II. THE TROLLEY PROBLEM: EXPECTED SURVIVAL AND CAUSAL STRUCTURES

We saw that libertarians committed to the mirroring view must defend the LPP. They may appeal to the following case, imagined by Judith Jarvis Thomson:

This case suggests that negative duties are very stringent. To use Thomson's words, ‘if David may not even choose to cut up one where five will thereby be saved, surely what people who say “Killing is worse than letting die” mean by it must be right!’ (Thomson's emphasis).Footnote ¹² Thomson observes, however, that this conclusion should be qualified to accommodate the following case:

If, as many of us believe, it is morally permissible for Bloggs to turn the trolley, then Trolley seems to undermine the LPP.

Thomson uses the pair of cases ‘Trolley-Transplant’ to formulate what she calls ‘the Trolley Problem’: why is it permissible to save the five in Trolley, but not in Transplant?Footnote ¹⁴ Otsuka observes that libertarians must address the Trolley Problem.Footnote ¹⁵ Along similar lines, I want to suggest that libertarians who insist that institutions must reflect the supremacy of negative moral duties over positive moral duties must show that political settings are relevantly akin to Transplant. Since, as I announced, I will not offer a full-fledged account of Trolley, I will not offer a full-fledged solution to the Trolley Problem either. I will nevertheless indicate the kinds of considerations that bear on solving it. This will suffice, I think, to bring out the difficulty that the Trolley Problem poses to defences of libertarianism based on the mirroring view.

Thomson's account of Trolley is as follows. She argues that there was an earlier time at which it was ‘to the advantage’ of all six to consent to a policy allowing bystanders to minimize deaths by redirecting runaway trolleys. For this reason, it is permissible now for the passer-by to redirect the trolley. According to Thomson, (a) if the existence of a policy increases one's life expectancy, then there is a presumption that the policy is to one's advantage, but (b) there are factors (e.g. an especial aversion to die cut into pieces by surgeons) capable of overturning this presumption.Footnote ¹⁶

Thomson's proposal would be circular unless she offered an independent account of the (b)-considerations against killing. In support of the claim that (b) is satisfied in Transplant, Thomson says that

This suggests that our considered judgements about various ways of dying determine whether (b) is satisfied. For Thomson, ex ante hypothetical consent of all six is neither necessary nor sufficient for the permissibility of saving the five by killing the one. Thus, she writes:

To show why hypothetical consent is not necessary either, she imagines the following case:Footnote ¹⁹

Thomson writes:

For Thomson, then, an appropriate account of Trolley can do without reference to hypothetical consent. She concludes that the passer-by may save the one because there was an earlier time at which it was ‘to the advantage’ of all six that the trolley would be turned at a later time should a Trolley-like situation arise.

To explicate the notion of ‘advantage’ that is at work here, Thomson invites us to consider a variant of Crushed Leg in which David is a Christian Scientist. She points out that it would be impermissible to cut off David's leg if David's religious objections to amputation were morally weightier than the fact that he will live if and only if he has his leg cut off.Footnote ²² Notice, however, that our belief that we must not cut off David's leg if he is a Christian Scientist does not commit us to believing that his religious beliefs are plausible. Rather, we must not do so because certain deeply held convictions (religious beliefs among them) ought to be respected, and this may sometimes require us to let believers die. But then, a circularity worry emerges. If Thomson's claim that it is not to the Christian Scientist's advantage to have his life saved by means of an operation pointed to facts that we can establish independently of the permissibility issue, then her (b)-considerations against killing would not render her account circular. Yet, we have just seen that Thomson's claim presupposes that it is morally impermissible to operate on Christian Scientists. Thomson's notion of ‘being to someone's advantage’ turns out to be moral in character.

Circularity would be less of a worry if Thomson's account allowed us to say that my action A is to X's advantage whenever X would welcome A if extraordinary circumstances that bar consent (e.g. being unconscious) ceased to exist. But Thomson's rejection of hypothetical consent rules out, as we saw, this interpretation of her notion of ‘advantage’.Footnote ²³ Accordingly, when we ask, in the context of Thomson's theory, whether the Christian Scientist's religious beliefs bear on what is to his advantage, we are inquiring about the permissibility of our deferring to such beliefs as a way to determine whether we may save his life by means of an operation. This is different from wondering if the Christian Scientist would have consented to being so saved. The former question is moral in character, whereas the latter is a (counter)factual question whose relevance, as we saw, Thomson denies. Her (b)-considerations refer us, then, to our considered moral judgments. They fail to furnish us with independent criteria of permissibility. It follows that Thomson's theory is circular.Footnote ²⁴

