The moral permissibility of banishment

This essay defends the moral permissibility, as a form of punishment, of banishment, namely the exclusion by a state of a citizen from its territory. I begin by outlining the prima facie case for banishment, consider for whom it may be appropriate, and acknowledge constraints on its permissibility. I then defend banishment against the main objections in the literature to banishment or the related measure of denationalization (stripping citizens of their citizenship): impermissible permanency; excessive severity; ineffectiveness; unfairness to those who are punished and the creation of two classes of citizens; unfairness among states; and that banishment without denationalization is incompatible with the nature of citizenship. I adopt a ‘cantilever strategy’: if incarceration is permissible notwithstanding a certain objection, so is banishment. In concluding, I sympathetically discuss the view that, despite the moral permissibility of banishment, the power to banish should not be instituted because of the risk of abuse.

The last two decades have seen a limited revival by democratic states of the practice of denationalization; that is, depriving citizens of citizenship. 1 The prevailing view among legal and political theorists who have written on the matter is that denationalization is never, or almost never, morally permissible. 2 In this essay, I defend the moral permissibility of banishment as a form of punishment for crime. By banishment, I mean the exclusion by a state of a citizen from all of its territory. (It would also be interesting, but is not attempted in this essay, to consider the ethics of the use of partial banishment -the exclusion by a state of a citizen from a proper part of its territory -as a punishment. 3 ) Banishment may be applied for a term or permanently. It may be applied to a citizen outside the country, by preventing his return, or to a citizen inside the country, by deporting him and preventing his return. Banishment is closely related to denationalization, and in most recent cases of denationalization, the motive of governments for denationalizing citizens was to be able to banish them. But banishment need not be applied in conjunction with denationalization; and I defend only banishment, not denationalization.
In section I, I outline the prima facie case for banishment, consider for whom banishment may be appropriate, and acknowledge constraints on the permissibility of banishment. In section II, I discuss my 'cantilever strategy': if imprisonment is permissible notwithstanding a certain objection, so is banishment. I then defend banishment against the main objections levelled in the literature against denationalization or banishment: that it is permanent (section III);  ', in Chris W. Surprenant (ed.), Rethinking punishment in the era of mass incarceration (Abingdon: Routledge, 2018), pp. 277-293. I do not discuss Honohan's republican objection because I am sceptical of her version of the non-domination requirement, but doubt that it would be worthwhile to discuss the issue with reference to banishment specifically. 3 McGinnis, 'Exile as an alternative to incarceration' discusses some polities that have practised partial banishment in recent history. that it is too severe (section IV); that it is ineffective (section V); that it is unfair and creates two classes of citizens (section VI); and that it creates unfairness among states (section VII). In section VIII, I defend the permissibility of banishing citizens without denationalizing them against the objection that the right to be in the state's territory is a necessary component of citizenship of a state. In concluding (section IX), I discuss the view that, despite the moral permissibility of banishment, it is impermissible to institute the power to banish -by which I mean to legislate to grant the government this powerbecause of the risk of abuse.

I. THE CASE FOR BANISHMENT
There are several reasons in favour of using banishment as a punishment.
Firstly, banishment is likely to be less expensive, and less demanding on facilities and human resources, than other forms of punishment and control like imprisonment or around-the-clock monitoring. The financial cost of a practice compared with its alternatives is a relevant consideration in morally assessing the practice, a consideration that has been largely neglected in the normative discussion of denationalization and banishment. It is also an advantage of a proposal that it might alleviate, even if only marginally, prison overcrowding, a significant problem in many jurisdictions. 4 Secondly, banishment by a state incapacitates the banished persons from crime that requires physical presence in the state's territory.
Thirdly, alternative punishments may be inexpedient. The imprisonment of a former dictator may encourage other dictators to fight to the bitter end rather than hand over power in the face of a democratic uprising; and in certain cases, imprisoning someone may strain relations with other countries (for example, the imprisonment by the United States of Jonathan Pollard, an American citizen who spied for Israel, strained US-Israel relations).
Fourthly, banishment is likely to be a more humane punishment than imprisonment, since a banished person will likely have substantially more freedom of action and movement, and freedom to pursue a normal life, than an imprisoned person.
Note that my prima facie case for banishment does not appeal to the claim that, by committing certain crimes, people express their renunciation, or intention to renounce, their citizenship. 5 I therefore do not consider objections that target arguments for banishment that rely on these putative expressive acts. 6 In the rest of section I, I clarify under what conditions banishment is, in my view, permissible, and for whom it might be appropriate.

A. Whom to banish?
Banishment is not appropriate for criminals whose safe incapacitation requires imprisonment.
Banishment may be appropriate for criminals who are unlikely to be dangerous in some other state's territory. I have in mind in this essay, as potential subjects of banishment, those whose crimes are political in nature, such as agents of an ancien régime, spies, those guilty of treason or sedition and those guilty of terrorism offences, provided that the danger that they will continue to do serious harm, either in their place of banishment or by illegally returning to the territory of the banishing state, is acceptably small.
