic demands of such a task, as the expectationdependent model is. My point is that it is precisely such equipment that renders that model inappropriate to characteristic legislative transitions.

IV. CONCLUSION
The recent discussion of legitimate expectations in legal transitions heralds a welcome return to two important topics in normative philosophy. But I have argued that these two topics should be separated, at least as far as characteristic legislative transitions are concerned. The best model of LE, I argued, is one that is expectation-dependent and practice-dependent. The focus on specific expectations and practices makes the concept uniquely useful as a means of dealing with interaction-based problems of interpersonal morality and private law involving the frustration of expectations associated with tacit arrangements. However, on this model of LE there are inherent limits to the concept's applicability. While it may be possible to scale the concept to certain analogous forms of interaction-based public functions, 86 it cannot be scaled to the domain of characteristic legislative enactments. I have argued, moreover, that attempts to develop conceptions of LE that locate the legitimacy basis in features other than social practices (practice-independent conceptions) and that avoid the onerous informational requirements associated with LE (expectation-independent conceptions) are poorly theoretically motivated and would have counterintuitive implications when applied to concrete instances of legislative transitions.
It follows that the project of building a good theoretical solution to the problem of legal transitions must look to other normative values and principles, which are appropriate to the political-institutional domain that the problem occupies. I suggested some viable candidates in Section IIB. These could usefully be explored in future work on transitional issues arising from characteristic legislative transitions. But if my analysis is correct, one thing is clear: this work should proceed without legitimate expectations.