I submit that the charge of circularity can be avoided by interpreting condition (b) as pointing to the causal structure of the situation, rather than to morally relevant objections to sacrificing persons. In her earlier work, Kamm put forward some ideas about causal structure that might be thought to flesh out appropriately Thomson's (b)-considerations against killing.Footnote ²⁵ Although, as we shall see, she later modified her views somewhat, a brief discussion of those earlier thoughts will help us to see what is wrong with an initially plausible strategy that the libertarian may adopt. Kamm argued that a ‘Principle of Impermissible Harm’ (‘PI/PH’) can accommodate our moral intuitions in cases in which we can save some only if we kill others. The PI/PH asserts that

The PI/PH's guiding idea is that the causal chain between the action and the harm must not be shorter than the causal chain between the action and the greater good. Kamm claimed that the PI/PH appropriately reconstructs the rationale behind our intuition that it is wrong to bring about good by harming others.Footnote ²⁷

Kamm points out that the PI/PH accommodates Trolley: the causal chain ‘trolley turned – one killed’ is not shorter than the causal chain ‘trolley turned – five saved’.Footnote ²⁸ It arguably accommodates Transplant as well: the causal chain ‘one cut up – one killed’ is shorter than the causal chain ‘one cut up – one killed – five saved’. Unfortunately, however, the PI/PH has two flaws. First, it cannot accommodate all of our considered judgements. Philippa Foot imagines in another context the following case:Footnote ²⁹

Many of us would find counterintuitive the PI/PH's verdict on this case, namely, that it is permissible to manufacture the gas – by construction, the chain ‘the gas manufactured – one killed’ is not shorter than the chain ‘the gas manufactured – five saved’.Footnote ³¹ Second, it is unclear how we are to measure the lengths of causal chains. In what sense is diverting the trolley in Trolley just one causal link away both from killing the one and from saving the five? Does it make sense to count causal links between diverting a threat and the threatened persons being saved? Indeed, measurement problems creep in even when all the alternatives involve killing others. Imagine a variant of Trolley in which the driver is the only person who can decide which of the two tracks to take. Imagine, further, that the trolley would hit each of the five across an extra layer of cloth. Shall we say that going through one layer of cloth constitutes a causal link, and so that the driver must direct the trolley onto the five?Footnote ³² This question suggests that fine-grained accounts of causal links are morally insignificant in the context of the PI/PH, but the PI/PH provides no clues as to why or to what extent this is so. There might be plausible characterizations of causal links under which the PI/PH can accommodate all imaginable cases, including Deadly Gas. However, it seems reasonable to require the proponent of the PI/PH to defend the theories of causality required by such accounts. He will have to show, on pain of circularity, that the requisite causal links can be individuated without appealing to the moral differences he wants to accommodate.

As I said, more recently Kamm put forward somewhat modified views. She now proposes a ‘Principle of Permissible Harm’ (‘PPH’). Roughly put, this principle lays down as a sufficient condition for the permissibility of killing an innocent that the greater good should cause the lesser harm, the non-fulfilment of this condition being prima facie impermissible.Footnote ³³ Unlike the PI/PH, the PPH yields in Deadly Gas a result that most of us would accept, namely, that it is prima facie impermissible to manufacture the gas. In addition, the PPH avoids questions about the relative lengths of distinct causal chains – it only requires that a causal chain (however its links are individuated) be part of another causal chain.

Kamm correctly observes, however, that we should refine the formulation of the PPH to neutralize some counterexamples. In Trolley, for instance, it is not the case that the greater good causes the lesser harm. Kamm offers the following case to help us see the direction that the requisite reformulation should take:

It seems impermissible for us to save the five here. Kamm writes that

So Kamm refines the formulation of the PPH to accommodate cases in which it is permissible to save five by killing an innocent despite the fact that, contrary to the above rough formulation, the greater good is not a cause of the lesser harm. In such cases, the greater good is, she says, a flip side of the act that saves the five.Footnote ³⁶

The PPH helps us accommodate our moral discriminations between cases for which Thomson's condition (b) (i.e. lack of objections to certain ways of saving the greater group) provides, as we saw, no guidance. Consider the following variant of Deadly Gas, imagined by Kamm:

Kamm observes that the PPH accounts for the moral discrimination that most of us would make between Deadly Gas and Healing Breath – only in the latter case it seems permissible to save the five.Footnote ³⁸ An ex ante maximizing policy such as that allowed by Thomson's condition (a) may well equate the two cases. On the other hand, as we saw, it is unclear how her condition (b) would ground the requisite discrimination. The PPH may be viewed as specifying Thomson's condition (b) to require that the greater good cause the lesser harm. Call this formulation ‘(b*)’. I suggest, then, that rescuing the five is permissible if and only if it instantiates a policy that ex ante increased, through causal routes sanctioned by the PPH, each of the six's life expectancy. Conditions (a) and (b*) are individually necessary and jointly sufficient for it being permissible to save five by killing one.

We can now accommodate all of our considered judgements on situations in which the agent harms some to save others. Trolley fulfils (a) and (b*). Transplant fails to fulfil (b*). A variant of Transplant (call it Transplant*) in which everyone's ex ante probability of survival would increase if surgeons were to proceed in Transplant-like situations also meets (a), but not (b*).Footnote ³⁹ Thomson observes that there was no time at which it would have been to the advantage of all six that the surgeon proceed in Transplant*, because persons would still object to certain ways of dying.Footnote ⁴⁰ As I argued, this reference to legitimate objections to certain ways of dying is circular. On my proposal, saving the five in Transplant* involves (as saving the five in Transplant does) a causal structure that (b*) condemns. Finally, a variant of Trolley in which the positions on the tracks are not determined at random (for instance, because a railway company contractually assigned each of the six workers to each of the two tracks) would satisfy (b*) but not (a). In this variant, it would be impermissible to save the five because there was no earlier time at which all six would have increased their probabilities of survival should the passer-by save the five as a flip side of an action that brings about the death of the one.

The view I am proposing (henceforth, the causal view) can therefore be formulated in this way: when killing and letting die are the only options and other things are equal, the agent may follow a policy that maximizes ex ante expected survival, provided that the death of the lesser number is either the causal consequence of the fact that the greater number was saved or this latter fact is the flip side of the act (e.g. throwing the switch in Trolley) that causes the death of the lesser number.Footnote ⁴¹ As I pointed out, I do not offer this account as a full-fledged solution to the Trolley Problem. Plainly, the natures of the relevant causal relationships and flip sides need elaboration. Nor shall I endeavour to elucidate what exactly counts as expected benefit besides expected survival – something required by formulations intended to apply to alternatives other than killing and letting die. I do think, however, that the causal view brings to light the kinds of considerations that a full-fledged solution to the Trolley Problem should contain, and it does so in a way that will help us see in sections III and IV why the Trolley Problem makes trouble for libertarians who rely on the mirroring view.Footnote ⁴²

III. DOES EXPECTED BENEFIT RULE OUT POLITICAL REDISTRIBUTION?

Let us assume that the causal structures licensed by the causal view obtain, and so focus on expected benefit. A suitably generalized version of the causal view would then require agents to aid others well beyond the scope tolerable by the LPP. This is so because most of us are members of groups of persons whose expected benefits would increase if a non-member followed a redistributive policy. In such cases, the causal view entails that it is permissible to benefit some members by harming other members. A political interpretation of the causal view would then sanction significant redistribution. Let us look more closely into this idea.

Under a political interpretation of expected benefit, the state can be seen as a bystander who may benefit certain persons by harming others if and only if at some earlier time everyone's expected gains would have been higher had such policies been adopted. The state may be viewed as the analogue of the passer-by in Trolley, i.e. as a third party in a position to redistribute burdens and benefits by taxation, public expenditure and various kinds of regulation. Indeed, the analogy can go quite far. For example, the state can ‘redirect’ a recession from the potential unemployed to individuals whose extra tax burden will be more than offset by unemployment benefits, given diminishing marginal utility of money.Footnote ⁴³ Notice that these analogies cannot be easily dismissed on the grounds that states are not really third parties vis-à-vis citizens, but rather agents of the citizenry. Our question is whether a state's redistributive policies can be seen as enforcing antecedent moral duties to aid, over and above other reasons for or against such policies. As I said, the libertarian argument that I explore in this article denies that such policies can be seen in that way.