What about criminals who are not dangerous, and whose punishment is motivated mainly by deterrence, rehabilitation, communication of censure or other goals, not incapacitation? It is perhaps true that the 'minimum severity' of banishment as an effective deterrent is higher than that of imprisonment; that is, banishment is likely to be effective in deterrence and communication only if the term of banishment is very long and therefore severe (for affluent citizens, banishment for two months might be no more than an extended holiday), whereas even a short term of imprisonment can be an effective deterrent. So, it might be argued, banishment might not be appropriate as a punishment for those who do not deserve severe punishment. But this argument misses the fact that, in the case of many minor crimes, the experience and the reputational effects of a judicial process and conviction are often more effectively deterrent than the punishment imposed. Moreover, banishment seems to be an effective means of communicating the community's censure of the criminal, which may itself be a goal of punishment and may also have deterrent or rehabilitative effects. Finally, even if banishment is less effective than imprisonment, this is to be balanced against the greater financial cost, and demand on prison facilities, of imprisonment. But although I do not rule out the possibility that banishment may be appropriate in some cases of the kind considered in this paragraph, I do not, in the rest of this essay, consider this kind of case, and restrict my attention to political crimes like those mentioned in the previous paragraph.

B. Statelessness and homelessness
A point of almost universal agreement in political theorists' discussion of denationalization is that it is impermissible to make a person stateless. Banishing citizens may leave them without any place where they may legally reside, which, it might be thought, is similarly impermissible.
I accept this constraint. That is, I defend the permissibility of banishment only in cases where the person banished nonetheless has a place where she may legally reside, either because she is a citizen of another country, or because another country has permitted or can be expected to permit her to reside there during her term of banishment; and the banishing state is obliged to re-admit a banished person should this condition no longer be met.

C. Due process
Provisions for denationalization in some countries allow a citizen to be denationalized without due process, for example merely by the decision of a minister. This, it is objected, 7 is unjust.
I agree that banishment should be imposed only after due process. Such due process would standardly involve a judicial process, or at least administrative decisions that are subject to appeal to a judicial process. But the same requirement holds with respect to imprisonment as to any other serious punishment. The objection is against certain current denationalization provisions, not the permissibility of banishment in general.

II. THE CANTILEVER STRATEGY
In what follows, a recurring strategy will be what David Miller and Joseph Carens call a 'cantilever strategy'. 8 I shall argue that, if a certain objection does not impugn the permissibility of imprisonment, neither does it impugn the permissibility of banishment. 9 Four considerations favour this approach.
Firstly: imprisonment is widely accepted as permissible by the public, policy-makers, and liberal political theorists.
Secondly: one theorist's modus ponens is another's modus tollens. For the growing minority of political theorists who believe that imprisonment is impermissible, my cantilever strategy may nonetheless have value in this way: if banishment is impermissible, this cantilever strategy, in contrapositive form, furnishes an argument against the permissibility of imprisonment.
Thirdly: perhaps there is tension between imprisonment and some normative commitments of contemporary liberal political philosophy. But the wide acceptance of imprisonment, I suggest, reflects other normative commitments, albeit ones that are not as often explicitly articulated or are neglected by contemporary liberal political philosophy. Appealing to imprisonment allows me to 9 A reviewer for this journal suggests an interesting line of argument, different from mine: people can be given a choice between imprisonment and banishment; and the permissibility of banishment thus chosen might be defended as follows:(1) If a punishment is chosen by the person punished in preference to a permissible alternative punishment, then the chosen punishment is permissible.(2) Imprisonment is permissible.(3) Banishment that is chosen by the person punished in preference to imprisonment is permissible.I do not pursue this line of argument, for two reasons. Firstly: The conclusion of this argument is only that banishment thus chosen is permissible, whereas my arguments are not restricted to banishment thus chosen. Secondly: (1) is controversial. It might, for one, be thought that a person's choosing x in preference to a permissible alternative y is not sufficient for x to be permissible; it might, for instance, be thought that a person's choosing to be killed in preference to being permissibly imprisoned would not make it permissible for him to be killed. It might, for another, be thought that allowing those who are to be punished to choose their punishment to their circumstances best is inconsistent with punishment being fair. A suitably qualified version of (1) may well be true, but it would take us too far from the main concerns of this essay to discuss the issues that have to be considered to formulate such a version. leverage those submerged commitments without engaging in extended critical discussion of contemporary liberal political philosophy, as I obviously cannot do in this essay.
Fourthly: imprisonment, even if permissible, is, as actually practised, often inhumane, expensive and ineffective. This essay is motivated not by attraction to banishment but by a wish to explore alternatives to imprisonment; I should like to show, therefore, not just that banishment is permissible, but that it is in some respects preferable to imprisonment. III. OBJECTION 1: PERMANENCE Elizabeth F. Cohen argues that denationalization, being a permanent measure, is incompatible with a core assumption of democracies, viz. that all citizens are capable of change and reform. 10 Banishment without denationalization need not, however, be permanent. Banishment can be imposed for a term. Indeed, it may even be open to a denationalized individual to apply for citizenship later. Admittedly, the possibility might in many cases be an empty one: the state would not be able to collect reliable evidence on someone's rehabilitation without removing a main motivation for banishment, its relative inexpensiveness. Still, this may be possible in some cases: for example, a repentant high official of the ancien régime whose life in exile is subject to extensive public scrutiny. Moreover, even permanent or effectively permanent banishment is not impugned by Cohen's objection, where an effectively permanent term of punishment is one such that, when it is imposed, it is predictable that the subject will die before the term is completed.
First: If it is sufficient for being compatible with the assumption that people are capable of reform that they be allowed, at some time before their death, to rebuild their lives, then even permanent banishment is compatible with assuming that those subject to it are capable of reform, because it allows them to rebuild their lives (more quickly, indeed, than does imprisonment), but merely requires that they do so elsewhere. (This point is due to a reviewer for this journal.) Perhaps, however, compatibility with the assumption that people are capable of reform requires that their punishment end at some time before their death -that they at some point in their lifetime cease to have the status of being punished. If so, the response of the previous paragraph is not effective, since although someone who is permanently banished can rebuild her life, her punishment -exclusion from the territory of the banishing state -does not end before her death.
Second: That one is capable of reform implies, I assume, that one is capable of reform during one's lifetime. If imposing a punishment that will last for someone's lifetime were incompatible with the assumption that people are capable of reform (during their lifetimes), imposing a fixed term of punishment of x years would, it seems, be incompatible with the assumption that people are capable of reform within x years. But it is common to impose punishment that does not end before some fixed length of time has elapsed, and indeed to impose imprisonment without the possibility of parole before some fixed length of time has elapsed, which is a punishment that does not allow people to rebuild their lives before the fixed length of time has elapsed. It does not seem in general impermissible to impose a term of punishment, or a non-parole period, of, say, fifteen or twenty years. If these practises are indeed not in general impermissible, then it is not necessary for a punishment to be permissible that it end within fifteen or twenty years, or that it allow people to rebuild their lives within fifteen or twenty years. But why not, if it is necessary for a punishment to be permissible that it end during people's lifetimes, or that it allow people to rebuild their lives during their lifetimes? Either or both of two answers can be offered.
One possible answer is that considerations of capacity for reform do urge against finite fixed terms and non-parole periods as well as against permanent punishment, but finite fixed terms and non-parole periods are permissible because considerations of capacity for reform do not impose absolute constraints, but can be outweighed by other considerations, such as ones of deterrence or retribution. But if such other considerations sometimes outweigh considerations of capacity for reform when finite fixed terms and non-parole periods are at issue, why would they never outweigh considerations of capacity for reform when permanent punishment is at issue? In the absence of a cogent argument for such a difference between the two cases, we do not have reason to think that considerations of capacity for reform make permanent punishment in general impermissible.
Another possible answer is that finite fixed terms of punishment and non-parole periods are permissible because it is consistent with the assumption that people are capable of reform (during their lifetimes) to expect that reform will not occur before a certain length of time has elapsed. If so, the assumption that those being punished are capable of reform does not tell against imposing on someone an effectively permanent term of imprisonment or banishment. For suppose that, in a given case, it is consistent with the assumption that those being punished are capable of reform to impose a fixed term, or a non-parole period, of x years, because it is justified to sentence on the basis of the expectation that reform will not occur before x years. But suppose also that, in view of the age or health of the person being punished, she will predictably die before x years elapse. In such a case, it is consistent with the assumption that those being punished are capable of reform to impose an effectively permanent sentence. Indeed, when an imprisoned person is released, or a relatively short sentence is handed down, on the ground of advanced age or terminal illness, the consideration at work is usually mercy, not respect for people's capacity for reform. If we imposed a sentence such that the criminal is likely to die in prison, we would (we may think) be unmerciful, but not necessarily failing to respect their capacity for reform.
Third: Cohen's assumption is not, it seems to me, a core assumption of democracies. Democracies need not assume that all sociopaths or fanatics are capable of reform. If Cohen's claim appears plausible to some readers, perhaps its plausibility derives from conflation with a weaker claim: that democracy, of the kind that contemporary liberal democratic polities are meant to embody, can be justified only on the assumption of a certain philosophical anthropology, one according to which citizens can revise their ends and thus are of a kind of agent that is capable of change and reform. If it were not the case that most citizens are capable of change and reform, it might be said, democracy of this kind would be pointless. In fact, even these claims are not plausible. There is still a good deal of 'point' to a democracy that satisfies only a minimalist conception of democracy that does not assume that most citizens can revise their ends, but does not satisfy more demanding deliberative conceptions; and it is questionable whether only those more demanding conceptions can be adequate interpretations of the ideology of contemporary liberal democratic polities. But I need not press these points. For even if the justification of a certain kind of democracy requires the assumption that most citizens are capable of change and reform, this does not entail that it requires the assumption that all citizens are so capable. Recognizing that the odd sociopath or fanatic is incapable of reform is compatible with even a deliberative democracy.
IV. OBJECTION 2: SEVERITY Banishment, it is objected, contravenes the right to security of residence. Patti Tamara Lenard writes: 'the right to citizenship is grounded first and foremost in the fundamental interest individuals have in possessing security of residence', an interest that is important because it 'protects the confident expectation that individuals will be able to continue living where they are for the foreseeable future and permits them to make decisions about how their lives will go '. 11 Firstly, this argument cannot impugn banishment without also impugning imprisonment, and is therefore dialectically unavailable to Lenard, who argues that states are morally obliged to use imprisonment rather than banishment. This point is recognized by Elizabeth F. Cohen, who writes, 'forced exile may not actually be any more disruptive or arbitrary than the most widely accepted forms of modern punishment, namely imprisonment. … People who are punished by imprisonment lose fundamental parts of their citizenship, including their right to free movement [and] important civil rights[.] … They are also removed from their entire social context, severing their most intimate ties.' 12 Indeed, insofar as the right to security of residence is grounded on the interest in the non-disruption of one's life plans, imprisonment would usually be more objectionable than banishment, since banished persons can still carry out many activities that that are typically parts of life-plans -for example, employment, starting a household, living with one's family, communicating freely with others, further education, or travel or pursuing interests and hobbies that require travel -that are either unavailable, or available only in attenuated form, to prisoners.