The upshot of this approach is gloomy for the LPP. Most likely, each of us belongs to many sets of persons (henceforth, mutual aid groups) whose members would have increased their expected gains by consenting at an earlier time to certain redistributive policies. Many of the usual redistributive arrangements for the unemployed, the sick, the aged, etc., could arguably be defended along these lines. Given that the causal view led us to exonerate the passer-by who kills someone in Trolley by throwing a trolley onto him, why not then authorize the state to impose some level of taxation on the rich in order to rescue some people from abject poverty? Such policies increase the expected gains of the members of the relevant mutual aid group, including those who turn out to be taxpayers. The sort of libertarian I am criticizing (i.e. the supporter of free markets on the grounds that we may not harm persons even to aid others) has to show what is special about political redistribution vis-à-vis Trolley. She cannot resort to strategies (such as independent theories of consent-based legitimacy, property rights or general welfare) that may be available to other libertarians.

Have I not overplayed the analogy between Trolley and political redistribution? The libertarian might adduce that, unlike saving the five in Trolley, political redistribution benefits some members of mutual aid groups at the expense of other members. To turn the trolley is to worsen the one's situation without improving the five's situations – it is not to provide the five with anything they lacked before the threat arose. By contrast, redistributive taxation worsens some persons and benefits other persons. But surely, so this libertarian contends, we may not benefit some persons at the expense of others. Libertarians who endorse the mirroring view might think that this manoeuvre blocks the route from Trolley to the redistributive state.

Different interpretations of this objection arise as we consider different baselines against which improvement and worsening occur. When the baseline is the fate that would await each of the six in Trolley in the event that the passer-by failed to intervene, diverting the trolley obviously improves the five's situations and worsens the one's situation. But the same can be said about redistributions within mutual aid groups: by definition, redistribution makes some people better off and others worse off compared to the situation they would face in the absence of redistribution. To distinguish Trolley-like redistribution from political redistribution, the baseline required by the objection must be people's situations until the passer-by/government decides what to do. Against this baseline, turning the trolley does not make the five better off. Rather, it preserves something they already have – i.e. their lives. On the other hand, it makes the one worse off, since it deprives him of something he already has, i.e. his life. By contrast, redistributive policies make some members of a mutual aid group better off against this baseline, for such policies provide those individuals with something they lacked beforehand, e.g. an unemployment subsidy. On the other hand, such policies deprive taxpayers of money they owned, thereby worsening their situations. Libertarians may invoke this difference between saving the five in Trolley-like cases and political redistribution. They may say that negative duties give in only in those situations, like Trolley, in which the agent minimizes evil without benefiting, against the ‘until redistribution’ baseline, anyone.Footnote ⁴⁴ So construed, the objection hopes to resist political extensions of the causal view.

No discernible moral principle, however, supports this version of the improvement/worsening criterion. It leads to the more libertarian conclusion but it does not justify it – it is ad hoc. It does not tell us what is so important about the ‘until redirection/redistribution’ baseline that renders it permissible to kill one person in Trolley while rendering it impermissible to tax a mutual aid group's member. What is more, this version of the improvement/worsening criterion fails to accommodate all of our considered judgements, as shown by the following case:

It seems at least permissible for the passer-by to take the cable out of the swimming pool, thereby ending the five's pains and inflicting a similar pain on John. However, the present version of the improvement/worsening criterion prohibits the passer-by to do so: if she does, she worsens, with regard to the ‘until redirection’ baseline, John's situation and improves the five's.

I conclude, then, that, given the mirroring view, the improvement/worsening criterion cannot be given an interpretation under which saving the five in Trolley and political redistribution within mutual aid groups are morally different.

IV. DO CAUSAL STRUCTURES RULE OUT POLITICAL REDISTRIBUTION?

In section II, I rejected the PI/PH, according to which the causal chain between the action and the lesser harm must not be shorter than the causal chain between the action and the greater good. I assumed, however, that the PI/PH accommodates both Trolley and Transplant. Footnote ⁴⁵ Moreover, the PI/PH, like the causal view, focuses on causal structures. If the libertarian who endorses the mirroring view could show that the PI/PH distinguishes between political redistribution and Trolley, she might then hope that a little tinkering with the PI/PH would rescue it from the objections raised in section II. To see the plausibility of this strategy, consider the fact that the causal chain ‘enactment of a tax law – decrease in taxpayers’ wealth’ is shorter than the causal chain ‘enactment of a tax law – decrease in taxpayers’ wealth – transfer to welfare recipients’. It follows that a political version of the PI/PH prohibits transfer spending financed through taxes.