Might taking into account the variable of duration impugn banishment but not imprisonment? It might be thought that, at least in cases of relatively short terms of imprisonment, the disruption to life plans is mitigated by the possibility of returning to one's previous life after release, even if in the interim the imprisoned person, unlike the banished person, can have little semblance of a normal life. Indeed. But the right comparison is not between a long or indefinite term of banishment and a short term of imprisonment. (If some people, initially, think that banishment is more severe than imprisonment, this is, one suspects, because they think that banishment must be permanent. But this is not so, as I noted in the previous section.) Alternatively, it might be thought that, in comparing the severity of banishment and imprisonment, we should hold all other variables, including duration, constant; and imprisonment for a certain term (or indefinitely) is more severe than banishment for the same term (or indefinitely). But this is not the right comparison either. The right comparison, rather, is between the term of banishment and the term of imprisonment that are appropriate alternative punishments in a given case. It is plausible that, typically, the appropriate alternative to a certain term of imprisonment would be a longer term of banishment, precisely because a term of banishment would be less severe than the same term of imprisonment. If the appropriate alternative to a certain term of imprisonment is a term of banishment calibrated to be equally severe, the consideration of severity cannot be used to impugn only one of the two punishments. Severity would impugn banishment but not imprisonment only if the term of imprisonment in a given case is such that any term of banishment that is an appropriate alternative to it would be more severe. This may be possible if, as suggested above, banishment is only effective if a quite long term is imposed, such that there are some terms of imprisonment that are less severe than any effective term of banishment. However, this concern would not impugn banishment in general, and would rarely arise for the cases for which banishment is being considered, which involve political crimes for which, typically, long terms of imprisonment would be imposed in lieu of banishment.
It might also be replied that imprisonment is likely to be in some respects a less severe disruption of life-plans than banishment. For example, an imprisoned citizen is likely to be able to live among people sharing the culture in which she hitherto lived, which might not be the case for a banished citizen. 13 An imprisoned citizen might also more easily be able to receive visits from family than a banished citizen. 14 On the other hand, it might be possible for the banished citizen continue to live with her immediate family if they migrate with her, as is not possible for the imprisoned citizen. Moreover, crucially, to sustain the argument that imprisonment is permissible but banishment impermissible, it is necessary 15 to show, not just that imprisonment is in some respects less severe than banishment, but that it is all-things-considered less severe than banishment. This latter claim is implausible. 16 Secondly, that citizens, in general, have a right to security of residence does not imply that those convicted of certain crimes also have such a right and this right of theirs cannot be justifiedly infringed. Plausibly, by committing certain crimes, either a citizen forfeits some important rights or makes these right liable to justified infringement. Insofar as one believes both that citizens, in general, have a right to freedom of movement, and that imprisonment is permissible, one must suppose that prisoners have either forfeited, or made liable to justified infringement, their right to freedom of movement. Are some rights such that they are never forfeited or (except perhaps temporarily, in emergencies) liable to justified infringement? Perhaps. But to defend imprisonment while impugning banishment, it would be necessary to claim that the right to freedom of movement, but not the right to security of residence, can be forfeited or made liable to justified infringement through committing certain crimes. That there is such a difference in alienability between the right to security of residence and the right to freedom of movement has not been shown and is prima facie implausible. V. OBJECTION 3: INEFFECTIVENESS 'In democratic theory,' Lenard argues, 'citizens are considered entitled to justifications for the policies their leaders intend to pursue, in particular where these laws are likely to have a coercive impact on them.' 'In order to justify a policy that threatens some citizens with severe harm, there must be reason to believe that (a) existing policies are inadequate to pursue the relevant objectives and (b) that the chosen policy is the least harmful way in which the relevant objectives can be met. … The burden on policymakers is especially high in cases where the rights of some individuals are at risk of being sacrificed, as in the case of revocation.' Thus, for introducing powers of revoking citizenship to be justified, an explanation must be offered to citizens 'of the significant difference the right to revoke will make in fighting terror'. Yet, 'it has not been made clear that the usual set of judicially applied sanctions (for example, imprisonment, rehabilitation, parole, or monitoring) is insufficient to punish and deter the bad actors targeted by revocation policies. If the state believes that certain crimes are underpunished or underdeterred at present, it can increase the associated prison terms, monitoring conditions, or parole durations. New powers, like the power to revoke, ought to be adopted only when they are more efficient at achieving goals or can do so in more cost-effective ways than the status quo', for which there is no evidence. 17 Underlying Lenard's demand that the moral costs of denationalization be justified by appeal to its effectiveness is the assumption that denationalization is more coercive, harmful or right-infringing than imprisonment. Consider the sentence: 'In order to justify a policy that threatens some citizens with severe harm, there must be reason to believe that (a) existing policies are inadequate to pursue the relevant objectives and (b) that the chosen policy is the least harmful way in which the relevant objectives can be met.' By 'relevant objectives', I take it, Lenard means the objectives (such as deterrence) whose fulfilment by the policy is to justify the harm threatened by the policy, and so the objective of minimizing the harm threatened by the policy is not itself a 'relevant objective'. Lenard's requirement (a) is plausible when the case that one has in mind is one where existing policies are not severely harmful. But suppose that existing policies are more severely harmful than, or as severely harmful as, the policy whose introduction is contemplated. Suppose, for instance, that people who violate parking regulations have hitherto been put to death, and the policy of fining them instead is contemplated. In such cases, it would not be necessary for justifying the new policy to show that existing policies are ineffective, or even that they are less effective than the new policy, in fulfilling the 'relevant objectives'. Lenard's requirement that the existing policies must be 'inadequate to pursue the relevant objectives' if the new policy is to be justified is plausible only when the new policy is more severe than existing policies. As I argued above, however, the form of banishment which I defend is not, all things considered, more severe than imprisonment.
Still, it might be argued that a policy that is as harmful as existing policies is justified only if it is at least as effective as existing policies; and that even a policy that is somewhat less harmful than existing policies may not be justified if it is much less effective than they are. How effective, then, is banishment compared with imprisonment? To answer this question, it is necessary to specify what their objectives are. 18 Lenard's 'Democracies and the power to revoke citizenship' takes a narrow view of their objectives: the only objective that she discusses is protecting the public through deterrence and incapacitation.
We should take a wider view. Minimizing financial cost is an objective that, while not special to the practice of punishment, is relevant to assessing most policy choices. 19 Banishment is superior to imprisonment in fulfilling this objective. Even if banishment is less effective in fulfilling the objectives that Lenard has in mind, then, it may still be justified overall. 18 Let us stipulate that, for comparative purposes, whenever something is an objective of one of the two policies under comparison, it is to be considered an objective of both policies. 19 Another way of taking a wider view is to allow retribution to be an objective of punishment. Elizabeth F. Cohen takes this view and argues that, insofar as banishment is an especially severe penalty, banishment is effective in exacting retribution ('When democracies denationalize', p. 254). But retributivism is a controversial thesis; and while the argument is forceful as an ad hominem argument against Lenard, who insists that banishment is an especially severe penalty, others might think that banishment is less severe than imprisonment and thus less effective in exacting retribution. Moreover, even on the narrow view, the argument that banishment is impermissible fails. Lenard argues that there is no reason to think that denationalization is an effective deterrent, or that the incapacitation afforded by banishment is necessary. 'One should be wary', she writes, 'of the suggestion that those dual citizens who might otherwise commit acts of terror and violence will be deterred from doing so simply because they are at risk of losing one of their citizenships. In perhaps the only analogous case of as severe a punishment-execution by the state-evidence suggests that it has only a minimal deterrence effect'; and 'there is not yet significant evidence to suggest that [citizens who participate in terrorism abroad] often return with an intention to harm their country of citizenship'. 20 As her mention of the capital punishment debate suggests, the evidence that Lenard claims is lacking is evidence that denationalization is more effective a deterrent than imprisonment, not that denationalization is an effective deterrent at all or that it is as effective a deterrent as imprisonment. But it is not necessary for justifying banishment, as I have discussed above, that banishment be more effective than imprisonment. Although reliable evidence is not available (for either the optimistic or the pessimistic view of the relative effectiveness of banishment or denationalization -Lenard's sentence beginning 'One should be wary' does not cite any reliable evidence either 21 ), insofar as being forbidden from returning to one's country is not a trivial penalty, banishment is likely at least somewhat effective as a deterrent. This point has particularly strong ad hominem force against Lenard, who insists on the severity of banishment.
Moreover, banishment is likely to be effective in incapacitating the person banished from committing crimes that require physical presence in the banishing state's territory; and the lack of 'significant evidence' that citizens who participate in terrorism abroad 'often return with an intention to harm their country of citizenship' does not entail that such incapacitation is never needed. For one thing, 20 Lenard, 'Democracies and the power to revoke citizenship ', pp. 85-86. 21 The footnote at the end of the sentence directs the reader to Betty de Hart and Ashley Terlouw, 'Born here: revocation and the automatic loss of Dutch nationality in case of terrorist activities', in Marjolein van den Brink, Susanne Burri and Jenny Goldschmidt (eds.), Equality and Human Rights: Nothing but Trouble? Liber Amicorum Titia Loenen (Utrecht: Netherlands Institute of Human Rights, Utrecht University, 12), p. 316. But this page, and this essay, are not about deterrence. those who participate in terrorism abroad are not the only people whose incapacitation through banishment may be in the public interest. For another, that most who participate in terrorism abroad will not, even if not banished, return to harm their country of citizenship does not entail that in no case will such a person, if not banished, return to harm their country of citizenship. In cases where it is reasonably thought that there is a non-negligible probability that someone potentially liable to banishment will, if not banished, return to harm their country of citizenship, it is reasonable to think that banishment will contribute to protecting the public.
The consideration of financial cost is also relevant on the narrow view. A less expensive measure, even if it is less effectively deterrent and incapacitating, can contribute more to the point of deterrence and incapacitation than a more effectively deterrent and incapacitating but also more expensive measure. This is because, plausibly, the point of deterrence and incapacitation is simply the protection of the public and they have no independent value (with which claim Lenard does not appear to disagree), and the savings effected by adopting the less expensive measure can be used to enhance the protection of the public in ways that are more efficient than adopting the more expensive measure. And there are presumably ways of enhancing public safety that are more cost-efficient than whatever marginal benefit to public safety is achieved by imprisoning rather than banishing certain criminals.

VI. OBJECTION 4: UNFAIRNESS
When the avoiding statelessness constraint is adopted, denationalization applies only to dual citizens. 22 When the avoiding homelessness constraint is adopted, banishment applies only to citizens who have somewhere else where they may legally reside. Permitting either measure, it is objected, would be unfair in two ways: (a) those who commit the same crime would be punished unequally; and (b) there would be two unequal classes of citizens, those who can and those who cannot be denationalized or banished.