However, some (actual) redistributive policies do conform to the PI/PH.Footnote ⁴⁶ Redistributive taxation is not the only way of redistributing wealth. Governments sometimes print money for redistributive purposes. When they do so, the causal chain leading to a group's being worsened is longer than the causal chain leading to another group's being benefited. The causal chain ‘printing money – some citizens subsidized – other citizens having their purchasing power lowered’ is longer than (indeed, it contains) the causal chain ‘printing money – some citizens subsidized’. Such measures conform, then, to a political version of the PI/PH. Similarly, governments sometimes subsidize certain groups with money borrowed at the expense of future taxpayers. Moreover, unlike the killing/letting die cases discussed so far, redistribution financed through taxes or through borrowing may confer individual benefits that are higher than the individual burdens they impose. This is arguably so because, given diminishing marginal utility of money, each taxpayer's loss is – within wide ranges of numbers and wealth of taxpayers and beneficiaries – more than offset by each beneficiary's benefit.Footnote ⁴⁷ Moreover, we can imagine that this is true of situations in which, by analogy with Trolley, the number of taxpayers is smaller than the number of beneficiaries. One would expect in these latter situations the PI/PH-based case for redistribution to be at least as strong as the PI/PH-based case for the claim that in Trolley it is permissible to save a greater number of lives by killing.

Consider now the PPH. Recall that the PPH states that it is permissible to save the greater number whenever the death of the lesser number is either the causal consequence of the fact that the greater number was saved or this latter fact is the flip side of the act that causes the death of the lesser number. As we saw, the causal view endorses the PPH as a causal constraint on policies that maximize life expectancy. I now want to argue that a political interpretation of the PPH fails to yield the priority of negative moral duties needed by the mirroring view. Consider again subsidies to the poor financed through taxes, and assume that benefits exceed costs by whatever amount is necessary to make the subsidy policy ex ante beneficial to both taxpayers and beneficiaries. If the subsidies could be seen as the flip side of (the enactment of) redistributive measures, a political version of the PPH would authorize such measures. Viewing the subsidies as flip sides is all the more plausible under legal systems that make it constitutionally mandatory (for example, by virtue of constitutional provisions regarding deficit financing and fairness of the tax system) for the legislature to finance the subsidies through, say, progressive taxes. The fact that a subsidy law makes it constitutionally mandatory for legislators who grant a subsidy to tax the rich mirrors here the fact that turning the trolley in Trolley makes it causally necessary to kill the one.

Perhaps the libertarian can resist this political extension of the PPH by arguing either that taxing the rich is not really a causal consequence of enacting the redistributive measures, or that the subsidies are not really flip sides of such measures. The libertarian might argue, that is, that law, being normative, does not involve causal necessities or flip sides. But a refined formulation of the PPH defuses this objection. To see the direction that such a refinement should take, consider that some laws redefine property rights, and so make some people better off and some others worse off by virtue of a single legislative act. By a ‘redefinition’ of property rights I mean a change in the bundle of rights regarding use and transfer of things. A rent control law, for example, redefines property rights to the extent that it deprives the landlord of the right to charge tenants above the legally stipulated rent.Footnote ⁴⁸ Let us assume that the benefit conferred to tenants is here the greater good. Borrowing from Kamm's terminology, we can then say that this benefit is an ‘aspect’ of a redefinition of property rights that worsens the landlords. To illustrate this notion of ‘aspect’ Kamm uses the following variant of Trolley:

I take it that most people would agree with Kamm's judgement that Lazy Susan is morally on a par with Trolley. She argues that in Lazy Susan it is permissible to save the five because their being saved is an aspect of the turning of the lazy Susan. The general point is that harms that are caused by actions whose flip sides or aspects are greater benefits are permissible.Footnote ⁵⁰

I am not sure how to explicate the notion of ‘aspect’ that Kamm has in mind here, or how to distinguish between aspects and flip sides. Kamm writes:

Apparently, saying that something is an aspect or a flip side is logically entailed by propositions to the effect that a certain action interrupts a threatening causal process. Aspects and flip sides presuppose causal notions, yet they are logically, rather than causally, connected with the relevant actions. Thus, in Lazy Susan the turning of the lazy Susan is, given certain geometrical and physical assumptions, to save the five. Kamm's passage also suggests that aspects are entailed by something done to the people being rescued, rather than to the threat from which they are rescued; flip sides, instead, are entailed by something that we do to a threat, such as diverting a trolley. Both kinds of rescue contrast with those performed by doing something to the one, such as performing blood transfusion from an involuntary donor, thereby killing her, into five patients to save their lives. Such transfusions are clearly impermissible.Footnote ⁵² Moreover, actions having both a greater good and a lesser harm as their aspects seem to be permissible. We can see this by considering a variant of Lazy Susan (Lazy Susan*) in which all six are attached to the lazy Susan in such a way that turning it will save the five if and only if it will place the one within the range of the threat. It seems natural to assume that whatever makes five saved an aspect of the turning of the lazy Susan in the original case should make both five saved and one killed aspects of the turning of the lazy Susan in Lazy Susan*. One killed does not seem to be here ‘more intimately causally related’ to the turning of the lazy Susan than five saved is.

My promised refinement of the causal view should then go along the following lines: agents may maximize ex ante expected survival, provided that the survival of the greater number causes the death of the lesser number, or is an aspect, or a flip side, of the act that has as its causal consequence, or aspect, the death of the lesser number.Footnote ⁵³ This formulation allows us to see legal redefinitions of property rights as analogous to the lazy Susan's movement: when conjoined with contextual information, both events logically entail benefits and losses to different groups. Rent control laws can thus be vindicated on the grounds that both the tenants’ gains (by assumption, the greater good) and the landlords’ losses (the lesser harm) are aspects of the enactment of the laws. In redefining property rights, a rent control law logically entails (legal) benefits and losses to different people, in much the same way as turning the lazy Susan in Lazy Susan* logically entails benefits and losses to different people (i.e. their new positions relative to the threat).

Let us finally consider the thought that political redistribution is objectionable because a wealthier group is treated as a means to benefit a less wealthy group. On this view, we should inquire about the agent's plans, rather than about causal structures. To be sure, an important tradition in moral philosophy embraces the Kantian injunction that persons not be treated merely as means to benefit other persons. Careful assessment of this idea would call for an explication of the notions of ‘treating merely as means’ and ‘benefiting others’ – something that I cannot pursue here. In any event, I do not believe that a prohibition against using persons merely as means, on any tolerably intelligible understanding of it, might help us solve the Trolley Problem, let alone solving it to the satisfaction of the libertarian who adopts the mirroring view. Consider the following case:

In this case, saving the five requires that the trolley jam into the one. There is, then, a natural sense in which the one would be treated as a means to save the five. Nonetheless, the moral standing of diverting the trolley does not seem here different from the moral standing of diverting the trolley in Trolley.Footnote ⁵⁴

Indeed, it is difficult to see how a prohibition against using others could be formulated in a way that withstands well-known objections to the doctrine of the double effect. This doctrine holds that although it is impermissible to bring about a greater good by intentionally using as a means a lesser evil, it is (at least sometimes) permissible to bring about a greater good even if the agent foresees that a lesser evil will take place as an unintended side effect of her behaviour. The doctrine of the double effect leads to the counterintuitive result that diverting the trolley is impermissible in Prevented Return: in order for the passer-by to save the five, it is necessary that the trolley jam into the one. The doctrine also has trouble accommodating Deadly Gas (section II). If we manufacture the gas, then John's death is not a means of saving the five; rather, it is a side effect of the means of saving them, namely, manufacturing the gas. Even so, most of us would find it impermissible to save the five in this way. It seems reasonable to conclude, then, that libertarians who rely on the LPP and the mirroring view cannot appeal to a prohibition to treat others as mere means.Footnote ⁵⁵