Firstly: although banishment differs from imprisonment, a penalty of banishment might be calibrated to be equally bad for someone as a penalty of imprisonment. 23 It might be replied that the differences between banishment and imprisonment are such that the as bad as relation never holds between a penalty of banishment and one of imprisonment. Even if so, however, one may also be neither better nor worse than the other. To banish some citizens and imprison others for the same crime would, if the punishments are appropriately calibrated, be to punish them differently but not unequally. 24 Secondly: Let us suppose that there is an ineliminable difference in severity between banishment and imprisonment. Even so, as Barry and Ferracioli and Miller have noted, treating people differently need not be unfair if there is a relevant difference between them, and while there is no relevant difference between naturalized and natural-born citizens, there is a relevant difference between single and dual citizens: the costs that would be borne by single and dual citizens in the event of denationalization are different. 25 If a single citizen were deprived of a nationality, she would be stateless; if a dual citizen were deprived of a nationality, she would not be stateless. Similarly, in the case of banishment, there is a relevant difference between citizens who have somewhere else they may legally reside and citizens who do not.
This reply by Barry and Ferracioli is on the right lines, but is too quick. The difference between single and dual citizens that Barry and Ferracioli raise is that denationalization for single citizens differs in severity from denationalization for dual citizens; it is not a difference in desert. The difference in the severity of denationalization for the two groups may make it all-things-considered justified for dual citizens to be subject to denationalization while single citizens are subject only to imprisonment; but there would still be unfairness in denationalizing dual citizens and imprison single citizens of equal desert if, as we are supposing, there is an ineliminable difference in severity between banishment and imprisonment, such that imprisonment for single citizens also differs in severity from denational-23 Cf. Miller, 'Democracy, exile, and revocation', p. 268. 24 Do we face here the spectre of 'separate but equal'? One thing that was wrong with certain purportedly 'separate but equal' historical arrangements is that they were in fact unequal. This is not the case for using banishment as a punishment as well as imprisonment if terms of banishment and imprisonment can be appropriately calibrated. Another thing that was wrong with those arrangements is that the separate treatment was arbitrary insofar as there was no morally acceptable reason for it. This is not the case for using banishment as a punishment as well as imprisonment, as I explain next. 25 Barry and Ferracioli, 'Can withdrawing citizenship be justified?'; Miller, 'Democracy, exile, and revocation'. ization for dual citizens. Compare: two criminals A and B, of equal desert, differ in that A has no dependents, while B has young children who would suffer if B were imprisoned. Imprisonment for B is more severe than imprisonment for A, which is more severe than community service for B. Imprisoning A while sentencing B to community service might be all-things-considered justified, but there would still be unfairness insofar as people of equal desert are punished with unequal severity. If community service for B were equal in severity to community service for A, and community service were an intrinsically appropriate punishment for both, then fairness between A and B would favour sentencing both to community service. Similarly, if imprisonment were equally severe for dual citizens and for single citizens, and imprisonment were an intrinsically appropriate punishment for both, then fairness would favour sentencing both to imprisonment.
But as what I have said suggests, unfairness is not decisive for permissibility, but is only one consideration to be weighed against others, and the extent of the unfairness -the number of people among whom the unfairness holds, and the severity of the unfairness among them -matters in the weighing. If, as I have argued, banishment is not an intrinsically impermissible punishment for certain crimes, then its being less expensive than imprisonment, and other advantages, may outweigh the consideration of unfairness. Consider, firstly, the unfairness labelled '(a)' four paragraphs back. This unfairness affects a very small number of people (those who commit banishment-liable crimes), and is plausibly outweighed by the advantages of banishment over imprisonment.
In the case of unfairness (b), the unfairness holds between the class of all citizens who can be banished and the class of all remaining citizens -not a small number of people. But there are other reasons that (b) does not impugn banishment. Firstly, whereas citizens can be divided into dual citizens who can, and single citizens who cannot, be denationalized subject to the constraint that no one be rendered stateless, no division into a class of citizens who can, and a class of those who cannot, be banished subject to the constraint that no one be left without somewhere he may legally reside obtains. For in principle, any citizen can be banished even subject to this constraint. Even a single citizen can be banished -all that is needed is the consent of a receiving state. At most, there is a division into single citizens whose banishment requires, and dual citizens whose banishment does not require, the consent of a foreign state.
Secondly and more importantly, let us grant arguendo that adopting a policy of banishing people who are convicted of certain crimes (call these people, for short, Convicts), subject to the constraint of not leaving people without somewhere they may legally reside, does generate a division into two classes of citizens -'Duals', who have somewhere else to live and so can be banished, and 'Singles', who do not have somewhere else to live and so cannot be banished -and that this division is unfair to Duals as a class. (The following justification also works, mutatis mutandis, if you think it is the Singles as a class to whom the difference in status would be unfair.) I take it that, if a measure is unfair to a group, but can be justified to the group on the basis of reasons that they rightly accept, then the unfairness does not preclude the measure from being allthings-considered justified. Adopting the policy of banishing Convicts who are Duals but not Convicts who are Singles, I contend, can be justified to Duals as a class, in the following way. We can say to them: To protect public safety, it is necessary to either imprison or banish Convicts. It is impermissible to banish Convicts who are Singles, so they must be imprisoned. Should we imprison or banish Convicts who are Duals? If we adopt the policy of banishing Convicts who are Duals, we will generate a division of all citizens into Singles and Duals that is, ex hypothesi, unfair to you, the class of Duals. But banishment is less expensive than imprisonment. You, the class of Duals, will benefit overall from adopting the policy, since the public economies outweigh the minute difference in status introduced and cannot be achieved without introducing the difference in status. The difference of status generated by adopting the policy can, thus, be justified to those to whom it is unfair on the basis that the policy is the permissible policy that is most to their advantage.
Note that we would not be able to justify in this way the unfairness between Convicts who are Singles and Convicts who are Duals, because the latter do not (we hypothesize) benefit overall. But this justification works because unfairness (b) holds, not between two sub-groups of a certain category of criminals, but between those with somewhere else to live as a class and those with nowhere else to live as a class. Thus, the people to whom justification needs to be made are those with somewhere else to live as a class, and as a class those with somewhere else to live do benefit overall from instituting banishment.
VII. OBJECTION 5: INTER-STATE RELATIONS Banishment or denationalization, Miller objects, 'potentially creates unfairness and/or a collective action problem between states'. 26 For (a) some states may use banishment to unfairly burden other states, and (b) if banishment becomes prevalent, states may start a race to rid themselves of undesirable citizens before being pre-empted by another state, which would have risks for due process.
But the fact that an action of a certain kind is such that, in some circumstances, carrying it out creates unfairness does not entail that the action is always impermissible. Indeed, the fact has no bearing on whether it is permissible to carry out such an action in circumstances where it does not create unfairness. And even in circumstances where it would create unfairness, the unfairness need not be morally decisive, but is only one moral consideration among several. In which circumstances would unfairness be created, and in which would it not be created? Miller's claim that it is in general fair to return someone to the state that had a responsibility to appropriately educate and socialize him and failed in doing so is plausible. Barry and Ferracioli's claim that it is fair to banish someone to the state that was complicit in his wrongdoing is also plausible. Likewise plausible is the claim that considerations of fairness discourage the passing of (uncompensated) burdens by wealthier states onto poorer ones. While it is difficult to give a general answer to the question, especially in the absence of a straightforward general theory of international ethics, it will in many cases be possible to reasonably believe that an act of banishment will create no unfairness.
Moreover, both the unfairness canvassed in (a), and the dangerous race to banish canvassed in (b), can be avoided through international agreements that establish fair procedures for how matters of banishment where the interests of two or more states conflict are to be handled. Such procedures may involve the requirement to seek a state's consent before banishing a person to that state, or compensation.
Finally, the interests of states need not conflict in cases of banishment: some people are undesirable only in one state. This goes especially for spies, some political criminals and certain kinds of ideological terrorists. We should recognize that we still live in a world of deep ideological differences among states, as well as differences of material interest. The United States gave refuge to Syngman Rhee, former dictator of the Republic of Korea, after the April Revolution of 1960 (to take one example among many); and may well be willing today to give refuge to an Iranian citizen who committed crimes motivated by opposition to the ideology of the government of Iran. Had the United States decided to banish rather than imprison Jonathan Pollard, Israel would no doubt have been willing to give him refuge.
VIII. OBJECTION 6: CITIZENSHIP I defend the moral permissibility of banishing citizens without denationalizing them. This course of action, it may be objected, contravenes certain international covenants. Article 13 of the Universal Declaration of Human Rights (1948) states: 'Everyone has the right … to return to his country.' Article 12 of the International Covenant on Civil and Political Rights (1966) states: 'No one shall be arbitrarily deprived of the right to enter his own country.' This course of action, it may further be objected, is incompatible with the nature of citizenship. 27 The right to enter the territory of a country cannot be alienated without losing the whole bundle of rights and duties that comprise the right of citizenship of the country. In the rest of this section, by 'banishment' I mean banishment without denationalization.
Consider first the appeal to international covenants. My concern is the moral permissibility of banishment. Positive law and quasilegal documents are a guide to what is widely accepted, but not 27 Lenard writes: 'It would be unthinkable for a state to permit an exiled individual-a person who was no longer permitted to reside in that state-to vote or to travel with its protection' ('Democratic citizenship and denationalization', p. 102).
decisive as to what is morally permissible. 28 We need to ask: what is the moral basis of the rights ascribed by international law? It is likely that the considerations discussed in the remainder of this section are believed to form part of the moral basis, and that the risk of abuse of the power to banish by states, discussed below, is believed to form another part. Now consider the claim that the right to be in a country's territory cannot be prised apart from the other rights and duties that comprise citizenship of the country. The objection under consideration is not the objection from equal citizenship discussed in the previous section. Rather, the claim is that there would be something wrong with someone having all the rights and duties of citizenship of a state except the right to be in the state's territory, regardless of whether others are in the same situation. Why might this be so? David Miller suggests 29 that the duties and rights comprising citizenship are in part justified by each other; it would be unjustified to hold someone who does not have the full bundle of rights of citizenship under the full bundle of duties of citizenship. Moreover, banishment, it might be argued, cannot be made permissible by simultaneously releasing the banished person from some duties of citizenship, because the justificatory relationship between duties and rights of citizenship may hold diachronically -a right in the present may justify and be justified by duties borne in the past.