V. LAW, MORALITY AND LIBERTARIANISM

I have argued that morality is non-libertarian, in the sense that the legal duties imposed by market institutions do not mirror the scope and strength of our moral duties to aid others. Of course, this is not to say that market institutions lack all-things-considered justification. David Schmidtz, for example, argues that, when combined with plausible assumptions about people's information and incentives, a version of rule-utilitarianism yields both our judgements in Trolley and Transplant, and a justification of free markets.Footnote ⁵⁶ Notice, however, that such arguments rest precisely on the kinds of considerations (e.g. the agent's informational shortcomings) excluded by the other-things-equal clause assumed in discussions of the Trolley Problem. I have tried to debunk an argument for market institutions that depicts legislation as reflecting here and now, as it were, the force of people's negative moral duties vis-à-vis their positive moral duties; we can ascertain such force only by considering situations in which things other than negative and positive moral duties are equal. When other things vary, market institutions may turn out to be justified: mirroring the relative stringency of negative and positive moral duties is not the only thing that matters in political decision-making. The rationale for the case method that I have been using is the need to isolate such stringency from other factors – this is why I assumed throughout that other things are equal.Footnote ⁵⁷ This assumption rules out the kinds of reasons that may play a role in a successful defence of a libertarian regime. As I said in section I, the mirroring relation obtains only when the morally relevant features of a political decision are salient. For example, I assumed that the alternatives faced by the agent, even in such political cases as that of a ‘redirection’ of a recession (section III), were equalized as to additional, less salient, effects, including long-term effects on production and distribution. Such effects may figure in an all-things-considered defence of free markets. Of course, the opposite may be the case as well: even if the LPP provided an adequate account of the relative stringency of our positive and negative moral duties, free markets might lack all-things-considered justification.

One may naturally wonder, then, why some writers, libertarian or otherwise, assume that libertarians are committed to the LPP and the mirroring view.Footnote ⁵⁸ I think that one major, though usually implicit, motivation behind the mirroring view is the belief that legal/institutional proposals are immune to the charge that they overlook the causal consequences of the law. The idea is that law ought to mirror morality, not because law will thereby realize morality's independent aims but as a matter of moral obligation itself. The justificatory structure that emerges from this picture is protected from causal challenge, and in that sense is non-consequentialist. On this view, morality requires that its own precepts be institutionally implemented through the certainty of the mirroring connection rather than through the vagaries of causal relationships.Footnote ⁵⁹ Indeed, this is a view that critics of libertarianism frequently share: they feel pressure towards showing that various forms of state interference with the market ‘reflect’, ‘enforce’, etc. the relative stringency of, say, moral duties to help the needy. Sometimes, those proposals are avowedly confined to ideal theory, leaving it undefined which institutions will best help the needy under realistic assumptions about compliance. Thus, Rawls takes the fact that the welfare state ‘may allow . . . large disparities of income that violate the difference principle’ to warrant more radical departures from a free-market economy, namely, a ‘property-owning democracy’ or a ‘liberal socialist regime’, leaving open the question of which of these two arrangements best realizes the difference principle.Footnote ⁶⁰ Rawls points out that this latter question must ‘be settled by historical conditions and the traditions, institutions, and social forces of each country’.Footnote ⁶¹ By contrast, his rejection of libertarianism (and the welfare state) is not empirically motivated: it is a rejection of the ‘ideal institutional description’ of the libertarian and welfare-state regimes, namely, ‘the description of how [the regime] works when it is working well, that is, in accordance with its public aims and principles of design’.Footnote ⁶² This suggests that Rawls would prefer property-owning democracy rather than libertarian institutions even if, in realistic situations, the former were less generous to the least advantaged. The claim that property-owning democracy or liberal socialism (i.e. legal/institutional regimes) realizes the difference principle (i.e. a principle of political morality) rests on the assumption that the political authorities of such regimes effectively fulfil their role descriptions. Their policies provide the greatest benefit to the least advantaged because that is the task officially assigned to them. Those authorities possess the relevant knowledge and do not take bribes or yield to other perverse incentives created by the institutional framework itself.

Other writers seem to endorse a version of the mirroring view that requires institutional proposals to express or symbolize the relative weights of various moral principles, as something different from effectively promoting (either in a full-compliance world or in the actual world) the outcomes recommended by those principles (for example, an egalitarian distribution, or improvements in the lot of the poor). It goes without saying that the symbolic function of law need not go hand in hand with the causal efficacy of law. Indeed, symbolic behaviour may be counterproductive in terms of the symbolized values or principles.Footnote ⁶³

This is not the occasion to discuss consequence-insensitive theories about the relationships between law and morality. I just mentioned the ideal-theory and the expressive interpretations of the mirroring view as a word of caution about the scope of my own argument. The Trolley Problem is a problem for the project of deriving libertarian institutions from the LPP along with the mirroring view. The causal view, which accommodates our moral intuitions in killing/letting die cases, hardly furthers that project. Under the causal view, too many and strong redistributive measures would be permissible. Nothing follows, however, as to the moral standing of libertarian institutions, all things considered, including the (properly theorized) relationships between law and morality.Footnote ⁶⁴