A problem with this argument is that it does not account for the permissibility of imprisonment. For also among the rights and duties comprising citizenship are the right to free movement within one's country and the right to place oneself outside one's country, rights of which imprisoned persons are deprived. If imprisonment is permissible, it must be allowed that a person's commission of certain crimes sometimes justifies depriving him of certain rights of citizenship at least insofar as (i) the deprivation is necessary for a certain punishment to be imposed, (ii) this punishment is appropriate for the crime, and (iii) this is the punishment actually imposed on the person. But then depriving someone of the right to be in his country of citi-28 Moreover, the ICCPR formulation does not say that people may never be deprived of the right to enter their own country, but only that they may not be arbitrarily deprived of it. If due process is followed and if there is a weighty cause for deprivation, then deprivation might perhaps not be arbitrary in a way that would render it in contravention of the ICCPR. 29 In private conversation.
zenship can also be justified as long as (iii) banishment is the punishment actually imposed on the person. For (i) depriving someone of this right is necessary for banishing him and, as I have argued, (ii) banishment is sometimes an appropriate punishment. It might be replied that the requirements of punishment can justify depriving someone of the right to internal freedom of movement and the right to leave, but not the right to be in the state's territory. But it is unclear why there should be this difference in alienability. It might be replied that banishment deprives someone of more rights of citizenship than imprisonment does, because losing the right to be in the country entails losing, or losing the ability to make use of, the right to internal free movement; and the right to leave is not a right of citizenship, because non-citizens also have it. On this view, imprisonment removes one right of citizenship, whereas banishment removes two.
But this view is wrong. Firstly, that non-citizens also have a right does not entail that the right is not a right of citizenship. Ex hypothesi, whether a right is a right of citizenship depends on whether it has a place in the web of mutual justification holding among the rights and duties of citizenship. Non-citizens also have the right that the police investigate crimes against them; but this does not entail that, for citizens, this right is not part of the justification of some of their duties of citizenship. Similarly, non-citizens also have the right to leave, but it is plausible that, for citizens, the right to leave is part of the justification of some of their duties of citizenship, and so is a right of citizenship.
Secondly, in mentioning right to internal freedom of movement and the right to leave I was not exhaustively enumerating the rights of citizenship lost by an imprisoned person. Plausibly, imprisoned people are also deprived of, deprived of the ability to make use of, or allowed to keep only in attenuated forms, other rights of citizenship. Among these may be many rights of social citizenship (including the right to a gainful working career and rights to access many public services and amenities), liberty rights of citizens in contemporary liberal societies (such as the right to spend one's money to improve one's living circumstances), and the right to privacy. 30 Some of the rights lost by imprisoned persons are similarly lost by banished persons, but some are not. The balance sheet between imprisonment and banishment in terms of rights of citizenship lost is not clear, but does not seem to support the permissibility of imprisonment but not banishment.

IX. OBJECTION 7: RISK OF ABUSE AND DISCRIMINATION
Insofar as we think that imprisonment is permissible, I conclude, we should also think that banishment (subject to the constraint that no one be left without somewhere he may legally reside) is permissible. But I want to address a final worry. This is that, while instituting the power to banish may be justified in a world where governments act justly, it will have baneful consequences given governments as they are. One such consequence is abuse. For example, some states unjustly use banishment to punish citizens with dissenting political views. Another such consequence is impermissible discrimination. Gibney argues that the practice of denationalization has historically been, and remains, marked by impermissible forms of discrimination, notably against certain ethnic and religious minorities. 31 This worry, while it counts against the moral permissibility of instituting banishment, does not count against the moral permissibility of banishment, the thesis that this essay defends. It is consistent to maintain both that it is sometimes permissible for a government with the power of banishment to banish someone, and that it is impermissible to grant any government this power, because governments are sure to use it in circumstances where its use is impermissible. Nonetheless, it might be wondered what the point of my argument is if banishment, though permissible, should not be instituted.
I am sympathetic to both of the concerns raised above. Insofar as the power to banish is likely to be more unjustly exercised than the powers that would be exercised in its stead, it ought not to be instituted. But do not similar concerns apply to those alternative 30 Perhaps some theorists will say that imprisonment is impermissible unless such rights, and the ability to make use of them, are secured; but that would be to take a view significantly different from the premise of my cantilever strategy, which is that actual practices of imprisonment, in accordance with law, that are widely accepted as permissible are indeed permissible. 31 Gibney, 'Denationalisation and discrimination'. powers? States that abuse banishment are likely also to abuse imprisonment, and states that impermissibly discriminate in imposing banishment are likely also to impermissibly discriminate in imposing imprisonment. If so, insofar as instituting banishment would not worsen injustice compared with the situation when only imprisonment is practised, instituting banishment might yet be permissible, even while the whole system of imprisonment and banishment would be unjust and require remedy if possible. Moreover, even if banishment at present ought not to be instituted, whether it is morally permissible still makes a difference. For if we think that banishment would be a permissible and useful practice which ought to be instituted but for the risk of abuse or discriminatory implementation, we may wish to address those risks to allow banishment to be instituted in future. A final consideration is a principle of prudential conservatism that urges us to be cautious in replacing existing practices or institutions that have worked tolerably well, even if we believe that the envisaged replacements are permissible and effective. This principle (which most political theorists, perhaps, do not take as seriously as they should) counsels caution in the case of introducing banishment, even if banishment is, as it has been argued, morally permissible. ACKNOWLEDGMENTS I am grateful to David Miller for commenting on an earlier version of this paper.

OPEN ACCESS
This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or